1. Introduction

Surface plasmon polaritons (SPPs) [1–3] have been studied to improve the performance of light-emitting devices [4,5], high-resolution microscopy techniques [6], surface characterization methods [7], and sensors [8]. Recently, SPPs were shown to mediate energy transfer from optically-pumped dipoles across a thick metal film to fluorescent dye molecules [9]. Here we study the plasmon-mediated energy cascade issuing from electrically-pumped molecular excitons, which can enable new integrated sensing and imaging applications with high signal-to-noise ratio as well as new routes for performance improvement in solar cells [10,11], plasmonic devices [12], and organic LEDs.

Consider the archetypical organic light-emitting device (OLED) structure shown in Fig. 1(a) that consists of active organic semiconductors sandwiched between two electrodes, one of which is transparent [13]. In the most efficient OLED architectures, the organic layers compose a heterostructure that uses separate compounds to efficiently transport electrons and holes, and to localize and maximize radiative recombination. Because the injection layers usually are thin, the electron-hole recombination zone (where light is emitted) is within tens of nanometers of the injecting electrodes, resulting in strong and often parasitic waveguiding, as well as coupling of exciton radiative energy to non-radiative surface plasmons [14]. One approach to removing parasitic losses is to engineer the optical microcavity to place the node of the optical field inside one of the thin electrodes, using dielectric coatings on the external surface of the electrode [15,16]. This approach is based on optical interference within the multiple layers, and ultimately improves the optical transmission of the stack. Considerable energy, however, is trapped in the SPP modes. To recover some of this dipole energy, electrode texturing has been employed [4,17], which causes the plasmon dispersion curve to be moved inside the light cone [18], leading to radiative outcoupling of surface plasmon energy to propagating modes.

 

Fig. 1 (a) Illustration of the energy coupling of exciton dipoles created at the interface of hole and electron transport layers. An energy flux diagram is superimposed on the corresponding layer structure (orange-red shading), indicating energy flux pathways for a normalized in-plane wave vector (u = k_x / k₀). For u < 1, the exciton energy decays through leaky light emission, which is more easily transmitted through the ITO electrode than the thick metal electrode. Waves with u ≈1.63 are guided in-plane through the device layers; for u ≈2.24, the emitted field strongly couples to bound surface plasmon modes at the two metal/organic interfaces; for higher u values, the energy couples to non-radiative modes. For each mode, the out-of-plane electric field component is drawn. The leaky mode propagates in both directions, the waveguided mode is confined in organic and ITO layers, the surface plasmons are bound at the metal interfaces, and the non-radiative modes are highly confined inside the structure. (b) Cross-section of the device under study. To study the energy transfer across the thick metal film, we separately consider the five pathways by which light can propagate from the top of the device by combinations of leaky and SPP-mediated transport. Straight lines indicate radiative coupling and curved lines indicate non-radiative coupling. In pathway 5 (the focus of this work), energy couples from decaying dipoles into SPP modes, which then evanescently couple to the emissive dye in the capping layer near the metallic surface.

Download Full Size | PDF

Another approach to enhancing the outcoupling of energy trapped in the SPP modes is to engineer the device so that these modes resonantly couple to light-emitting dye molecules outside of the device (i.e. outside of the electrodes and the electrically pumped layers). The coupling of SPPs to dye molecules on the outer surface of the metal electrode can occur by two mechanisms: (A) evanescent coupling and (B) absorption of light that originates from SPP modes propagating in the metal electrode that are scattered by the roughness of the metal film. The reciprocal of mechanism (A), in which external dye layers are optically excited by solar radiation and couple to excitons within a device through a plasmon-mediated process in a metal electrode, has been studied in organic photovoltaic devices [10,19]. The measured energy transfer efficiency across the metal electrode was reported to be 46%.

2. Device design

The device we use to demonstrate the SPP-mediated transfer of electrically-pumped excitons by evanescent coupling [Fig. 1(b)] is based on a conventional OLED structure of ITO (150 nm) / α-NPD (50 nm) / Alq₃ (15 nm) / Ag (t_m nm) / α-NPD (t_c nm), where the final (capping) α-NPD layer contains a thin layer of an emissive dye (DCM2). The thickness of Alq₃ is reduced compared to conventional OLEDs in order to achieve better dipole-dipole energy transfer from the excitons to the metal. We separately consider the five pathways by which light can propagate from the top of the device, as shown in Fig. 1(b). In the leaky pathway (1), light energy is radiated by decaying dipoles in the direction of the metal electrode and leaks through this electrode and the capping layer. In the leaky dye excitation pathway (2), light that leaks through the electrode is absorbed and re-emitted by the emissive dye. In the SPP radiation pathway (3), energy couples from decaying dipoles into SPP modes, which then radiate light due to surface roughness (texture). In the SPP radiation dye excitation pathway (4), light emitted by SPPs due to surface roughness is absorbed and re-emitted by the emissive dye. In the energy transfer pathway (5), the focus of this work, energy couples from decaying dipoles into SPP modes, which then evanescently couple to the emissive dye in the capping layer near the metallic surface.

To maximize the fraction of energy flow through pathway 5, we first maximize the fraction of energy transferred into SPP modes by defining a figure of merit F = “SPP”/(“SPP” + “leaky”) and calculating the capping layer and metal electrode thicknesses that maximize F. We use a classical dipole model [20] and dyadic Green’s function (DGF) approach [21,22] for exciton decay rate and electric field distribution calculations, fullyconsidering dyadic eigenfunctions in the Green’s function formula [22]. The energy flux is then obtained using the Poynting vector, which is readily formulated using these dyadic functions [21]. We place the emissive dye layer within the capping layer 10 nm above the metal electrode (to capture the energy cascade behavior of pathway 5) and begin by examining the dependence of the exciton decay rate on in-plane wave vector and the thicknesses of the capping layer and metal electrode, as shown in Figs. 2(a) and 2(b). From Fig. 2(a) it is apparent that SPP modes on the interior and exterior of the metal electrode become increasingly coupled and approach a nearly equal in-plane wavevector as the capping layer thickness increases, due to the increasingly matched boundary conditions on either side of the metal electrode. A similar increase in SPP mode coupling occurs as the metal electrode thickness is increased, as shown in Fig. 2(b). This enhancement of the evanescent field is due to the negative real part of the complex permittivity of silver, and is consistent with prior work using thin films of silver in the context of superlensing [23–25].

 

Fig. 2 (a,b) Calculated decay rate (in log scale false color) as a function of in-plane wave vector (k_x) for exciton dipoles in the device structure: glass / ITO (150 nm) / α-NPD (50 nm) / Alq₃ (15 nm) / Ag (t_m nm) / α-NPD (t_c nm), where the metal (t_m nm) and capping layer (t_c nm) thicknesses are varied to optimize coupling to surface plasmons. As the capping layer and metal thicknesses increase, surface plasmons (near k_x/2π = 4) at the organic/metal and metal/capping surfaces start to couple, leading to increased electric field strength and efficient energy transfer from decaying excitons. (c) Rates of exciton energy coupling to leaky and surface plasmon modes as a function of capping layer thickness. Optical interference effects for leaky radiation with increasing capping layer thickness are characteristic of the general layer structure, as shown (upper graph) in calculations of the far-field transmittance of Alq₃ (20 nm) / metal / capping layer structures. (d) Similar calculation as a function of metal electrode thickness. (e) Three device structures designed to study energy transfer mechanisms, using an archetypal OLED structure as a common substrate, but observed from the top. Devices I and II have a 150 nm thick capping layer of α-NPD, leading to a strong electric field at both interior and exterior surfaces of the metal film (blue color electric field). DCM2, a red-emitting dye, is doped into the α-NPD capping layer with 5% mass ratio at different locations for Devices I and II. While the dye-doped region of Device I is adjacent to the metal electrode to promote resonant near-field energy transfer, in Device II it is far from the electrode. Device III is a bare OLED, with no capping layer, resulting in an SPP mode that is confined only at the interior of the metal electrode as shown in the (red) electric field.

Download Full Size | PDF

To calculate the strengths of the “leaky” and “SPP” modes that determine the figure of merit (F), we integrate the energy flux at the top of the metal electrode across the radiative and SPP ranges of wavevectors. The rate of energy coupling to leaky modes has a periodic dependence on the capping layer thickness due to optical interference effects, as shown in Fig. 2(c). In contrast, the rate of energy coupling to SPPs shows a monotonic rise with increasing capping layer thickness due to the increasingly matched boundary conditions. The rate of energy coupling to leaky modes exponentially decreases due to absorption losses as the metal electrode thickness (t_m) is increased, as shown in Fig. 2(d). For SPP modes, these losses compete with the field resonance, producing a maximum in energy coupling at an electrode thickness of 37 nm. Based on an iterative solution to these energy coupling calculations, we derive the thickness values (t_m and t_c) which optimize F: 65 nm for the metal electrode and 150 nm for the capping layer. Note that these values do not correspond to the maximum SPP coupling but rather the maximum fraction of SPP coupling, which yields the highest signal-to-noise ratio for our measurements and also for potential applications of the SPP-mediated energy transfer mechanism in microscopy and sensing.

Having maximized the fraction of energy transferred to SPPs, we now analyze the pathways (3-5) by which this energy in SPP modes can subsequently couple out, using a set of glass / ITO (anode) / α-NPD / Alq₃ / 65 nm Ag / α-NPD devices designed as shown in Fig. 2(e). For efficient injection of electrons into Alq₃, a thin interfacial layer (0.3 nm LiF / 0.6 nm Al) is inserted before the cathode layer. Devices I and II have an α-NPD capping layer thickness set at 150 nm to maximize F and support a relatively strong electric field (depicted in blue) at both the interior and exterior surfaces of the cathode caused by coupled surface plasmons. In Device I, a thin layer of the fluorescent dye DCM2 (doped at 5% concentration by mass, with an emission wavelength of λ = 630 nm) is inserted in the capping layer approximately 4 nm above the metal electrode. In Device II, the dye is inserted in the capping layer approximately 112 nm above the metal electrode. DCM2 is chosen because its absorption peak when doped into α-NPD (λ = 530 nm) is very close to the emission peak of Alq₃ (λ = 525 nm), therebyallowing SPP-assisted resonant coupling between electrically pumped excitons in the interior of the device and excitons on DCM2. Device III has no α-NPD capping layer, resulting in SPP modes that are confined to the electrode’s interior surface as depicted by the red-shaded electric field of Fig. 2(e).

Calculations were performed to predict the pathways for exciton energy decay in the above three devices. The dispersion diagrams for Devices I and II [which are almost identical regardless of dye position, and given in Fig. 3(a) ] and Device III [Fig. 3(b)] show that the interior and exterior surface plasmon modes in the metal electrode are only coupled appreciably in devices that have a capping layer. These coupled modes provide a path for energy transfer to the DCM2 dye molecules, as shown in the energy flux diagrams of Figs. 3(c) and 3(d). Because of the evanescent nature of the SPP field, this energy transfer path is much stronger when the dye layer is close to the electrode (Device I) than when it is further away (Device II). Since the surface plasmons in Device III are confined to the interior surface of the metal electrode and have no path for outcoupling, most of the energy coupled to them is dissipated as heat [Fig. 3(e)].

 

Fig. 3 Dispersion diagrams of energy coupling for Device I or II (a) and Device III (b). While surface plasmons are generated only at the interior (Alq₃/Ag) metal interface in Device III, coupled surface plasmons are created at both (Alq₃/Ag and Ag/α-NPD capping) metal interfaces in Devices I and II. (c), (d) and (e) show the energy flux (outside of the active organic region) through the layers of Devices I, II, and III, calculated from the Poynting vector, clearly indicating energy transfer from decaying electrically-pumped dipoles into leaky, waveguided, surface plasmon, and lossy modes. From (c) and (d) it is clear that a small amount of energy is transferred via coupled surface plasmons to the dye in Device II (in which the dye is far from the metal electrode), and a large amount of energy is transferred to the dye in Device I (in which the dye is near the metal electrode) due to near-field coupling. The scale bar in (c) is logarithmic.

Download Full Size | PDF

3. Experimental results

Devices I, II, and III were simultaneously deposited onto the same substrate, thereby eliminating fabrication variability and facilitating their comparison at identical operating conditions. Figures 4(a) and 4(b) show electroluminescence and emission spectra of the devices operating under forward bias (7 volts), imaged from the top. The spectra labeled “A”, “B”, and “C” correspond to the top emission of Devices I, II, and III, respectively. Spectrum “D” corresponds to light emission through the bottom (ITO side), and was nearly identical among the three devices.

 

Fig. 4 Measurement of energy transfer. a) Photograph, (b) Emission spectra, and (c) Emission diagram of the devices deposited onto a common substrate, operating under forward bias (7 V) and imaged from the top (silver electrode side). Top emission for Devices I, II, and III are labeled “A”, “B”, and “C, while bottom emission to the ITO side (which is nearly equal among the three devices) is labeled “D”. In Device I, evanescent SPP fields efficiently couple energy into the DCM2 dye through a resonant near-field process, leading to strong red dye emission (“A”). (d) Device II (“B”) shows a 6-fold increase in green light emission compared to the leaky emission of Device III (“C”) as the increased energy content in SPP modes at the exterior metal surface of Device II is scattered outward into propagating light modes due to metal film roughness. Atomic force microscopy (AFM) measurements of the topography of the Ag electrode shared by all three devices show an RMS roughness of 6 nm, which is sufficient to scatter out SPP modes to propagating light [5]. The good overlap of the scaled “C” spectrum with the “B” spectrum indicates that effects such as downconversion or wavelength-dependent scattering are relatively minor during the SPP-assisted emission process (pathway 3). However, even with this enhanced green emission (which is at the absorption peak of DCM2), very little red emission is observed in Device II. A comparison of the red (DCM2) emission in Devices I and II (in which both have been corrected by subtracting the spectral tail contribution of the green peak) shows a factor of 6.5 enhancement in Device I, significantly greater than the maximum predicted Purcell enhancement factor of 1.9, indicating the near-field transfer of energy to the dye in Device I through pathway 5.

Download Full Size | PDF

The emission of Device I (“A”) exhibits a significant peak in the red portion of the spectrum due to DCM2 dye emission. Because this peak occurs in the red and not the green, pathways 1 and 3 are ruled out as contributors. Furthermore, Devices I and II are expected to have nearly the same leaky and directly scattered SPP light intensity (even after accounting for microcavity effects), since they differ only in the location of the DCM2 layer. This is supported by DGF models of these devices. Dye photoluminescence due to the absorption of either leaky (pathway 2) or directly scattered SPP (pathway 4) light is therefore ruled out as a primary contributor to the much larger red peak seen in Device I. The fact that the red peak is much stronger when the dye is closer to the metal contact indicates a near-field (evanescent) energy transport mechanism, which is a signature of the remaining mechanism for energy transfer: pathway 5. We therefore ascribe this peak primarily to a resonant plasmon-assisted transfer of energy from decaying excitons in the device across the thick metal film to the DCM2 dye adjacent to the metal.

While thus far we have focused on using SPP modes in a metal film as a near-field pathway to couple energy into a fluorescent dye layer, it is also important to consider the possible contributions of these modes to the subsequent outcoupling of energy from the dye. In particular, it is well known that the decay rate of a dipole can significantly increase due to SPP coupling when it is in proximity to a metal film through the Purcell effect [20,26]. Using the DGF model discussed above, we calculate an increase of a factor of 1.9 in the DCM2 dye decay rate for Device I due to optical microcavity effects and energy transfer back to SPP modes in the cathode. While only a fraction of these SPPs will re-radiate due to metal electrode roughness, we view this as an upper limit on the enhancement of red emission due to the Purcell effect.

In order to compare the calculated factor of 1.9 enhancement of DCM2 emission via the Purcell effect to the enhancement of DCM2 emission observed in Device I, we first isolate the dye emission contributions to the spectra of Devices I and II. In Fig. 4(d), we plot the spectra of Device II (“B”), in which there is a capping layer with a DCM2 layer at a distance from the electrode, and Device III (“C”), in which there is no capping layer and no DCM2. Because the interior and exterior SPP modes in Device III are uncoupled as discussed above and provide a negligible pathway for emission, spectrum “C” is expected to primarily consist of dipole energy that leaks through the metal electrode. In Device II, green emission is significantly enhanced due to strong coupling to SPP modes that scatter to propagating light due to electrode roughness. Assuming that this scattering due to surface roughness is approximately wavelength-independent, we uniformly scale the “C” spectrum by a factor of 6 and compare it to “B” [Fig. 4(d)]. There is good overlap of the spectra in the green region, indicating that effects such as downconversion or wavelength-dependent scattering are relatively minor during the SPP-assisted emission process (pathway 3). In the red region, DCM2 emission that is visible in the “B” spectrum can be isolated by subtracting the spectral tail of the scaled “C” spectrum.

By similarly subtracting the spectral tail of the “C” spectrum scaled to the height of the green peak in “A”, the DCM2 (red) emission in Device I can be isolated, resulting in a measured intensity that is a factor of 6.5 higher than that in Device II. This compares favorably with calculations based on Figs. 3(c) and 3(d) that show the average evanescent field intensity in the dye layer of Device I to be 5 times stronger than the average field intensity in the dye layer of Device II. Since this measured enhancement of DCM2 emission is significantly larger than the upper limit of 1.9 calculated for the Purcell effect, we are led to conclude that the primary source of enhanced red emission is the plasmon-assisted dye pumping of pathway 5, rather than the increased coupling of red emission to re-radiating SPP modes in the electrode. This latter effect is expected to partially account for the small difference between the measured emission enhancement and the calculated field intensity enhancement.

To further support the assertion that energy transfer from decaying excitons by coupled surface plasmons is the main cause for the observed dye emission in Device I (as well as the enhanced green light emission in Device II), a further set of calculations and experiments were performed for devices incorporating an aluminum cathode instead of silver. Devices BI, BII, and BIII used a 30 nm Al cathode instead of 65 nm Ag, and 25 nm Alq₃ instead of 15 nm, all other layers being identical with Devices I, II, and III. The dispersion and energy flux diagrams of Figs. 5(a) and 5(b) indicate that no significant coupled surface plasmon mode exists for devices using the Al cathode, and consequently we expect no significant SPP-mediated transfer of energy to the external dye layer.

 

Fig. 5 For a device using an aluminum (rather than silver) cathode, the dispersion diagram (a) and energy flux diagram (b) predict that surface plasmon modes are uncoupled even when a capping layer is present. Surface plasmons excited at the interior interface of Al (Al/Alq₃) are confined to that interface, releasing only a tiny amount of energy into the exterior dye. This is confirmed by the observed emission spectra (c) and electroluminescence (inset) of control devices BI, BII, and BIII, which were made with 30 nm Al cathodes instead of 65 nm Ag, and 25 nm Alq₃ instead of 15 nm, all other layers being identical with Devices I, II, and III of Fig. 4. All control devices have a very similar green intensity, with only a faint red component observed from the DCM2 dyes in Devices BI and BII. Red emission from BII is presumably from PL of the dye, with some red emission from BI due to energy transfer via surface plasmons, albeit at much lower intensity compared to the devices that use a silver cathode.

Download Full Size | PDF

Electroluminescence and emission spectra shown in Fig. 5(c) for devices with Al cathodes exhibited no change in green emission intensity and only a weak increase in red emission for the devices with DCM2 dye layers (BI and BII). These measurements are consistent with the calculations presented above. Furthermore, the observation that green emission is not stronger in the BI and BII devices than the BIII device is consistent with the fact that coupling to SPP modes is weak for all devices using Al cathodes and, therefore, the outcoupling of SPP modes due to surface roughness scattering is weak as well.

In summary, surface plasmon mediated energy transfer of electrically-pumped excitonic energy across a thick metal film was demonstrated for the first time, showing a 6.5-fold enhancement of light emission from an external dye. The magnitude of this enhancement is not attributable to or attainable via optical microcavity effects or the Purcell effect. The demonstrated mechanism holds promise for improving top-emission in OLEDs in display and lighting applications. With an appropriate combination of dyes, white light emission could potentially be achieved by this mechanism. Furthermore, this mechanism could be used as an integrated electrically-pumped evanescent wave generator, with applications in very high resolution optical microscopy and lab-on-a-chip systems.