We synthesized and used an emitter molecule 4,7-bis(7,12-bis(9,9-dioctyl-7-(trimethylsilyl)-9H-fluoren-2-yl)-5,5,10,10,15,15-hexahexyl-10,15-dihydro-5H-diindeno[1,2-a:1',2'-c]fluoren-2-yl)benzo[c][1,2,5]thiadiazole (MHeB14), Fig. 2 (a), which is a suitable material for use in OLEDs. Polymethyl methacrylate (PMMA) layers (∼ 200 nm thickness), which were either undoped or doped with nominal 1, 10 and 50% w/w of MHeB14, were spin coated onto two types of nanostructured substrates. The substrates, which have been described in detail elsewhere25, consist of ∼ 100 nm thick Au films, deposited on to polycarbonate templates which contained periodic square arrays of either left- or right-handed shuriken indentations. We have used two types of templated structures, they both contain identical shuriken indentations, but with periodicities of 1000 and 1500 nm respectively, Fig. 2(b). Subsequently, the two spin-coated periodic cavities will be referred to as 1000 nm and 1500 nm metafilms. As will be shown later, although the shuriken structures are identical, these substrates have dissimilar EM environments. Effectively they allow the EM environment to be modified without altering the structure of the nano-cavity, mitigating against possible influences of changing the nano-cavity on the polymer film structure.
For reference purposes we have compared luminescence and absorption spectra for solutions of MHeB14 with the equivalent data from PMMA doped films deposited on unstructured Au surfaces. This reference data, Fig. 2(c), demonstrates that the optical absorption and emission properties of MHeB14 immobilised in the PMMA are not significantly modified compared to that of the free molecule in solution, Fig. 2(c), which is indicative of an absence of molecular aggregate formation. The slight red shift of the luminescence compared to solution can be attributed to different dielectric environments and the reduced conformational flexibility of MBeH14 in the matrix. The MBeH14 does not display significant optical absorption for wavelengths > 450 nm, which is the region occupied by resonances associated with Au d→ sp inter-band transitions, Fig. 2(d), and plasmonic modes. Consequently, reflectance (linearly polarised) spectra collected from undoped and doped metafilms are only dominated by features associated with inter-band transitions (∼500–680 nm) and plasmonic resonances (> 680 nm), as shown in Supplementary Fig. 8.
The Figure 3(a) shows the schematic of the doped RH-metafilms. Circularly polarised luminescence spectra derived from detecting either left or right-handed light (LCP and RCP) are collected from both the undoped and doped metafilms. For clarity only the spectra for the 50% doping are discussed, other data can be found in supplementary information (Supplementary Figure 9), but all dopant levels display qualitatively similar behaviour. The luminescence from the undoped films is derived from the radiative decay of holes in the 5d band, which produces very weak emission (two orders of magnitude weaker than from the 1% doped PMMA films) spanning the blue to IR regions of the spectrum 26. The doped metafilms display a relatively larger amount of emission over the interband and plasmonic regions compare to the reference films on unstructured Au. A fraction of molecular luminescence from the doped metafilms displays helicity dependent structure, which is not observed in reference spectra in the absence of the nanocavities. In the case of doped 1000 nm metafilms, the luminescence showed two distinct regions of structure centred at ∼ 560 and 620 nm, figure 3(b, c), which are regions associated with d → sp transitions (at the L and X symmetry points), with the most intense and obvious structure occurring at the L region. For the L region, the structure observed for mis-matched combinations of metafilm handedness and emitted light helicity (LH/RCP and RH/LCP) is a single peak, which appear to split into two less intense components for matched combinations, figure 4, with the splitting being equivalent to ∼ 107 meV (Δλ∼ 28 nm). Relatively weaker structure occurs in the region of the X 5d-6sp transition, with a single peak observed for the mis-matched combinations and weaker less defined features observed for the matched combinations. For the 1500 nm metafilms, the structure is less pronounced and the most noticeable feature, a single peak, occurs close to the X 5d-6sp transition, while a weaker peak is observed in the vicinity of the L 4d-6sp transition. In contrast to 1000 nm periodicity, switching helicity does not result in the splitting of the peaks, but rather causes only a change in relative intensities of the structure. As with the 1000 nm case, the most intense structured luminescence is observed for the matched combinations of metafilm handedness and light helicity. For both periodicities the positions of the structured emissions do not significantly change with increasing level of doping. This is consistent with the structures being associated with inter-band transitions, which unlike plasmonic resonances, are relatively insensitive to the refractive index of the surrounding environment.
Circularly polarized extinction spectra were derived by monitoring the intensity of circularly polarised scattered light generated by linear polarised incident beam, from doped and undoped metafilms. As shown in Supplementary Fig. 10, these spectra are dominated by features associated with plasmonic and X d → sp resonances for both 1000 and 1500 nm metafilms, and as expected identical spectra are observed for symmetry equivalent combinations of light helicity and nanostructure handedness. Consistent with previous observations the L d → sp resonance makes a far weaker contribution to the overall extinction spectrum27, making it difficult to discern splitting of this band.
While splitting of peaks in extinction spectra can arise through mechanisms other than strong coupling, e.g. Fano interference28, equivalent splitting in luminescence is a definitive fingerprint of strong coupling16. This is because PL is an incoherent process, in contrast to light scattering, and thus does not display Fano interference29.
Two significant inferences can be drawn from the experimental data about the nature of the EM environments within the cavities that define the strength of the coupling. The absence of strong coupling in the 1500 nm cavity indicates relatively weak electric fields compared to those for the 1000 nm. Most significantly, the intensities of the E fields in the 1000 nm cavity have a strong dependency on both cavity handedness and emission helicity. These phenomenological arguments can be justified by using EM numerical simulations to calculate the properties of the near fields, with the caveat that EM numerical simulation cannot fully describe excitation of inter-band transitions. This is because quantum mechanics is required to rationalise the transition between two electronic states and cannot be accounted for with classical Maxwell’s equations 30, 31. However, EM simulations can provide a guide to the nature of the near fields generated by light scattering, albeit they will overall underestimate the strength of the fields, since the effects of the inter-band transitions are not taken into consideration. Near field electric field enhancement\({(\left|E\right|}^{2}/{\left|{E}_{ref}\right|}^{2})\) for LCP and RCP light for both enantiomorphs of 1000 and 1500 nm metafilms have been calculated and compared to experimental luminescence spectra, figure 3, where \(\left|E\right| \text{a}\text{n}\text{d} \left|{E}_{ref}\right|\) are the volume averaged electric fields in the polymer films surrounding the nanocavities and an unstructured Au surface respectively. In addition, the level of chiral asymmetries of near fields have been parametrised with the optical chirality factor (C) 32, 33. For the 1000 nm substrate there are peaks in \({\left|E\right|}^{2}/{\left|{E}_{ref}\right|}^{2}\) which overlap with the regions of enhanced structured emission. The peaks in the simulate spectra are slightly blue shifted with respect to the experimental data, which is a common feature of numerical simulations, and can be attributed to the idealised model used in the simulation not accounting for the surface roughness of real structures34. Consistent with the proposal that splitting of the luminescence structure is due to strong coupling, the largest peaks in \({\left|E\right|}^{2}/{\left|{E}_{ref}\right|}^{2}\) occur for matched combinations of light helicity and nanocavity handedness (i.e. LCP/LH and RCP/RH), figure 4. The greater intensities of the single luminescence bands observed for the weakly coupled mis-matched combinations can be attributed to the Purcell effect amplifying emission35. For the 1500 nm substrates the peaks in \({\left|E\right|}^{2}/{\left|{E}_{ref}\right|}^{2}\) also overlap with the peak of enhanced emission. However, the peaks in \({\left|E\right|}^{2}/{\left|{E}_{ref}\right|}^{2}\) are smaller than for the 1000 nm substrates, which is consistent with the weaker nature of the coupling, implied by the absence of peak splitting. Hence, for this weak coupling regime only the Purcell effect operates leading to enhanced emission coinciding with the peaks in the \({\left|E\right|}^{2}/{\left|{E}_{ref}\right|}^{2}\).
A comparison is made between dichroic spectra derived from the scattering and luminescence of circularly polarised light, Fig. 5, which in accordance with literature are plotted in terms of asymmetry factor (g) and will be referred to as CD and CPL spectra respectively. The spectra provide a means from discriminating between weak and stronger forms of coupling. In the weak coupling regime, such as that observed in the present work in the absence of the emitter, and in previous studies of emitter and chiral nanostructures 11, 12 there is a direct correlation between the CPL and CD spectra in terms of the position and sign of resonances. This is because under the weak coupling regime both the CD and CPL responses are governed by the C of the near field environments36. For the 1500 nm metafilm there is a correlation between CPL and CD spectra for all emitter doping concentrations, in agreement with the assignment of weak coupling. In contrast, there is a clear disparity between the CPL and CP spectra in the region covered by the L and X inter-band transitions at all dopant levels in the 1000 nm metafilm, consistent with stronger coupling. At longer wavelength, the region where plasmonic modes dominate, the level of correlation between CD and CPL is dependent on dopant concentration. For the 1% there is a correlation between CD and CPL, which is absent for higher dopant levels. The dependence of strong coupling on emitter concentration is well established37. CPL and CD data, which display similar qualitative behaviour, obtained for Rhodamine 6G in PMMA on related chiral nanocavities is displayed in supplementary information. This points to the coupling beyond the weak regime is not molecule specific.
Thus, using a luminescence spectral signature, band splitting, of strong coupling, supported by numerical simulations we demonstrate the control of strong coupling based on the symmetry combination of nanocavity handedness of light helicity. Strong coupling is achieved for Left Nanocavity / LCP and Right Nanocavity / RCP combinations, and therefore these can be considered to be left and right EM-enantiomers respectively.