Dynamic photonic barcodes for molecular detection based on cavity-enhanced energy transfer

Optical barcodes have demonstrated a great potential in multiplexed bioassays and cell tracking for their distinctive spectral fingerprints. The vast majority of optical barcodes were designed to identify a specific target by fluorescence emission spectra, without being able to characterize dynamic changes in response to analytes through time. To overcome these limitations, the concept of the bioresponsive dynamic photonic barcode was proposed by exploiting interfacial energy transfer between a microdroplet cavity and binding molecules. Whispering-gallery modes resulting from cavity-enhanced energy transfer were therefore converted into photonic barcodes to identify binding activities, in which more than trillions of distinctive barcodes could be generated by a single droplet. Dynamic spectral barcoding was achieved by a significant improvement in terms of signal-to-noise ratio upon binding to target molecules. Theoretical studies and experiments were conducted to elucidate the effect of different cavity sizes and analyte concentrations. Timeresolved fluorescence lifetime was implemented to investigate the role of radiative and non-radiative energy transfer. Finally, microdroplet photonic barcodes were employed in biodetection to exhibit great potential in fulfilling biomedical applications.

The vast majority of WGM resonators reported to-date are classified as passive resonators; they require evanescent wave coupling and operate based on changes of the modes induced by perturbations occurring inside the evanescent light fields. 3,6 In contrast, active resonators that utilize analytes as a gain medium can support free space excitation and collection to acquire more biological information from emission signals. 8 Modulated fluorescence (MFL) refers to the fluorescence emission modulated by resonators, in which enhanced spontaneous emission at resonant modes is introduced by the Purcell effect. 1,[34][35][36][37] When considering molecular detection, the mode occupation factor 38 of the analyte outside the cavity is only a few tenths of that inside the cavity (see Fig. S1 in the Supplementary Material), leading to reduced effective Q-factor and Purcell factor and unsatisfactory signal-to-noise ratio (SNR). Hence, we introduce resonant energy transfer by separating donors and acceptors at the cavity interface, where radiative fluorescence energy transfer plays a dominant role, as compared to non-radiative Förster resonance energy transfer (FRET). In conventional cases, FRET is a non-radiation mechanism interpreting energy transfer from excited-state donor to ground-state acceptor through resonant dipole-dipole interaction. Unlike FRET, radiative energy transfer is accompanied by electromagnetic radiation and thereby can occur even if the donor and acceptor are far apart, as shown in Fig. 1(a). In the presence of both energy transfer mechanisms, efficient energy transfer and coupling between donors and acceptors (light-matter interaction) may lead to enhanced SNR and detection limits. Taking advantage of the light-harvesting effect, the high concentration of dye (donor) inside the microdroplet triggers cavity-enhanced energy transfer to excite the molecules (acceptor) attached to the cavity interface. The number of binding molecules therefore alters the amount of energy transfer between donor-acceptor, resulting in distinctive MFL emission peaks.
In this study, we propose the concept of a dynamic photonic barcode based on cavity energy transfer and subsequently employ it in molecular detection using a microdroplet. A new encoding rule that focuses on energy density was established for converting the WGM spectra into photonic barcodes (Fig. 1). Time-resolved fluorescence lifetime measurements were also implemented to investigate the role of FRET and its role in cavity energy transfer [ Fig. 1(b)]. Dynamic spectral barcoding was achieved by a significant improvement in terms of SNR upon binding to target molecules. The contribution of cavity size and donor/acceptor concentration ratio to cavity energy transfer was investigated through theoretical simulations and experiments. In contrast to laser emission-based detection (where a threshold was required), 12 the achievable detection limit could be lower (nanomolar level). Lastly, potential biosensing applications with streptavidin-biotin conjugates were demonstrated.
For the preparation of the streptavidin-coated poly(L-lysine) microdroplet, 10 μL BODIPY-R6G-doped liquid crystal was added into 1 mL SDS/PBS solution and sonicated for 5 min. After centrifugation at 5000 rpm for 4 min, the supernatant was replaced with a 0.01% PLL solution and incubated at room temperature for 30 min. The resulted PLL-modified LC microdroplets were washed by adding PBS solution to remove excessive PLL. Then, 0.1 mg∕mL streptavidin (SA) solution was added to the PLL-modified LC microdroplets and slowly mixed for 1 h. Finally, the microdroplets were washed with PBS solution twice to remove the unbound SA before biosensing. 39,40 For cavity energy transfer experiments, Rhodamine 6G (mediator) and Rhodamine B (acceptor) solutions were prepared by mixing 20 μM Rhodamine 6G with 20 μM Rhodamine B solution in equal volumes.
A typical upright microscopic system (Nikon NI-E) with a 50 × 0.6 NA objective was used to excite the sample and collect the WGM emission from the microdroplets. Light from a SOLA light engine was passed through a filter cube to obtain blue excitation light. The excitation wavelength was 430 to 490 nm, with an average light intensity of 2 W∕m 2 for all measurements in this study. All of the WGM emission spectra were collected through a high-resolution spectrometer (Andor Kymera 328i/ Newton 970). All of the fluorescence images were captured by a ToupTek charge-coupled device (CCD) camera.

Concept of Photonic Barcode via Cavity Energy
Transfer Figure 1(a) illustrates the interaction between donor molecules and acceptor molecules with (bottom) and without (top) a cavity under different molecular distances. The top row shows that nonradiative FRET occurs when donor and acceptor molecules are extremely close (≤ 7 nm). However, when the molecular distances become larger, only weak radiative energy transfer can be sustained between donor and acceptor molecules in free space. Another scenario is when a cavity interface is formed between the donor and acceptor molecules [bottom of Fig. 1(a)]; the photonic environment becomes inhomogeneous due to the cavity modified density of photonic modes. Excited photons from donor molecules tend to be confined into resonant modes, thus energy transfer will mostly exist at the cavity interface where resonances are located ( Fig. S1 in the Supplementary Material). As such, acceptor molecules could still be strongly excited even at a relatively larger distance (≫7 nm). In Fig. 1 the interface. Second, only the donor-acceptor molecules near the droplet interface will contribute to the FRET process (see Fig. S2 in the Supplementary Material). Therefore, our findings indicate that strong radiative energy transfer that takes place in the evanescent field region is much more dominant than nonradiative FRET process. Based on the physical mechanism in Fig. 1(a), the conceptualization of the dynamic photonic barcode was developed based on cavity energy transfer, as illustrated in Fig. 1(c). We adopted free-space coupling to excite the gain medium within the microcavity and collected the photoluminescence spectrum from the leaky modes and non-directional emissions emitted from the non-resonant region, as shown in Fig. S2 in the Supplementary Material. Energy transfer occurs at the interface of the microcavity, where donor and acceptor are separated, leading to a change in the PL spectrum as well as the optical barcode. Each solid bar is predefined at the center wavelength of the corresponding resonance mode, where the width is determined by its integration of resonance intensity after subtracting the associated fluorescence background. To facilitate the quantification of energy transfer, we correlated the energy integration with a colormap, where the darker color represents a larger spectral integrated intensity (from 0.1 to 1.0). As a proof-of-concept, Coumarin 6 (C6) and Rhodamine B were selected as the donor-acceptor pairs, respectively. To optimize the energy transfer efficiency, a mediator (Rhodamine 6G) was used (Fig. S3 in the Supplementary Material). Figure 1(d) presents a typical MFL spectrum with quasiperiodic WGM peaks embedded in the fluorescence emission background (spectrum 1, only donor). In the presence of the microcavity, both spontaneous emission rate and photon directivity of the gain medium would be anisotropic, owing to the inhomogeneous local density of states, which is known as a cavity quantum electrodynamic phenomenon called the Purcell effect. The enhancement of the spontaneous emission rate could be expressed by the Purcell factor, 36 where λ free is the wavelength in the vacuum, n is the refractive index of the liquid crystal cavity, and Q and V represent the quality factor and mode volume of resonant modes of the cavity, respectively. As such, the sharp peaks result from the edge of the cavity where the modes are located, while the rest of the emission from the gain medium contributes to the fluorescence background ( Fig. S2 in the Supplementary Material). However, in the presence of acceptor molecules adhered to the droplet cavity surface by physical absorption (spectrum 2), the WGM peaks beyond 570 nm are evidently enhanced under the same lightemitting diode (LED) pump, indicating that energy has been transferred from the donor band to the acceptor band via cavity energy transfer. The strongest bandwidth of MFL also shifts from 520 to 580 nm, in which several new peaks arise between 670 and 700 nm in the inset of Fig. 1(d).

Physics of Cavity-Enhanced Energy Transfer
For more comprehensive insights into cavity energy transfer, we applied rate equations to simulate the intensity for every single mode and thereby obtain the efficiency of cavity energy transfer. Equations (2)-(5) describe the dynamics of excited-state molecule density and photon density of the donor and acceptor. The first term in Eqs. (2) and (4) represents direct excitation by the pump source. The second and third terms in Eq. (4) are derived from non-radiative FRET and radiative energy transfer from donor to acceptor, respectively. In Eqs. (2) and (4), the integral terms κ rad;D n D ðtÞ R ∞ 0 F p;D ðωÞL D ðωÞdω and κ rad; n A ðtÞ R ∞ 0 F p;A ðωÞL A ðωÞdω represent spontaneous emission in an inhomogeneous photonic environment according to In the above equations, n D ðtÞ, n A ðtÞ, q D ðω; tÞ, and q A ðω; tÞ represent the densities of donor and acceptor molecules in the excited state and densities of the photon at frequency ω emitted by the donor and acceptor molecules, respectively. σ abs;D ðωÞ and σ abs;A ðωÞ are the absorption cross-sections of the donor and acceptor at frequency ω, respectively. I ph ðt; ωÞ is the timedependent pump at the frequency ω in units of photons∕ ðcm 2 · sÞ. N D0 and N A0 are the total concentration of donor and acceptor molecules, respectively. η 1 and η 2 represent the refractive index of the microcavity and the surrounding medium, respectively. κ F is the FRET rate, and κ rad;D , κ nrad;D , κ rad;A , and κ nrad;A denote radiative decay rate and nonradiative decay rate of the donor and acceptor, respectively. To further demonstrate the significance of cavity energy transfer, we compared the WGM spectra with and without the donor inside a microdroplet cavity in Figs. 2(a) and 2(b), respectively. Figure 2(a) shows the dynamic MFL spectra of a Coumarin 6 (C6, donor) microdroplet upon adding 5 μM Rhodamine molecule (acceptor). Over time, the increased amount of acceptor molecules continuously binds to the surface of the microdroplet due to gradient diffusion, as shown in the time-dependent fluorescence image (inset). According to the dynamic spectra, the energy distribution of the MFL tends to redshift due to the increase in energy transfer efficiency. The corresponding photonic barcodes are provided in the right panel, where one can clearly observe the process of energy transfer. As a comparison, Fig. 2(b) presents the modulated emission spectra of a pure LC microdroplet upon adding 5 μM Rhodamine. Slight WGM modulation of fluorescence emission can be observed in Fig. 2(b), where the corresponding photonic barcodes are plotted on the right panel. Without donor excitation from the droplet, the wavelength remains the same in the photonic barcodes. Given that both results in Figs. 2(a) and 2(b) show that external acceptors can be involved in cavity resonance through the evanescent coupling, Fig. 2(a) presented a much higher SNR and dynamic range. More significantly, the number of optical barcodes that could be generated with the existence of donor excitation is estimated to be 10 9 times more complex than that without a donor (details can be found in the Supplementary Material). In addition, the establishment of equilibrium for a transparent droplet was simulated under a green-LED pump (suitable for acceptors and equivalent to excitation from the donor) in Fig. S4 in the Supplementary Material. The density of excited states of the acceptor in equilibrium in the absence of donor was found to be 8 times less than that with the donor, implying that the WGM emission from Rhodamine is supposed to be 8 times different as well.
Note that since Rhodamine molecules will continue to bind on the droplet surface [in Fig. 2(a)], it is expected that the WGM spectrum will continue to change until it reaches an equilibrium. As shown in Fig. 2(c), we demonstrate that the WGM spectra and converted barcodes will reach an equilibrium after 130 s, which could possibly be used to determine the concentration. In particular, we specifically chose three different peaks extracted from Fig. 2(c) and traced the energy transfer efficiency over 7 min in Fig. 2(d). The energy transfer efficiency was calculated by E ¼ 1 − I DA ∕I D , where I D and I DA represent the intensity of the mode in the absence and presence of the acceptor, respectively. Obviously, the peak wavelengths become stabilized eventually; hence, we believe that the barcodes can read out repeatedly after detailed calibrations.

Investigation of the Cavity Size and Molecular Concentration
Here, we explore the minimum cavity size that can support MFL and cavity energy transfer. Figures 3(a)-3(c) demonstrate the MFL spectra of Coumarin 6 droplets with different sizes before and after adding 5 μM acceptor solution. The insets represent the fluorescence image captured by a monochromatic CCD (pseudocolor), illustrating the diameter of each droplet. The free spectral range (FSR) agrees well with the droplet diameter in each figure, as shown in the inset. For instance, in Fig. 3(a), the corresponding eigenmodes of each resonance peak were numerically solved and fitted with a diameter of 3.52 μm. As the cavity size increases, the number of resonance modes, as well as the emission intensity, increases due to the multimode nature of WGM (Fig. S5 in the Supplementary Material). For larger droplets, higher-order TM modes (s ¼ 2, TM 2 l ) also participate in the cavity energy transfer due to the significant evanescent field outside the cavity [see Figs. S1(b) and S1(d) in the Supplementary Material]. The changes of spectra through barcodes with multimodes become complex and distinguishable. Note that the blueshift shown in Fig. 3(a) is due to the temperature effect under LED excitation. Since the ambient temperature will affect the refractive index (RI) of liquid crystals, a blueshift for TM and redshift for TE is sometimes expected regardless of droplet size. 41 Theoretical simulation outcomes are also in good agreement with the experimental results. As presented in the simulation (Fig. S6 in the Supplementary Material), larger droplets generated a more significant intensity change than smaller droplets under a concentration of 5 μM. Nonetheless, in the experiment, it is noteworthy that large droplets suffer from sensitivity when the acceptor concentration is relatively low (nM level). At lower acceptor concentrations, the acceptor signals will be immersed by the strong donor emission from larger droplets. In this case, a smaller cavity size shall be more effective to the binding events due to a larger surface-to-volume ratio. The occupation factor of the evanescent field increases as the cavity size decreases, therefore the ratio of Purcell factors F p;A ðωÞ∕F p;D ðωÞ becomes closer to 1.
Next, we investigated the effect of different acceptor concentrations (0.5, 1, 2.5, and 5 μM) under similar-sized droplets in Figs. 4(a)-4(d). By extracting the WGM emission spectra from Figs. 4(a)-4(d), the converted photonic barcodes after binding to different acceptor concentrations were plotted, respectively [see Fig. 4(e)]. As the concentration of acceptors increased, the relative intensity of peaks at longer wavelengths increased significantly, implying an increased efficiency of cavity energy transfer. The simulation result also agrees well with this phenomenon (Fig. S7 in the Supplementary Material).

Proof-of-Concept for Biomolecular Detection
Finally, we demonstrated the potential application of cavity energy transfer by choosing streptavidin-biotin conjugates as the target molecules. As shown in Fig. 5(a), a streptavidin (SA)coated dye-doped droplet was used as the donor, while biotin molecules labeled with Atto 550 (Biotin-Atto 550) were employed as the target acceptor. Details of droplet modification and coating are provided in Sec. 2. Note that BODIPY-R6G was selected as the donor dye, owing to its huge spectral overlap with Atto 550 [ Fig. 5(b)]. Figure 5(c) presents the WGM spectra when Biotin-Atto 550 molecules were applied to the donor droplet. Significant changes can also be observed via the optical barcode, as shown in the bottom panel of Fig. 5(c), where cavity energy transfer between BODIPY-R6G and Atto 550 occurred. As shown in the inset, new modes TM l 59 , TM l 60 , and TM l 61 were   generated after specific binding to Biotin-Atto 550, demonstrating the potential for utilizing cavity energy transfer in biosensing technology.

Conclusions
In this study, we have reported a concept of dynamic photonic barcodes based on cavity energy transfer, and subsequently employed it in molecular detection. By separating the donor and acceptor at the cavity interface, cavity-enhanced energy transfer was observed between the microcavity and binding molecules. When biomolecules bind to the cavity interface, the MFL spectra shift from green emission to longer wavelengths. An encoding rule was therefore developed to identify the changes of the WGM MFL emission spectra, where more than 10 9 distinctive barcodes could be generated. Different droplet sizes and molecular concentrations were also investigated theoretically and experimentally. Finally, a simple proof-of-concept for biomolecular detection was demonstrated using dynamic photonic barcodes. We envisage that the proposed concept in this study can be widely applied in many biosensing applications and optical encryption. To improve the detection limit, it may be possible to improve the Q-factor of the droplet material or to increase the overlap region between selection of the donoracceptor pair spectrum. This study aims to focus on the physical concept and show the possibility of using this as a more distinguishable readout. Multiplexed detection can be easily achieved by choosing multiple dye-doped droplets (donor) and fluorophores (acceptor). With the addition of one fluorophore (donor or acceptor), the total number of optical barcodes could increase dramatically by several orders. Such complex barcodes could therefore provide a better way to identify and monitor molecular interactions in real time.