Correction to: Vibro-polaritonic ir emission in the strong coupling regime (ACS Photonics (2018) 5:1 (217-224) DOI: 10.1021/acsphotonics.7b00677)

The method of fitting the microcavity transmission spectrum to derive the cavity absorption spectrum by means of transfer-matrix analysis was flawed. Consequently, in contrast to our initial conclusions, the polaritonic emission appears thermalized and there is no evidence for polariton emission that is blue-shifted with respect to the polariton absorption. For this reason, the interpretation of the blue-shift as a signature of polariton-polariton interactions is obsolete. Two errors have been made: First, the dielectric function of PMMA was only modeled in the range from 1400 to 2100 cm^-1 (original Figure 2(b)). However, by virtue of the Kramers-Kronig relations, the relatively strong absorption lines at lower energies (cf. original Figure 2(a)) influence the refractive index in the vicinity of the strongly coupled C=O stretching band. Therefore, the simulated free spectral ranges between the cavity and the polariton modes were incorrect. Second, a small aperture was kept inside the Fourier transform infrared (FTIR) spectrometer transmission compartment for both emission and transmission measurements. As a consequence, only a small area of the prepared cavity was sampled. It was not checked which part of the emitting sample was imaged onto the aperture in the transmission compartment. Therefore, it is likely that different spots of the same sample were measured in transmission and emission, respectively. A film thickness inhomogeneity on the order of 10 nm (the polymer film was about 4 μm thick) can cause a spectral shift of the modes on the order of the initially derived polariton blue-shift (17 and 6 cm^-1) . Both errors added up in a way that the transmission spectrum fit yielded a match of emission maxima and reflectivity minima at the first and the third cavity mode, as expected from thermalized emission (cf. original Figure 3(b)). The emission maxima appeared however blue-shifted with respect to the reflection minima (cf. original Figure 3(d)). It should be noted that absorption maxima and reflection minima overlap. In an analogous manner to the procedure described in the original paper, we spin-coated an about 1 μm thick PMMA film on a 0.5 mm thick silicon substrate. We measured the sample transmission with an FTIR spectrometer and fitted it by transfer-matrix analysis while optimizing the model of the dielectric function of PMMA. The silicon substrate was treated as an incoherent layer. Its complex refractive index was taken from the literature1 and was in good agreement with experimental data. To approximate the dielectric function of PMMA, we started from the Lorentz-oscillator model used in the original paper. Next, we adjusted the imaginary part of the PMMA complex refractive index and finally applied Kramers- Kronig relations to obtain the real part of the refractive index.2 Figure 1 shows the measured transmission spectrum, the fit with the originally used simplified dielectric function, and the fit with the refined dielectric function. (Figure Presented). Consequently, we fitted again the cavity transmission and emission data originally published. The fit of the polaritonic transmission peaks is shown in Figure 2(a). The horizontal black bar denotes the spectral region that has previously been considered (original Figure 2(b)). Subsequently, the polaritonic peaks are modeled well by both dielectric functions. However, the transmission peaks of the first and the third cavity mode are only well matched by the refined dielectric function, whereas the simulation predicts blue-shifted peaks for the originally used function. Figure 2(b) shows the derived normalized absorption spectra and compares them to the experimentally determined emissivity. The use of the original dielectric function results in a good agreement between the first and the third cavity mode peak locations, but a larger discrepancy between the measured and simulated lower polariton emissivity peaks. This led to the original conclusion of nonthermal polariton emission. By contrast, the refined dielectric function shows a spectral offset not only for the (Figure Presented). polariton peaks but also for the peaks of the bare cavity modes; that is, the whole simulated absorption spectrum is red-shifted with respect to the experimentally determined emissivity spectrum. In Figure 2(c), this shift is corrected by enforcing the overlap of the third cavity mode absorption and emissivity peaks. This results in excellent agreement between the emissivity peaks and the absorption peaks modeled with the refined dielectric function. The original dielectric function, however, shows offsets between absorption and emissivity peaks of the first cavity mode as well as the polaritons. Finally, the corresponding transmission spectra are shown in Figure 2(d). They are computed by means of the cavity parameters used in Figure 2(c). The clear offsets between experimental and simulated third cavity mode peaks manifest that the (Figure Presented). originally measured transmittance and emissivity spectra were not directly comparable, as the locally sampled cavity lengths were different. To attain consistence among simulations, emissivity, and transmittance measurements, we prepared another Fabry- Perot cavity comparable to the one used to collect the original data. We used a ZnS substrate, sputter-coated an about 10 nm thin Au film on it, spin-coated an about 4 μm thick PMMA layer on top, and finally closed the cavity by sputter-coating another about 10 nm thin Au film. We removed the aperture from the transmission compartment of the FTIR to sample the same cavity area in emission and transmission. As a consequence, the signal-to-noise ratio in the emission measurement was reduced, and thus we performed an additional correction for atmospheric absorptions in the FTIR and we subtracted the reference spectrum from the signal of the sample and the experimental blackbody spectrum (cf. eq 13 of the original paper). Measurement results and simulations are shown in Figure 3. As in Figure 2(c) and (d), the peaks of the third cavity mode emissivity and simulated absorption were matched. Figure 3(a) exhibits again an overlap of simulated and measured polariton emission peaks if the refined dielectric function is used, whereas the original dielectric function predicts blue-shifted emission of the lower polariton. This is highlighted in Figure 4, where only the spectral range from 1400 to 2000 cm^-1 is considered, like in Figure 3(d) of the original publication. The absence of the aperture in the transmission compartment leads to a clearly improved agreement between measured and simulated transmittance spectra, as the comparison between Figures 2(d) and 3(b) demonstrates. The shift between measured and simulated third cavity mode as well as the line width difference is much smaller in the latter case. The refined dielectric function leads also to a good agreement between measured and simulated transmittance of the polariton modes, as Figure 4(b) shows. In addition to the funding acknowledged in the original publication, M.S. acknowledges support from the Marie Sklodowska-Curie actions of the European Commission (project 753228, PlaN).