Coupling of charge fluctuations probed in the THz range is correlated to fluorescenceWe carried out THz absorption and NIR fluorescence measurements (Fig. 1) to unravel possible correlations between the charge-induced coupling of local solvation and the SWCNT’s fluorescence. We colloidally stabilized the hydrophobic SWCNTs by adsorbing biopolymers such as (GT)10 ssDNA or surfactants (deoxycholate, DOC; sodium cholate, SC, and sodium dodecyl benzene sulfonate, SDBS). This procedure creates an organic corona around the SWCNTs and makes them soluble in water and high ionic strength buffers.Fig. 1: Schematic of carbon nanotube-based molecular sensors and their spectroscopy.Biofunctionalized carbon nanotubes change their fluorescence when they interact with analytes. THz spectroscopy probes how water around them changes. This interplay is studied to understand the sensing mechanism.Unlike in our study of solvated salts, in which rattling of the atomic charges in the water network caused a characteristic anion or cation absorption mode25, for the solvated SWCNT, we observe a broad increase in absorption, extending over the entire frequency range. This indicates a broadband coupling in the frequency range of the translational (100 cm−1), intermolecular stretch, and vibrational modes of water, as predicted by Bocquet et al.21 and Cox et al.22. More quantitatively, the amplitude of Δα is a measure of the charge/carrier coupling. Based on our measurements we find that this coupling depends on the functionalization of the SWCNT, with functionalization by (GT)10 and SDBS resulting in a more efficient coupling to the hydration shell compared to that of DOC and SC (Fig. 2a).Fig. 2: Single wall carbon nanotube (SWCNT) functionalization inversely governs THz absorption and intrinsic fluorescence.a THz difference spectrum, ⧍α (= αsample – αwater) of (6,5)-chirality enriched semiconducting single-walled carbon nanotubes SWCNTs with different surface chemistries/functionalization. Spectra is a mean of two measurements of distinct samples carried out on the same day (b) Fluorescence spectra of SWCNTs with the same functionalization as shown in a. Spectra are mean values of three subsequent measurements. c Maximal fluorescence of the samples shown in b plotted against the difference in THz absorption. Fluorescence data are represented as mean values ± SEM (n = 3).Upon optical excitation (fluorescence measurements) all functionalized SWCNT will create additional charge carriers, the so-called excitons, which can recombine under the emission of a photon in the NIR. In Fig. 2b, we plot the NIR fluorescence intensity for the same functionalized SWCNTs. Interestingly, we find that SWCNTs, which show a more efficient coupling to the charge fluctuations in the hydration shell, as evidenced by a higher ∆α, have a lower fluorescence intensity (quantum yield, Fig. 2c). This gives a hint to the underlying molecular mechanism: We propose that the energy relaxation of excitons can take place either via an efficient coupling into the solvent (as probed in the THz) or via recombination of the charge carriers (as probed by NIR fluorescence). Therefore, the spectroscopic observables of each of these alternative energy pathways will be inversely correlated.We propose that the different spectral properties of functionalized SWCNTs can be explained by the fact that DNA is more loosely bound to SWCNTs compared to sodium deoxycholate (DOC) or sodium cholate (SC). This enables the charge density fluctuations in the water to couple more strongly with the SWCNTs for DNA or SDBS functionalized SWCNT than in the case of DOC or SC. Because of this coupling of SWCNTs to the charge density fluctuations of water, the number of charge carriers in SWCNTs that undergo recombination is reduced. This corresponds to the quenching of fluorescence and consequently, a decrease of the fluorescence quantum yield given that the absorption cross-section does not depend on the surfactant30. In the case of DOC or SC, the coupling between the charge fluctuations of water molecules in the solvation layer and the SWCNTs is reduced, as confirmed by the smaller Δα (Fig. 2a) but in turn, the quantum yield for fluorescence is increased (Fig. 2b).To validate this model, we tested (GT)10-SWCNTs, which are highly selective NIR fluorescent biosensors for the neurotransmitter dopamine and the vitamin riboflavin. It is known that upon addition of the analytes dopamine or riboflavin, the impact on the quantum yield is opposite i.e., the fluorescence intensity increases for dopamine and decreases for riboflavin13,31,32. While the effect is known, the underlying mechanism has not been understood so far. In Fig. 3a, we plot Δα, i.e., the THz difference spectra upon the addition of either dopamine or riboflavin to the (GT)10-SWCNTs concerning water. Per our working hypothesis that a decrease/increase in THz absorption goes along with an increase/decrease in fluorescence, we found that upon addition of dopamine, epinephrine, and ascorbic acid Δα decreased and for riboflavin Δα increased, compared to the solvated (GT)10-SWCNTs (Fig. 3a and Supplementary Fig. 1). For direct comparison, we also measured changes in the fluorescence of the DNA-SWCNTs upon the addition of these analytes (Fig. 3c, and Supplementary Figs. 2, 3). All analytes were added at the same concentration (100 µM) as in the THz spectra. Additionally, we carried out the above-mentioned experiments with (AT)15-SWCNTs (Fig. 3b, d, Supplementary Fig. 4). The fluorescence and THz absorption responses of (AT)15-SWCNTs followed the same trend as seen for (GT)10-SWCNTs, but the magnitude of change in fluorescence upon the addition of analytes is smaller. Since the coupling of the electronic fluctuation of SWCNTs to charge density fluctuations in water depends on the specific SWCNT corona (Fig. 2), the alterations induced by the analytes can vary in magnitude. The chirality of the SWCNTs could play a certain role in the sensing mechanism. In our experiments we used (6, 5)-SWCNTs. It was shown before that the chirality of DNA-SWCNT-based sensors for small molecules plays a less important role than the surface chemistry (e.g., DNA sequence33, compare Supplementary Fig. 5). Other factors such as orientation could also play a role but should be negligible in solution34.Fig. 3: Analyte binding affects coupling of the hydration shell and single wall carbon nanotube (SWCNT) photophysics.a Difference spectra, ⧍α (= αsample – αwater) of (GT)10-SWCNTs, i.e., change in THz absoption upon addition of analytes (100 µM) in physiological buffer. (All spectra are corrected for water vapor lines). b Change in difference spectra upon addition of the analyte at ν = 150 cm-1 compared to the absorption of the solvated (GT)10- and (AT)15-SWCNTs. Data is represented as mean values ± SEM (n = 2). c Representative NIR-fluorescence spectra of (GT)10-SWCNTs before and after analyte addition (100 µM) in physiological buffer. d Normalized fluorescence changes of (GT)10 and (AT)15-SWCNTs after addition of 100 µM respective analyte. Data are represented as mean values ± SEM (n = 3).In the THz spectra (Fig. 3a), we observed a broad increase in absorption between 100 cm−1 and 200 cm−1. Remarkably, these two frequencies correspond to the expected center peak frequencies for the translational and intermolecular stretch of bulk water, respectively. A further increase in Δα is observed for frequencies >350 cm−1, i.e., at the onset of the librational mode. This experimental observation supports our hypothesis that Δα probes the coupling between the THz spectrum of water and the SWCNTs, which is predicted theoretically. To quantify the coupling, we plotted in Fig. 3b the change in difference spectra upon the addition of the analyte at 300 cm−1 (for ν = 150 cm−1 see Supplementary Fig. 6) compared to the absorption of the solvated (GT)10−SWCNTs, i.e., (Δα-Δα0)/Δα0, which yielded an inverse correlation between both quantities.Fig. 4: Fluorescence quantum yield is inversely proportional to coupling between single wall carbon nanotube (SWCNT) and water as probed by (Δα-Δα0)/Δα0.a Plot of the relative change of α (i.e., change in THz absorption) vs relative change of I (i.e., change in fluorescence intensity, which is a measure for quantum yield) upon addition of analytes dopamine, ascorbic acid, and riboflavin. In this plot, all values for Δα are taken at 150 cm-1 wavenumbers. The data points shown include data of the SWCNTs functionalized with two distinct DNA sequences, (AT)15 and (GT)10. Dashed lines represent linear fits for both DNA sequences with slopes m(GT)10 = −4.11557 and m(AT)15 = −2.07549. Data are represented as mean values (n = 3). Error bars = SEM (n = 3). b Simplified energy state schema for excited SWCNTs suggesting an alternative non-radiative decay (NRD). GS: Ground state; E11/E22: SWCNT excited states; VIS: Visible excitation; NIR: NIR fluorescence emission. Arrows indicate the transition between states.Thus, based on our experimental results, we conclude that the addition of analytes impacts the coupling between the DNA-SWCNT and the hydration water, which in turn changes the fluorescence quantum yield as can be seen by the fluorescence intensity changes (Fig. 3c, d). Note that fluorescence intensity and fluorescence quantum yield are proportional to each other because the absorption spectra of the SWCNTs are not changed by the analyte (Supplementary Fig. 7)13.For SWCNTs, it has been proposed that there is a strong coupling between surface plasmon modes in SWCNTs and excitons35. As the energy of (axial) surface plasmons of SWCNTs lies in the THz region36 we propose that this is a dissipation pathway. The spectral overlap between excitons and THz modes of water alllows coupling. An alternative could be the large number of (dark) states within the bandgap of the SWCNT.In Fig. 4a, we plotted the normalized change in fluorescence intensity against the normalized change in Δα. We observe an inverse correlation between the change in quantum yield/NIR fluorescence and the change in Δα. Whereas the same trend is seen for analytes with (AT)15 and (GT)10 SWCNT, the correlation factor, i.e., the value of the slope, is found to be DNA-sequence specific. The (GT)10 functionalized SWCNT responds strongly to the binding of the analyte. Therefore, the slope, which serves as a guide to the eye is steeper than for (AT)15 functionalized SWCNT (Fig. 4a).This pattern further corroborates our theory of competing pathways for the relaxation of charge carriers/excitons in SWCNTs. Based on the changes the analytes create in the local environment, the coupling between water and SWCNTs changes. The fluorescence signal is higher when the coupling is lower and vice versa (Fig. 4b). To explain these differences, we carried out accompanying molecular dynamics (MD) simulations (Fig. 5).Fig. 5: Effects of ligand binding on the SWCNT’s (single wall carbon nanotube) hydration shell probed by molecular dynamics simulations.a Snapshot of a dopamine ligand directly contacting the SWCNT surface. Carbon, oxygen, nitrogen, and hydrogen atoms of dopamine are depicted in red, green, yellow, and white, respectively. DNA appears in orange, SWCNT in white, water in dark blue, and the dark gray region represents the calculated localized water contact area. b Snapshot of two riboflavin ligands in direct contact with the SWCNT surface. The color scheme mirrors panel a, with riboflavin’s carbon atoms in blue. c Schematic of the two preferred binding modes for dopamine (SWCNT surface) and riboflavin (DNA corona). d Bulk contact areas between the SWCNT surface and all dopamine (DA-red) or riboflavin (RBF-blue) molecules. Data are presented as mean values ± SEM (n = 30,000 configurations extracted from 3 μs MD trajectories), where averaging is performed over the trajectory time.Hence, the SWCNT corona is impacted by binding to an analyte. In general, analytes that increase fluorescence, such as dopamine, reduce the observed THz coupling efficiency, whereas quenching analytes, such as riboflavin, increase the coupling efficiency. The underlying molecular alterations of the corona are not revealed by the spectroscopical data. To understand the changes happening in the corona we carried out accompanying molecular dynamics (MD) simulations (Fig. 5 and Supplementary Fig. 8).They underline the differences in the interactions between (GT)10-wrapped SWCNTs and two analyte molecules, dopamine, and riboflavin. In simulations of DNA-SWCNTs with dopamine, three out of ten dopamine molecules and four out of ten dopamine molecules are directly bound to the SWCNT surface, as seen in Fig. 5a. The remaining dopamine molecules are bound to the DNA corona. In simulations of DNA-SWCNTs with riboflavin, two out of ten molecules and three out of ten riboflavin molecules directly bound to the SWCNT surface (Fig. 5b). However, riboflavin molecules were either tilted with respect to the SWCNT surface (Fig. 5b) or were associated purely with the DNA corona. Overall, riboflavin molecules exhibited a preference for binding to the DNA corona. Based on these observations, we propose two general binding modes: Mode 1 where the analyte binds directly to the SWCNT surface, and mode 2 with analytes binding to the corona. Figure 5c provides a schematic of each of these modes. In the SI (Supplementary Tables 1 and 2) we provide the details of the simulated systems and binding events in the systems characterized by the type of the binding mode, and the number of carbon atoms of the SWCNT in the vicinity of the bound ligands.To assess the influence of the analyte on the bulk contact area between SWCNT and water/ions, we performed an analysis on five simulated systems, including a control system without a ligand. Two systems contained dopamine molecules, and two systems contained riboflavin molecules. While no significant differences were observed in the contact areas between SWCNT surface to water and sodium ions across all systems, the contact areas between ligands and the SWCNT surface exhibited substantial variations. Specifically, dopamine molecules demonstrated a higher contact area with the SWCNT surface compared to riboflavin (Fig. 5d). This direct binding was confirmed experimentally by isothermal titration calorimetry (ITC). Dopamine was titrated in (GT)10-SWCNTs to a total concentration of 100 µM (Supplementary Figs. 9, 10 for riboflavin, and Supplementary Table 3). Based on these experiments we deduce an interaction stoichiometry ranging from 1:30 to 1:40 (SWCNTs: dopamine), meaning that a single SWCNT covered with ssDNA harbors on average 37 strong interaction sites for dopamine. They correspond to higher affinity states on the SWCNT surface with a larger thermodynamic footprint. The lower affinity binding sites in the DNA corona showed a smaller thermodynamic footprint and did also not saturate at the high concentrations up to 100 µM. This result verifies the two different binding modes found in the MD simulations.The analysis of the overall contact areas between the SWCNT surface and water or sodium ions for modeled systems, including the control (GT)10-wrapped SWCNT system (no analytes) and (GT)10-wrapped SWCNT systems binding to dopamine or riboflavin analytes. These bulk analyzes of the systems revealed no significant overall effects of analytes on the environment of the SWCNTs. Given these results of the bulk analysis, the localized contact areas between the SWCNT surface segments and surrounding water molecules were examined next (Supplementary Fig. 11). In this approach, we tracked the number of water molecules near local regions of SWCNTs, comprised of a fixed number of SWCNT carbon atoms in the vicinity of the bound ligands. We identified local regions for all ligand binding events that last longer than 400 ns. In Supplementary Fig. 12 we plotted the change in the average contact area between water and the local SWCNT surface when analytes are binding to the regions and when they are not binding. Since the sizes of the local SWCNT regions can differ, we normalized the water contact area values by dividing each value by the maximum value observed for the contact area of water. Local SWCNT regions had less direct contact with water when the analytes bind to the local region (in binding mode 1, as for dopamine) than when they bind to the DNA corona (mode 2, as for riboflavin). In fact, the water-region contact areas were similar when the ligands bound to the DNA corona (binding mode 2) and when the ligands left the region, indicating that ligand binding to the corona does not affect the exposure of local SWCNT regions to water. These results together with the THz results indicate that one analyte (riboflavin) affects the water structure in the DNA corona while the other analyte (dopamine) affects mainly the water structure directly on the SWCNT surface.In the case of dopamine, we expect a bigger impact on the coupling between charge fluctuations of the water and the SWCNTs than in the case of the more loosely bound riboflavin. Therefore, in the case of dopamine, we observe a reduction in the coupling of the SWCNT and the water, as quantified by Δα, and thus an increase in quantum yield, whereas for analytes that quench SWCNT fluorescence (e.g., riboflavin) the opposite holds. Thus, local solvation changes at the SWCNT surface affect the fluorescence of SWCNTs. However, in contrast to previous common bulk solvatochromic models of SWCNTs that predict wavelength shifts if SWCNTs are exposed to more or less water, there is no overall net change in SWCNT exposure to water5,11. The solvatochromic models were unable to predict the fluorescence increases or decreases observed for many SWCNT-based sensors including the ones studied here1. The model presented here gives now a conclusive explanation for the underlying molecular mechanism.
