Observation of Mode Splitting in Photoluminescence of Individual

Dec 22, 2016 - Department of Astronautical Engineering, University of Turkish Aeronautical .... James T. Hugall , Anshuman Singh , and Niek F. van Hul...
0 downloads 0 Views 2MB Size
Letter pubs.acs.org/NanoLett

Observation of Mode Splitting in Photoluminescence of Individual Plasmonic Nanoparticles Strongly Coupled to Molecular Excitons Martin Wersal̈ l,† Jorge Cuadra,† Tomasz J. Antosiewicz,*,†,‡ Sinan Balci,§ and Timur Shegai*,† †

Department of Physics, Chalmers University of Technology, 412 96 Göteborg, Sweden Centre of New Technologies, University of Warsaw, Banacha 2c, 02-097 Warsaw, Poland § Department of Astronautical Engineering, University of Turkish Aeronautical Association, 06790 Ankara, Turkey ‡

S Supporting Information *

ABSTRACT: Plasmon−exciton interactions are important for many prominent spectroscopic applications such as surfaceenhanced Raman scattering, plasmon-mediated fluorescence, nanoscale lasing, and strong coupling. The case of strong coupling is analogous to quantum optical effects studied in solid state and atomic systems previously. In plasmonics, similar observations have been almost exclusively made in elastic scattering experiments; however, the interpretation of these experiments is often cumbersome. Here, we demonstrate mode splitting not only in scattering, but also in photoluminescence of individual hybrid nanosystems, which manifests a direct proof of strong coupling in plasmon−exciton nanoparticles. We achieved these results due to saturation of the mode volume with molecular J-aggregates, which resulted in splitting up to 400 meV, that is, ∼20% of the resonance energy. We analyzed the correlation between scattering and photoluminescence and found that splitting in photoluminescence is considerably less than that in scattering. Moreover, we found that splitting in both photoluminescence and scattering signals increased upon cooling to cryogenic temperatures. These findings improve our understanding of strong coupling phenomena in plasmonics. KEYWORDS: Strong coupling, Rabi splitting, plasmon−exciton interactions, photoluminescence, plexciton

P

plasmonic nanoparticles and individual quantum emitters. Observation of individual plasmonic nanoparticles in turn requires an appropriate spectroscopic technique, which in practice almost exclusively reduces to some variant of dark-field (DF) microscopy. However, strong coupling data obtained via DF scattering is sometimes difficult to interpret, as these might be confused with other effects such as enhanced absorption, plasmonic quenching,33−35 Fano resonances in the weak coupling regime,36,37 or various polarization and symmetry artifacts. These subtle effects are different from strong coupling in their physical nature but may still result in similar experimental observations, namely asymmetry or split-like line shape of spectroscopic signal.15,34,35,37 The absence of alternative experimental techniques combined with simplicity and robustness of DF explains the popularity of the method; however, at the same time makes it non trivial to draw conclusions on the nature of plasmon−exciton interactions. In part, this is because plasmonic cavities, as opposed to their photonic counterparts, are multimode and lossy.38−41 An additional complication is a lack of tunability in plasmonics, which makes it difficult to

lasmon−exciton interactions and in particular strong coupling phenomena in plasmonics have recently attracted significant research interest, motivated mainly by room temperature performance and deeply subwavelength confinement of electromagnetic modes.1−5 Strong coupling is a nonperturbative regime of light−matter interactions in which the coupling strength dominates over any dissipative processes within the system.5,6 To achieve strong coupling, two requirements need to be satisfied: strong mode confinement and strong transitional dipole moment μe in accordance with g = μe|Evac|∝ μe/√V, where g is the coupling strength, Evac is the vacuum field, and V is the mode volume.5,6 The former is achieved via plasmonic nanoparticles, while the latter is provided by a certain class of materials possessing a high transition dipole moment of the electronic excitation, such as J-aggregates,7−16 quantum dots,17,18 perovskites,19 and most recently 2D transition metal dichalcogenides.20,21 Demonstration of strong plasmon−exciton coupling has been shown using surface lattice resonances,22−24 propagating surface plasmons,9,12 Fabry-Pérot (FP) cavities,7,8,11,15,25,26 and localized surface plasmons.27−29 Dynamic observations of Rabi oscillations in the strong plasmon−exciton coupling regime have also been demonstrated.30−32 Studying quantum optical phenomena with hybrid plasmon− exciton systems naturally requires experiments on individual © 2016 American Chemical Society

Received: November 8, 2016 Revised: December 16, 2016 Published: December 22, 2016 551

DOI: 10.1021/acs.nanolett.6b04659 Nano Lett. 2017, 17, 551−558

Letter

Nano Letters experimentally verify anticrossing behavior using one and the same hybrid nanostructure. Previously, several works have demonstrated strong coupling on a single nanoparticle level,27−29 and claims have been made even at the single nanoparticle−single quantum dot/molecule level.42−44 These works, however, utilized DF rather than photoluminescence (PL) for strong coupling investigations, which is in sharp contrast to past established works in solid state and atomic physics where strong coupling in PL has been demonstrated in the quantum regime.6,45 Although in our previous studies on a single particle level significant broadening of PL was observed, which could be interpreted as signs of strong interactions, no obvious splitting was observed.28,29 In bigger systems, such as thin metallic films and FP cavities, fluorescence originating only from the lower polariton (LP) branch was reported, as the upper polariton (UP) is inactive at room temperature due to the fast nonradiative energy transfer to uncoupled incoherent states.9,11,46,47 Although some recent data suggest observation of complex spectra in PL of nanoparticle solutions on ensemble level,18,48 PL splitting on a single particle level has not been reported. Here, we demonstrate splitting in PL of strongly coupled plasmon−exciton hybrids at an individual nanoparticle level. We achieve this by fully saturating the mode volume, which results in DF scattering splitting of up to 400 meV, which approaches the ultrastrong coupling regime.16,49 The mode splitting observed in PL is substantially smaller (nearly 2-fold) than the obtained DF values. We study hybrid plexciton systems in both DF and PL as a function of temperature in vacuum. In addition, we perform a polarization-resolved study, which shows a nearly isotropic response, suggesting random orientation of J-aggregates with respect to the local field. The data are highly consistent and reproducible, which is in line with the near complete mode volume saturation with J-aggregates observed in SEM. This result is the first demonstration of the strong coupling regime observed in the PL of individual nanoparticles, which is analogous to earlier works in solid state and atomic systems.45,50,51 One thus could envision the emergence of quantum optical applications such as nonlinearities at the single photon level and quantum optics using plasmonics at room temperature in similar plasmonic systems.4,22,52,53 The spectral scattering data originating from single bare Ag nanoprisms and a single plexcitonic particle are displayed in Figure 1. SEM shows that in both cases the spectra originate from individual triangular nanoprisms. The scattering spectrum reveals a massive splitting (∼400 meV) that indicates that the structure is deep in the strong coupling regime. This type of spectral change is not possible to attribute to other DF artifacts, as morphological analysis performed by SEM shows that Ag prisms are isolated nanostructures and are completely embedded in the J-aggregate matrix (gray regions around the Ag nanoparticle, see inset in Figure 1). For comparison, bare Ag nanoprisms are not surrounded by a corresponding gray region. Hyperspectral imaging signals in this study were obtained using a liquid crystal filter and an EM-CCD detector (see Methods). The liquid crystal filter plays a role of a linear polarizer, which allows studying polarization dependence of the optical signal. By performing the measurements using two different orthogonal orientations of the filter, we recorded polarization resolved data. Most of the structures displayed a nearly isotropic response, which is consistent with a homogeneous distribution of TDBC molecules around the Ag nanoprisms in a core−shell geometry. This was confirmed by SEM

Figure 1. Dark-field scattering spectra of an uncoupled plasmonic Ag nanoprism (cyan) and a coupled single nanoprism plexciton structure (blue). Insets display SEM images of the nanoparticles imaged on top of Au surface to visualize J-aggregate shell (right) as well as graphical sketches of the nanoprisms with J-aggregates shown in red. The splitting reaches nearly 400 meV, a value significantly above the fullwidth at half-maximum of an uncoupled nanoprism.

image visualizing the distribution of dye molecules around the Ag prism in Figure 1. However, several symmetric triangular nanostructures exhibited a nonisotropic response, which we attribute to a nonhomogenous distribution of molecules around the plasmonic structures. Indeed, a polarization dependent coupling constant may arise from orientation of the transition dipoles with respect to vacuum field, gi = |μ⃗ e||E⃗ vac(ri)| cos θi, where θi denotes the angle between the ith J-aggregate transition dipole moment and vacuum electric field vectors and E⃗ vac(ri), the vacuum field at the position of the ith J-aggregate (see Supporting Information Figures S2−3). The collective coupling constant gtot is given ⃗ | ⟨cosθi 2⟩N b y g = g 2 + g 2 + ... + g 2 = N |μ ⃗ ||Evac tot

1

2

N

e

and thus depends on the local molecular orientation with respect to the vacuum field. Hence, polarization-resolved data allow us to conclude that this orientation is random and homogeneously averaged out over many possible orientations for a majority of nanoparticles. We observe several interesting properties in the investigated hybrid systems. The splitting overall is significantly greater than previously reported for similar structures,29 which supports the hypothesis of mode volume saturation with J-aggregates. Furthermore, the size of the nanoparticles here is smaller than in our previous studies,29 which means that the mode volume is further compressed, which in turn explains a higher coupling strength in this case in accordance with g tot ∝ N /V , since for small single particles V approaches the (decreasing) geometrical volume.39 Having studied the elastic optical response of the coupled structures, we now turn to PL. The PL signal originating from an individual plexcitonic nanoprism at room (293 K) and liquid He temperatures (4.5 K) is shown in Figure 2a. Corresponding DF and SEM data are also displayed. Note that room temperature PL was recorded both prior to and after the cooling to ensure that photodegradation issues play no role in the temperature-dependent analysis (see Figure 2a, blue and red solid lines). The data show observations of splitting both in DF 552

DOI: 10.1021/acs.nanolett.6b04659 Nano Lett. 2017, 17, 551−558

Letter

Nano Letters

Figure 2. (a) PL of a single hybrid nanoprism at room and liquid helium temperatures. Inset shows a SEM image of the nanoparticle (scale bar is 50 nm); (b) dark-field scattering recorded at room temperature for the same nanoprism as in panel a. The green vertical dotted line shows the position of the excitation laser. (c) PL signal recorded at room and liquid helium temperatures (inset: SEM scale bar is 50 nm); (d) dark-field scattering recorded at room and liquid helium temperatures for the same nanoprism as in panel c. Black vertical dotted lines in all panels show the position of bare J-aggregate emission (λ = 588 nm). In the shown examples, nanoparticles’ plasmon resonances were approximately degenerated with the J-aggregate resonance at room temperature, that is, ωpl ≈ ω0. Smooth Lorentzian lines below each PL spectrum are obtained through fitting of the data to the coupled oscillator model.

both DF and PL were measured as a function of temperature (Figure 2c,d). Importantly, here we observe that splitting in DF and position of the plasmon resonance are also temperaturedependent (Figure 2d). This implies that splitting in PL is not solely due to temperature-dependent q, but also due to stronger coupling at low temperature. It is known from the literature that plasmon resonance is weakly temperature-dependent due to insignificant electron−phonon contribution to the total Ohmic loss at optical frequencies.57,58 Thus, stronger DF splitting at low temperature is likely due to excitons. In particular, μe in J-aggregates is temperature dependent.54,55 It is important to note that in addition to stronger coupling we also observe a red shift in plasmon resonance frequency upon cooling (Figure 2d). This is in line with earlier works on individual gold nanostructures.57 This red shift correlates with red shift in the lower energy PL peak and thus results in observation of stronger splitting in PL at low temperature (high energy PL peak is relatively temperature independent). We further report statistics on all the data measured in this study. Mode splittings shown in Figure 3a were extracted using the coupled oscillator model that takes losses into account. The splitting between the polariton branches

scattering and PL at cryogenic temperatures (Figure 2a,b), which we attribute to plasmon−exciton hybridization observed in PL of individual plexcitonic nanostructures. However, the splitting in PL at both room and low temperatures is significantly smaller than splitting in scattering for the same nanostructure. Moreover, we observe that the PL signal depends on the temperature. At room temperature, we observe significant PL broadening (in comparison to bare J-aggregate’s PL), but only small splitting, while at low temperature, PL splitting is increased. Note also that the higher energy peak in the PL signal is close to the uncoupled molecules, which indicates that it may be not due to emission from the upper polariton but rather due to uncoupled molecules. We return to this point in the Discussion section. There can be several reasons behind these temperaturedependent observations. One possibility is increased exciton delocalization in J-aggregates at low temperatures. Indeed, at room temperature, the exciton in TDBC J-aggregates is delocalized over ∼15 monomer molecules,54 while at low temperature, the delocalization is increased to ∼30−45 molecules.55 Another possibility is that the quantum yield q of the upper polariton branch is highly temperature dependent, which was demonstrated in various microcavity systems.8,9,46,47,56 To understand the reasons for stronger PL splitting at low temperatures, we performed additional experiments in which

is given by ω+ − ω− =

4g 2 + δ 2 − (γpl − γ0)2 , where g is

the total coupling strength, δ is the detuning between plasmon and exciton resonance frequencies, that is, δ = ωpl − ω0, 553

DOI: 10.1021/acs.nanolett.6b04659 Nano Lett. 2017, 17, 551−558

Letter

Nano Letters

Figure 3. (a) Distribution of mode splitting in DF and PL (at T = 4.5 and 293 K) as a function nanoprisms’ side length L. Note that a typical value for γpl ≈ 150 meV, while γ0 ≈ 100 meV. (b) Anticrossing in DF and PL (at T = 4.5 and 293 K) as a function of detuning. Red and blue dashed lines show plasmon and exciton frequency correspondingly. Green dotted line shows 532 nm laser excitation. (c) Blue shift between PL and DF signal of DF the lower polariton branch, Δ = ωPL − − ω− , shown as a function of splitting in scattering. Blue and red dashed lines are linear fits of the data and serve as guides for an eye.

and γpl,0 are the plasmon and exciton line widths correspondingly. At first, we extracted resonance positions of upper and lower polaritons, ω+ and ω−. Further, by assuming that ω+ + ω− = ωpl + ω0 and noting that the exciton position is always fixed at ω0 ≈ 2.11 eV, we calculated the detuning and then estimated splitting at zero detuning as Ω R =

particles inside the cryostat, see Methods). We return to these observations in the Discussion section. The nature of the higher energy resonances in PL is not entirely clear based on this data. The high-energy peaks may be due to uncoupled molecules46,59 or dark polaritons60 and not due to UP emission as was suggested previously.48 The fact that its dispersion is essentially flat speaks in favor of the former, despite being slightly blue-shifted with respect to uncoupled molecules. In such a case the mode splitting in PL calculated earlier does not fully account for the energy difference between the upper and lower polariton states, thus explaining a nearly two-fold difference between splitting in scattering and PL (see Figure 3a). The line width of the higher energy resonance is also slightly reduced upon cooling−from 107 ± 11 meV at 293 K to 99 ± 10 meV at 4 K, which is in agreement with free J-aggregates temperature induced narrowing. Note that the fluorescence filter cube used in our measurements limits spectral range of observations to wavelength longer than ∼545 nm, which in turn limits the detuning range in PL measurements (see Figure 3b). Note also that because of the unclear nature of the high energy PL peak, estimation of the ωpl and correspondingly δ = ωpl − ω0 = ω+ + ω− − 2ω0 in case of PL may be inaccurate. To elucidate on this issue, we plot δPL versus δDF for the same particles in Figure S4 (see Supporting Information). We find that the offset typically does not exceed ∼50 meV, which implies that the interpretation provided in Figure 3b is valid. To summarize, we observe temperature dependent mode splitting in PL. In all measured cases, splitting in PL at low temperature was always stronger than at room temperature. Moreover, splitting in PL was always smaller than in DF. These observations are consistent and reproducible. We further discuss these observations, as well as several related issues, such as relaxation within the excited states, polariton lifetime, and quantum yield.

4g 2 − (γpl − γ0)2 = 2 (ω+ − ω0)(ω0 − ω−) . In

this manner, we extracted ΩR for both DF and PL signals. We observed that splitting is almost independent of the particle size, thus supporting the hypothesis of full saturation of the mode volume with molecules (Figure 3a). This contrasts with our previous observations, where saturation of Rabi splitting was not reached.29 Mode splitting in DF and PL as a function of detuning is shown in Figure 3b. The data exhibit an anticrossing behavior in both cases; however, splitting in DF is always greater than in PL. The lower energy peak in PL displays dependence on the detuning based on which we attribute it to the lower polariton branch. The average line width of the lower energy peak is slightly reduced upon cooling, from 104 ± 16 meV at 293 K to 88 ± 20 meV at 4 K. Note that the lower energy PL peaks are blue-shifted with respect to the corresponding scattering peaks. This may be due to emission from not fully thermally relaxed lower polariton states as well as due to differences in mode splittings of different experimental observables that we discuss further. In Figure 3c, we plot this blue shift in terms of Δ = ωPL − − ωDF − . We observe a clear correlation between Δ and splitting in scattering, which suggests that the blue shift may totally disappear at lower coupling strengths. Moreover, Δ depends on temperature such that the blue shift is higher at room temperature. This may be related to the red shift of ωpl upon cooling,57 as Δ at both room and liquid helium temperatures was calculated using the room temperature value of ωDF − (T = 293 K) (due to complexity of DF measurements of small 554

DOI: 10.1021/acs.nanolett.6b04659 Nano Lett. 2017, 17, 551−558

Letter

Nano Letters Discussion. Splitting in Scattering versus Photoluminescence. We first turn to the question of why splitting in PL is smaller than splitting in DF. To shed more light on this problem, we performed numerical calculations of plasmon−exciton coupling (see Methods). Indeed, Figure 4 shows splitting both

measuring the actual structure of plasmon−exciton polariton energy levels. In our previous works, splitting in scattering, extinction, and absorption have been shown to differ from one another, such that splitting in scattering was typically stronger than that in absorption.28,34,35,37 Observations in Figure 4 are consistent with these earlier results. Hence, we conclude that mode splittings for different experimental observables may be different. A simple explanation for this can be inferred based on the coupled oscillator model.36 However, in addition to these difficulties, there may be other reasons, related to excited state dynamics, that further complicate the picture in case of PL. We consider these further. Photoluminescence Mechanism. When N J-aggregates strongly couple to a plasmon field, two bright, |+> and |−>, and (N − 1) dark polariton states arise. Although these dark states are weakly absorbing, they may contribute to the PL, as was suggested recently.60 In addition, an ensemble of incoherent uncoupled molecules may be present in our structures due to random molecular orientations and thick dye overlayers spanning beyond the plasmon near-field (see SEM image in Figure 1 inset). We also note that strong coupling to dark plasmonic states, that is, quenching, can play an important role here, especially for the molecules in the nearest vicinity of the metal surface.63 Additionally, exciton−exciton interactions in high density molecular layers may lead to self-quenching.61 With this complex structure in mind, we start discussing the mechanism of PL in the strongly coupled plexcitons. For PL to occur, absorption of a laser photon must be followed by some inelastic nonradiative energy loss mechanism(s). These processes have been studied by Agranovich and co-workers for the case of strongly coupled organic microcavities filled with J-aggregates.46,59 Following these works, the dynamics in the present case can be expected to proceed as follows. At first the hybrid system is excited to the vibronic polariton state by absorbing a 532 nm laser photon |g, 0 > ν=0 → |+ > ν≥0 (here, ν denotes vibrational quantum number along some vibrational normal mode). Excitation is then followed by a rapid relaxation (∼50 fs) to uncoupled molecules,46 where the population can reside for a relatively long time (∼1−10 ps).59 An alternative route of direct relaxation to the lower polariton branch |+ > ν≥0 → |− > ν≥0 was estimated to occur on pico second time scale and is thus inefficient.64 The subsequent dynamics strongly depends on the relation between the Rabi splitting and the vibrational modes of the molecules. In particular, if Rabi splitting exceeds the phonon energy, in addition to standard photoluminescence of uncoupled molecules, there appears a relaxation pathway consisting of nonradiative transition to the lower polariton branch accompanied by emission of a high-energy intramolecular phonon.59 In the present case, < 180 meV,28,65 while ΩR > 300 meV, so ΩR ≈ 2ωJ−agg ωJ−agg vib vib , which implies that this pathway is indeed viable. Both processes, namely direct PL from the uncoupled molecules and population of lower polariton branch, give rise to PL. On the basis of the arguments above, as well as on data shown in Figures 2 and 3, we assign the high-energy PL peak to emission from uncoupled molecules, Process 1, while the lowenergy PL peak to emission from the lower polariton branch |−>, Process 2, correspondingly. We note that the PL of the uncoupled molecules is modified with respect to the free space emission due to interaction with the plasmon (Purcell effect); however, keeping in mind that relaxation of J-aggregates is typically dictated by fast nonradiative processes,55 this modification is expected to be small. Both Processes (1−2) occur on

Figure 4. FDTD calculations of the optical response of the coupled silver nanoprism J-aggregate system. Silver core is homogeneously covered by a finite molecular (TDBC) shell (thickness 3 nm). The nanoprism side length and the thickness were taken to be 40 nm while 11 nm correspondingly, in agreement with HR-TEM data. The corners were rounded (r = 6 nm). Vertical dashed lines show the position of upper and lower resonances for total absorption (blue), scattering (purple), and J-aggregate absorption (orange); J-aggregate resonance at 588 nm (black). Evidently, splitting in scattering is greater than splitting in total absorption, which in turn is greater than splitting in absorption of the J-aggregate layer.

in absorption and scattering, indicative of the system being in the strong coupling regime. Furthermore, we are able to separate the absorption in the coupled system into absorption in the dye and in the silver nanoparticle contributions. Both of these contributions exhibit splitting, although the former shows a smaller one. Additionally, a peak/shoulder at 588 nm is observed, which results from absorption in uncoupled dye. The similarity between the relatively weak splitting of absorption in the dye and the PL signal (when compared to elastic scattering) is striking. To observe the PL signal, the energy first needs to be absorbed by the dye. Assuming that in the coupled system the Stokes shift is also close to zero as it is for the free J-aggregate,61 one can envision how the PL spectrum should look alike. Hence, one could expect that the splitting in PL would be smaller than that of the elastic scattering as shown in Figure 4. To elucidate on the issue of the uncoupled molecules, we have performed additional FDTD calculations, in which molecular aggregate shell thickness was systematically varied (see Figure S5). These calculations show that contribution of uncoupled molecules to both absorption and scattering of strongly coupled systems become significant in thick molecular shells. However, for thinner shells, molecules mostly contribute through the collective strong coupling. In addition to FDTD data, previous works on strong coupling in quantum well−semiconductor microcavity systems suggested that different quantities, such as absorption, transmission, reflection, and PL, all exhibit different values of observable mode splittings.62 Moreover, all these splittings are actually different from the Rabi splitting itself, Ω R =

4g 2 − (γpl − γ0)2 .

This raises an important practical question about the means of 555

DOI: 10.1021/acs.nanolett.6b04659 Nano Lett. 2017, 17, 551−558

Letter

Nano Letters

understanding of the dynamics in single strongly coupled plasmon−exciton structures. Our measurements ensure that the spectral modes indeed arise from plasmon−exciton interactions rather than from any ensemble related issues such as complexity in particle clusters or structures with odd morphology. Our experimental approach allows for a careful correlation between scattering and photoluminescence spectra with morphological details of each individual nanoparticle. The system parameters in terms of particle size, J-aggregate shell characteristics, temperature, and vacuum were reproducible, controllable and stable, which greatly simplified the analysis. We expect that future studies could focus on systematic variation of the splitting with respect to phonon energy, excitation spectra, temperature effects, quenching, lifetime, and quantum yield measurements. Also proper theoretical understanding of strongly coupled plasmon−exciton systems is required. This work is the first important step in confirming strong coupling behavior on the single plexciton particle level in photoluminescence, which opens up prospects for future realizations of quantum plasmonics systems provided the mode volume is further compressed. Methods. Synthesis of Plasmon−Exciton Hybrids. We prepared the samples following ref 16. Briefly, the Ag nanoprisms with precise control over the prism edge length were wet-chemically synthesized by using a seed-mediated protocol. High-resolution TEM shows that Ag nanoprisms are high quality single crystalline nanostructures (dimensions: ca. 50 nm side length and 10 nm height), which is essential for minimizing Ohmic losses (see Supporting Information Figure S1). At first, small seed Ag nanoparticles were synthesized, which was then followed by a slower growth of nanoprisms. Plasmon−exciton hybrid nanoparticles were synthesized by self-assembly of a J-aggregate dye (TDBC, 5,5′,6,6′-tetrachloro-di(4-sulfobutyl) benzimidazolocarbocyanine, FEW Chemicals) on Ag nanoprism surfaces. The color of the solution changed immediately indicating plexcitonic nanoparticle formation. The excess TDBC molecules were removed by centrifugation. This allowed saturation of the mode volume of the nanoparticles with the dye and to stabilize J-aggregates against degradation, which resulted in formation of core−shell Ag J-aggregate nanostructures. Optical Measurements. Hybrid nanoparticles were immobilized on the oxidized Si substrate and analyzed using an upright microscope. To achieve an appropriate density of nanostructures, a drop of nanoparticle solution was applied on a substrate precoated with polylysine (0.25 mg/mL). The droplet was subsequently removed by an intense nitrogen flow after remaining on the substrate for about 2 min. The substrate contains a lithographically defined square pattern (100 × 100 μm2), which enables spectroscopic and morphological correlation. DF scattering measurements were performed using a hyper-spectral imaging technique based on a liquid crystal tunable filter.29 This allows for a high throughput parallel recording of DF spectra for all structures within the field of view. All room temperature measurements were performed using a reflective DF objective (Nikon, 100× NA = 0.8) in an upright microscope. All low-temperature measurements were performed in the optical cryostat and a long working distance objective (Nikon, 20× NA = 0.45) in an upright microscope. We note that these low-temperature measurements allowed the signal only from the biggest nanoprisms to be recorded with sufficient signal-to-noise ratio.

a pico second time scale, that is, much slower than the polariton lifetime (∼10 fs in accordance with γ± ≈ (γpl + γ0)/2) and thus represent the bottleneck of the PL in strongly coupled systems. Such picture agrees with the lifetime measurements in strongly coupled microcavities at room temperature reporting τ = 1−3 ps.66,67 An alternative view involves coherent dark polariton states in addition to incoherent uncoupled molecules.60 This process also gives rise to PL at around J-aggregate resonance of uncoupled excitons, which thus can be responsible for the higher energy PL peak. Spectral Line Shape and Its Dependence on Temperature. Previous works on free J-aggregates showed that their optical properties can be tuned by temperature. In particular, an increase in the q, narrowing of the PL, and shortening of τ were observed at low temperatures.55,56,68 It is thus possible that similar effects can be important for temperature-dependent strong plasmon−exciton interactions studied here. When the lower polariton branch is pumped via emission of a phonon from a reservoir of uncoupled molecules, the emitting state will be determined by the rate of the competitive processes involved. A good measure characterizing this competition could be the blue shift between the energy of the lower DF polariton in PL and in DF Δ = ωPL − − ω− . This quantity is shown in Figure 3c. We clearly observe that Δ grows with the splitting. By extrapolating this result to smaller splittings, we estimate values at which the blue shift should totally disappear. This happens for splitting of about 150 meV, that is, approaching the high-energy phonons of the system. In the same figure we also observe that the blue shift is offset by about 40 meV at T = 4.5 K, which could be explained by the ∼10 nm temperature-induced red shift in ωpl.56 Previous literature has reported observations of similar blue shifts in some cases,47,66 while in others these were not observed.8,9,67,69 As we show in Figure 3c, the blue shift is reduced for smaller coupling strength, which may be determined by the relation between the Rabi splitting and the phonon energy. In the above-mentioned examples, cavities with the largest Rabi splittings of 330 ÷ 500 meV47,66 indeed showed blue shifts in PL. In cases of weaker couplings,8,9,67,69 blue shifts were not observed, although splitting could have been as large as 230 meV, that is, above the phonon energy. In our previous works, splitting was much weaker than here, yet the blue shift was still observed.28,29 In those cases, however, partial or complete photodegradation of the dye rapidly occurred and thus could be responsible for this blue shift. Conclusions. In conclusion, we demonstrated vacuum Rabi splitting in PL of individual plexciton nanoparticles. This was achieved due to full saturation of the plasmonic mode volume with J-aggregates, which resulted in a giant Rabi splitting of up to ∼400 meV in DF measurements. Moreover, we showed that splitting in PL is temperature dependent and that it is smaller than splitting in scattering. On the basis of the anticrossing data, we assigned the lower energy resonance in PL to the lower polariton emission, while the higher energy resonance to emission of uncoupled molecules. We analyzed the blue shift between the lower polariton emission in PL and DF as arising due to competition between relaxation and emission rates and due to differences in splitting measured in absorption and scattering of nanoparticles. At lower temperature, ωpl is slightly red-shifted,57 which explains a stronger PL splitting observed at low temperature. Even though previous works performed on ensemble level have suggested visualization of both UP and LP branches in PL,18,48 the current study gives a more thorough 556

DOI: 10.1021/acs.nanolett.6b04659 Nano Lett. 2017, 17, 551−558

Letter

Nano Letters Photoluminescence. It is well-documented that J-aggregates are unstable under humid, oxygen, and elevated temperatures conditions.61 Thus, to increase the photostability of the molecular aggregates, the hybrid systems were investigated in an optical cryostat chamber. All measurements were conducted using a 532 nm laser in epi-illumination mode with an irradiance of ∼200 W/cm2 and a long working distance objective (Nikon, 20× NA = 0.45) in an upright microscope. Note that under these conditions, no obvious photodegradation was observed after several hours of continuous irradiation. Numerical Calculations. We performed numerical calculations of plasmon−exciton coupling using the finite-difference time-domain method (FDTD Solutions, Lumerical). On a semi-infinite glass substrate (n = 1.45), we place a J-aggregate coated silver nanoprism (Figure 4). The side length a = 40 nm and the thickness measured h = 11 nm. The corners are rounded (r = 6 nm), and the permittivity of silver is taken from literature.70 Such a nanoprism, when placed on the substrate, has a resonance wavelength of ∼530 nm, a value that does not change significantly with polarization. The coupled plasmon− exciton system is properly modeled by assuming a J-aggregate shell that conformally surrounds the nanoprism. We use linearly polarized light (both polarizations, total-field/scatteredfield formulation) to excite the structure and measure the scattered signal to match the calculated spectrum with the measured one in terms of peak splitting, peak width, and relative amplitude (see Figure 4). This matching is obtained by varying the shell thickness as well as the optical properties of the dye, which are described by a Lorentzian function, εj−agg(ω) = ε∞ − fω02/(ω02 − ω2 − iγ0ω). Matching of the experimental and calculated scattering spectra is obtained for a 3 nm conformal J-aggregate layer with the following optical properties. The background refractive index is 1.45 (ε∞ = 2.10) with an oscillator strength of f = 0.1 and peak width γ0 of 75 meV. The absorption line ω0 is at 588 nm. We then calculate absorption within the structure.



resolution TEM measurements correspondingly. We acknowledge financial support from Swedish Research Council (VR grant: 2012-4014) and Knut and Alice Wallenberg foundation. T.J.A. thanks the Polish National Science Center for support via Project No. 2012/07/D/ST3/02152 and the Polish Ministry of Science and Higher Education via the Iuventus Plus Project No. IP2014 000473. S.B. acknowledges The Scientific and Technological Research Council of Turkey (TUBITAK) (112T091).



(1) Barnes, W. L.; Dereux, A.; Ebbesen, T. W. Nature 2003, 424 (6950), 824−830. (2) Lal, S.; Link, S.; Halas, N. J. Nat. Photonics 2007, 1, 641. (3) Halas, N. J.; Lal, S.; Chang, W.-S.; Link, S.; Nordlander, P. Chem. Rev. 2011, 111 (6), 3913−3961. (4) Tame, M. S.; McEnery, K. R.; Ozdemir, S. K.; Lee, J.; Maier, S. A.; Kim, M. S. Nat. Phys. 2013, 9 (6), 329−340. (5) Törmä, P.; Barnes, W. L. Rep. Prog. Phys. 2015, 78 (1), 013901. (6) Khitrova, G.; Gibbs, H. M.; Kira, M.; Koch, S. W.; Scherer, A. Nat. Phys. 2006, 2 (2), 81−90. (7) Lidzey, D. G.; Bradley, D. D. C.; Skolnick, M. S.; Virgili, T.; Walker, S.; Whittaker, D. M. Nature 1998, 395 (6697), 53−55. (8) Lidzey, D. G.; Bradley, D. D. C.; Virgili, T.; Armitage, A.; Skolnick, M. S.; Walker, S. Phys. Rev. Lett. 1999, 82 (16), 3316−3319. (9) Bellessa, J.; Bonnand, C.; Plenet, J. C.; Mugnier, J. Phys. Rev. Lett. 2004, 93 (3), 036404. (10) Wiederrecht, G. P.; Wurtz, G. A.; Hranisavljevic, J. Nano Lett. 2004, 4 (11), 2121−2125. (11) Tischler, J. R.; Bradley, M. S.; Bulović, V.; Song, J. H.; Nurmikko, A. Phys. Rev. Lett. 2005, 95 (3), 036401. (12) Aberra Guebrou, S.; Symonds, C.; Homeyer, E.; Plenet, J. C.; Gartstein, Y. N.; Agranovich, V. M.; Bellessa, J. Phys. Rev. Lett. 2012, 108 (6), 066401. (13) Fofang, N. T.; Park, T.-H.; Neumann, O.; Mirin, N. A.; Nordlander, P.; Halas, N. J. Nano Lett. 2008, 8 (10), 3481−3487. (14) Symonds, C.; Bonnand, C.; Plenet, J. C.; Bréhier, A.; Parashkov, R.; Lauret, J. S.; Deleporte, E.; Bellessa, J. New J. Phys. 2008, 10 (6), 065017. (15) Schwartz, T.; Hutchison, J. A.; Genet, C.; Ebbesen, T. W. Phys. Rev. Lett. 2011, 106 (19), 196405. (16) Balci, S. Opt. Lett. 2013, 38 (21), 4498−4501. (17) Gómez, D. E.; Vernon, K. C.; Mulvaney, P.; Davis, T. J. Nano Lett. 2010, 10 (1), 274−278. (18) Zhou, N.; Yuan, M.; Gao, Y.; Li, D.; Yang, D. ACS Nano 2016, 10 (4), 4154−4163. (19) Niu, W.; Ibbotson, L. A.; Leipold, D.; Runge, E.; Prakash, G. V.; Baumberg, J. J. Phys. Rev. B: Condens. Matter Mater. Phys. 2015, 91 (16), 161303. (20) Liu, W.; Lee, B.; Naylor, C. H.; Ee, H.-S.; Park, J.; Johnson, A. T. C.; Agarwal, R. Nano Lett. 2016, 16 (2), 1262−1269. (21) Wang, S.; Li, S.; Chervy, T.; Shalabney, A.; Azzini, S.; Orgiu, E.; Hutchison, J. A.; Genet, C.; Samorì, P.; Ebbesen, T. W. Nano Lett. 2016, 16 (7), 4368−4374. (22) Rodriguez, S. R. K.; Feist, J.; Verschuuren, M. A.; Garcia Vidal, F. J.; Gómez Rivas, J. Phys. Rev. Lett. 2013, 111 (16), 166802. (23) Rodriguez, S. R. K.; Rivas, J. G. Opt. Express 2013, 21 (22), 27411−27421. (24) Väkeväinen, A. I.; Moerland, R. J.; Rekola, H. T.; Eskelinen, A. P.; Martikainen, J. P.; Kim, D. H.; Törmä, P. Nano Lett. 2014, 14 (4), 1721−1727. (25) Kena Cohen, S.; Forrest, S. R. Nat. Photonics 2010, 4 (6), 371− 375. (26) Shalabney, A.; George, J.; Hutchison, J.; Pupillo, G.; Genet, C.; Ebbesen, T. W. Nat. Commun. 2015, 6, 5981. (27) Schlather, A. E.; Large, N.; Urban, A. S.; Nordlander, P.; Halas, N. J. Nano Lett. 2013, 13 (7), 3281−3286.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.6b04659. High-resolution TEM image of Ag nanoprism; polarization resolved data; additional statistical data (PDF)



REFERENCES

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. ORCID

Tomasz J. Antosiewicz: 0000-0003-2535-4174 Timur Shegai: 0000-0002-4266-3721 Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We would like to thank Dr. Ruggero Verre and Dr. Andrew Yankovich for their help with sample preparation and high557

DOI: 10.1021/acs.nanolett.6b04659 Nano Lett. 2017, 17, 551−558

Letter

Nano Letters

(61) Würthner, F.; Kaiser, T. E.; Saha-Möller, C. R. Angew. Chem., Int. Ed. 2011, 50 (15), 3376−3410. (62) Savona, V.; Andreani, L. C.; Schwendimann, P.; Quattropani, A. Solid State Commun. 1995, 93 (9), 733−739. (63) Delga, A.; Feist, J.; Bravo-Abad, J.; Garcia-Vidal, F. J. Phys. Rev. Lett. 2014, 112 (25), 253601. (64) Agranovich, V.; Benisty, H.; Weisbuch, C. Solid State Commun. 1997, 102 (8), 631−636. (65) Coles, D. M.; Meijer, A. J. H. M.; Tsoi, W. C.; Charlton, M. D. B.; Kim, J.-S.; Lidzey, D. G. J. Phys. Chem. A 2010, 114 (44), 11920− 11927. (66) Wang, S.; Chervy, T.; George, J.; Hutchison, J. A.; Genet, C.; Ebbesen, T. W. J. Phys. Chem. Lett. 2014, 5 (8), 1433−1439. (67) George, J.; Wang, S.; Chervy, T.; Canaguier-Durand, A.; Schaeffer, G.; Lehn, J.-M.; Hutchison, J. A.; Genet, C.; Ebbesen, T. W. Faraday Discuss. 2015, 178 (0), 281−294. (68) Kamalov, V. F.; Struganova, I. A.; Yoshihara, K. J. Phys. Chem. 1996, 100 (21), 8640−8644. (69) Hobson, P. A.; Barnes, W. L.; Lidzey, D. G.; Gehring, G. A.; Whittaker, D. M.; Skolnick, M. S.; Walker, S. Appl. Phys. Lett. 2002, 81 (19), 3519−3521. (70) Palik, E. D. Handbook of Optical Constants of Solids I; Academic Press: San Diego, CA, 1998.

(28) Zengin, G.; Johansson, G.; Johansson, P.; Antosiewicz, T. J.; Käll, M.; Shegai, T. Sci. Rep. 2013, 3, 3074. (29) Zengin, G.; Wersäll, M.; Nilsson, S.; Antosiewicz, T. J.; Käll, M.; Shegai, T. Phys. Rev. Lett. 2015, 114 (15), 157401. (30) Fofang, N. T.; Grady, N. K.; Fan, Z.; Govorov, A. O.; Halas, N. J. Nano Lett. 2011, 11 (4), 1556−1560. (31) Vasa, P.; Wang, W.; Pomraenke, R.; Lammers, M.; Maiuri, M.; Manzoni, C.; Cerullo, G.; Lienau, C. Nat. Photonics 2013, 7 (2), 128− 132. (32) Balci, S.; Kocabas, C.; Kücu̧ ̈köz, B.; Karatay, A.; Akhüseyin, E.; Gul Yaglioglu, H.; Elmali, A. Appl. Phys. Lett. 2014, 105 (5), 051105. (33) Liu, G. L.; Long, Y. T.; Choi, Y.; Kang, T.; Lee, L. P. Nat. Methods 2007, 4 (12), 1015−1017. (34) Antosiewicz, T. J.; Apell, S. P.; Shegai, T. ACS Photonics 2014, 1 (5), 454−463. (35) Zengin, G.; Gschneidtner, T.; Verre, R.; Shao, L.; Antosiewicz, T. J.; Moth-Poulsen, K.; Käll, M.; Shegai, T. J. Phys. Chem. C 2016, 120 (37), 20588−20596. (36) Wu, X.; Gray, S. K.; Pelton, M. Opt. Express 2010, 18 (23), 23633−23645. (37) Yang, Z.-J.; Antosiewicz, T. J.; Shegai, T. Opt. Express 2016, 24 (18), 20373−20381. (38) Maier, S. A. Opt. Express 2006, 14 (5), 1957−1964. (39) Koenderink, A. F. Opt. Lett. 2010, 35 (24), 4208−4210. (40) Sauvan, C.; Hugonin, J. P.; Maksymov, I. S.; Lalanne, P. Phys. Rev. Lett. 2013, 110 (23), 237401. (41) Kristensen, P. T.; Hughes, S. ACS Photonics 2014, 1 (1), 2−10. (42) Hartsfield, T.; Chang, W.-S.; Yang, S.-C.; Ma, T.; Shi, J.; Sun, L.; Shvets, G.; Link, S.; Li, X. Proc. Natl. Acad. Sci. U. S. A. 2015, 112 (40), 12288−12292. (43) Santhosh, K.; Bitton, O.; Chuntonov, L.; Haran, G. Nat. Commun. 2016, 7, 11823. (44) Chikkaraddy, R.; de Nijs, B.; Benz, F.; Barrow, S. J.; Scherman, O. A.; Rosta, E.; Demetriadou, A.; Fox, P.; Hess, O.; Baumberg, J. J. Nature 2016, 535, 127−130. (45) Yoshie, T.; Scherer, A.; Hendrickson, J.; Khitrova, G.; Gibbs, H. M.; Rupper, G.; Ell, C.; Shchekin, O. B.; Deppe, D. G. Nature 2004, 432 (7014), 200−203. (46) Agranovich, V. M.; Litinskaia, M.; Lidzey, D. G. Phys. Rev. B: Condens. Matter Mater. Phys. 2003, 67 (8), 085311. (47) Schwartz, T.; Hutchison, J. A.; Leonard, J.; Genet, C.; Haacke, S.; Ebbesen, T. W. ChemPhysChem 2013, 14 (1), 125−131. (48) Melnikau, D.; Esteban, R.; Savateeva, D.; Sánchez-Iglesias, A.; Grzelczak, M.; Schmidt, M. K.; Liz-Marzán, L. M.; Aizpurua, J.; Rakovich, Y. P. J. Phys. Chem. Lett. 2016, 7 (2), 354−362. (49) Cacciola, A.; Di Stefano, O.; Stassi, R.; Saija, R.; Savasta, S. ACS Nano 2014, 8 (11), 11483−11492. (50) Thompson, R. J.; Rempe, G.; Kimble, H. J. Phys. Rev. Lett. 1992, 68 (8), 1132−1135. (51) Aoki, T.; Dayan, B.; Wilcut, E.; Bowen, W. P.; Parkins, A. S.; Kippenberg, T. J.; Vahala, K. J.; Kimble, H. J. Nature 2006, 443 (7112), 671−674. (52) Martikainen, J. P.; Heikkinen, M. O. J.; Törmä, P. Phys. Rev. A: At., Mol., Opt. Phys. 2014, 90 (5), 053604. (53) Ramezani, M.; Halpin, A.; Fernández-Domínguez, A. I.; Feist, J.; Rodriguez, S. R. K.; Garcia-Vidal, F. J.; Gomez-Rivas, J. arXiv preprint 2016, arXiv:1606.06866. (54) van Burgel, M.; Wiersma, D. A.; Duppen, K. J. Chem. Phys. 1995, 102 (1), 20−33. (55) Moll, J.; Daehne, S.; Durrant, J. R.; Wiersma, D. A. J. Chem. Phys. 1995, 102 (16), 6362−6370. (56) Coles, D. M.; Michetti, P.; Clark, C.; Adawi, A. M.; Lidzey, D. G. Phys. Rev. B: Condens. Matter Mater. Phys. 2011, 84 (20), 205214. (57) Liu, M.; Pelton, M.; Guyot-Sionnest, P. Phys. Rev. B: Condens. Matter Mater. Phys. 2009, 79 (3), 035418. (58) Khurgin, J. B. Nat. Nanotechnol. 2015, 10 (1), 2−6. (59) Litinskaya, M.; Reineker, P.; Agranovich, V. M. J. Lumin. 2004, 110 (4), 364−372. (60) Herrera, F.; Spano, F. C. arXiv:1610.04252 2016. 558

DOI: 10.1021/acs.nanolett.6b04659 Nano Lett. 2017, 17, 551−558