Letter pubs.acs.org/NanoLett
Time-Resolved Emission Reveals Ensemble of Emissive States as the Origin of Multicolor Fluorescence in Carbon Dots Syamantak Khan, Abhishek Gupta, Navneet C. Verma, and Chayan K. Nandi* School of Basic Sciences, Indian Institute of Technology Mandi, Himachal Pradesh 175001, India S Supporting Information *
ABSTRACT: The origin of photoluminescence in carbon dots has baffled scientists since its discovery. We show that the photoluminescence spectra of carbon dots are inhomogeneously broadened due to the slower relaxation of the solvent molecules around it. This gives rise to excitationdependent fluorescence that violates the Kasha−Vavilov rule. The timeresolved experiment shows significant energy redistribution, relaxation among the emitting states, and spectral migration of fluorescence spectra in the nanosecond time scale. The excitation-dependent multicolor emission in time-integrated spectra is typically governed by the relative population of these emitting states. KEYWORDS: Carbon dots, solvent relaxation, inhomogeneous broadening, spectral migration, red-edge effect, time-resolved fluorescence emission
C
solvent. This suggests a major role of solvent polarity in CND fluorescence (Figure S2). After the absorption of light (10−15 s), two simultaneous nonradiative processes (i) vibrational relaxation and (ii) solvent relaxation take place to the band edge (10−12 s) followed by a radiative fluorescence (10−9 s).5,15 Vibrational relaxation, which is the intrinsic property of the fluorophore molecule, leads to a red-shifted emission compared to the absorption wavelength, known as Stokes shift (Figure 1A, left panel).16 Solvent relaxation or dipolar reorganization of the surrounding solvent molecules further promotes the Stokes shift to a higher value. A rapid solvation (10−12 s) will cause the complete relaxation or lowering of energy well before the fluorescence decay (10−9 s). The net result will be a redshift of the emission spectrum, commonly known as solvatochromic shift (Figure 1A, right panel).17 The greater the polarity of the solvent, the greater the red shift according to the Lippert−Mataga equation.18,19 When solvation dynamics slows down to the time scale of fluorescence decay (10−10−10−9), a set of ensemble of substates is formed. The fluorescence is observed from each of these states with a sharp maximum. When their contribution is added, broad-band emission is observed, known as inhomogeneous broadening. This depends on temperature, solvent dipole moment, and change in fluorophore dipole moment on excitation.20
arbon dots (CND), a new class of carbogenic nanomaterial, have extensively been used in recent years for bioimaging, catalysis, light harvesting, photovoltaics, and drug delivery. They have been proven as an extremely bright, highly photostable, and less toxic multicolor fluorescent nanomaterials.1−3 However, the origin of excitation-dependent fluorescence of CND, violating the Kasha−Vavilov rule of excitationindependent emission4,5 (Figure 1A, middle panel), is not clear until date and remains a topic of debate. Starting from the quantum confinement,3,6 the recent accepted mechanism is the presence of different surface defects that give rise to different chromophoric groups.6−8 Whether the fluorescence originates from a single particle with multichromophoric groups or from different particles, is still debatable.9 In addition, recently observed fluorescence blinking violates the earlier reported nonblinking phenomena.9,10 The important role of either individual surface groups and core or their combined effect has also been established.11,12 A general trend in CND bulk fluorescence is the gradual excitation-dependent shifting of emission spectra from its emission maxima to the longer wavelength with a decrease in intensity (Figure S1). Sometimes, a shorter wavelength component is also observed that has either little or no excitation-dependent behavior. These are basically two different overlapping emissive species as confirmed by the polaritydependent chromatographic separation13,14 (Figure S2). The shorter wavelength component was less polar due to the presence of amine groups, but the longer wavelength component was highly polar due to the presence of −CO and −COOH groups as confirmed by FTIR spectroscopy (Figure S3). The excitation-dependent emission of the shorter wavelength component reappeared when dissolved in polar © XXXX American Chemical Society
Δν = Δμa−3/2(kT )1/2
and Received: September 26, 2015 Revised: November 12, 2015
A
DOI: 10.1021/acs.nanolett.5b03915 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 1. Schematics of the dynamical aspects of various fluorescence phenomena. (A) Solvent relaxation with a time scale of fluorescence decay (τR ≈ τF) results in emission from various substates with variable extent of relaxation, thus violating both Kasha−Vavilov rule and resulting in inhomogeneous broadening. (B) Emission occurred following the Vavilov rule, whereas (C) shows the violation of the Vavilov rule by allowing the excitation dependent emission. (D) Homogeneously broadened time integrated spectra. (E) The fluorescence emission maxima migrate toward a more relaxed state with higher wavelength in time, giving rise to inhomogeneously broadened time integrated spectra. (F,G) The dipolar rotation causes depolarization or energy relaxation with respect to time. A state with higher depolarization has less anisotropy and low energy, which corresponds to a red-shifted spectrum in the time-resolved spectra.
A=
⎛2⎞ 1/2 ⎜ ⎟[(ε − 1)/(2ε + 1)] ⎝h⎠
of substates are significantly populated, which give rise to a spectral migration of ∼100 nm in TRES. The time-dependent shift of the emission position is a function of both relaxation time as well as fluorescence lifetime. The center of mass of the steady-state spectrum5 can be calculated using the following equation:
where ε is the dielectric constant of the medium, and a is the Onsagar sphere radius. So in the case of a slower solvation dynamics, the initial energy state migrates toward a lower and relaxed state in the same time scale of fluorescence decay. The implication of this includes a red-shift of times resolved spectra (TRES) as well as time integrated spectra (TIS) as a function of time and excitation wavelength, respectively. Figure 2 shows the steady state and time-resolved spectra in water and dimethylformamide (DMF). An increase in brightness as well as fluorescence lifetime of graphene quantum dots (GQD) in DMF is already observed in previous studies21 where DMF is used for the synthesis of the GQD in a hydrothermal treatment. In our study DMF has been used as an exchanged solvent. Due to the higher dipole moment of DMF, a number
vs = νt →∞ − (νt → 0 − νt →∞)τR /(τR + τF)
where νt→∞ is the emission maxima at infinite time, νt→0 is emission position at time zero, τR is the relaxation time, and τF is the fluorescence lifetime. It could be intuitively understood from the equation that only when the time constants are of the same order, a time dependency should arise. At τF≫ τR, ν(t) = νt→∞, and at τR≫ τF, ν(t) = νt→0. At low temperature and high viscosity, the red-edge effect is vanished (Figure S4) and emission is observed from a fixed wavelength. This unique B
DOI: 10.1021/acs.nanolett.5b03915 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
lifetime in the two mentioned components. For instance, they show average lifetime of 3.5 and 6.4 ns (Figure S2, Table S1) when monitored at 350 and 463 nm, respectively (assuming the distinct populations of the two components in these wavelengths). The CND with a high degree of oxidation and higher polarity (Figure 3E−H) invariably show higher lifetime values, which notably increase with increasing emission wavelength (Table S1). This implies that the relaxed substates would be significantly populated before fluorescence decay occurs. Therefore, a higher number of substates will contribute to photon emission, eventually giving rise to the edge effect and an inhomogeneously broadened spectrum (Figure 1, right panel). In contrast, the less oxidized and less polar CND fraction (Figure 3A−D) shows a considerably shorter lifetime component and displays a little or no dynamical red shifting. This is rather intuitive, as radiative photon emission is quicker in this case and most of the photons decay before a significant relaxation can take place (Figure 1, upper panel). Though the time-resolved spectra show a visible spectral migration, the relaxed low energy substates are not fairly populated to have an observable impact on the steady state time integrated spectra. Even if the edge effect be observed, the intensity of the redshifted emission would drastically drop with increasing wavelength. The time-resolved anisotropy decay, which reflects the rotational correlation time of a fluorophore in the process of relaxation, is studied at different wavelength. The rotational time is found in the nanosecond scale and decreases with respect to emission wavelength (Figure 4A−C, Table S2). The value of anisotropy also decreases along with rotational time at a higher wavelength. This could be attributed to the fact that at a higher emission wavelength only those fluorophores that emit photons only after completing a certain degree of rotation can be probed. Fluorophores of this substate start from a low anisotropy value as mentioned earlier and complete the final part of its rotation in rather shorter time. In contrast, at lower wavelength, one can probe the substates, which decay photons
Figure 2. Role of solvent dipole moment. (A,C) Steady state spectra and (B,D) TRES of carbon dots in water and dimethylformamide, respectively. The spectrum is heavily dependent on the polarity of the solvent. Both energy of relaxation and spectrum broadening increases, with higher dipole moment of the solvent molecules.
observation is actually universal for any CND irrespective of starting material. The reason behind this is a slower relaxation process and a higher relaxation time, which is a function of viscosity (η), temperature (T), and volume (V). τR =
ηV kBT
Both of the purified components of CND1 (Supporting Information) show spectral migration and relaxation to a lower energy state with time. Nevertheless, their time integrated spectra are distinctly different in terms of red-edge effects, which is of course due to their difference in intrinsic polarities (Figure S2). This happens due to a mismatch of fluorescence
Figure 3. Inhomogeneous broadening and spectral migration of carbon dots fluorescence. (A,E) The steady state fluorescence spectra of two separated components (CND1) with different polarities. The component with highly excitation-dependent emission (E) clearly shows an increase in fluorescence lifetime with respect to the emission wavelength (F,G) in contrast to the other component (B,C). The lifetime components are also larger in (G). (D,H) TRES of both components shows spectral migration, while the depopulation of the substates is quicker in (D). Inhomogeneous broadening and higher excitation dependence arise from the higher contribution of the relaxed substates (E−H). A shorter fluorescence lifetime causes most of the photons decay from the initial states, resulting a little excitation dependence of the steady state spectra (A−D). C
DOI: 10.1021/acs.nanolett.5b03915 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 4. (A−C) Time-resolved fluorescence anisotropy decay of CND1 (purified) measured at different emission positions. The rotation of dipole with respect to time give rise to various relaxed substates, which contribute to the time integrated fluorescence spectra. At a higher wavelength (C), the overall anisotropy value decreases as most of the species which emit at this wavelength have completed their rotation causing depolarization. (D−F) Effect of solvent pH. At a higher pH a blue shift is visible in time integrated spectra (E) with a distinct change of the spectrum. (F) TRES shows a higher number of substates are formed at this higher pH with a quick depopulation of the states with respect to time. A large (∼70 nm) shift of TRES at pH 12 in comparison to pH 3 (30 nm) suggests a possible association of excited state proton transfer mechanism with carbon dot photoluminescence.
intensity is linearly dependent on concentration. However, the overall spectrum is found to shift toward lower energy at very high concentration, with a drop in quantum yield (Figure S6). At higher concentration a number of emitters form a cluster thus ensuring the energy is transferred to the emitters, which emit at higher wavelength. In dilute solution the clusters become smaller and less effective. The model of “nanocluster” has already been proposed in a previous study.25 Blue shift at higher pH as mentioned above (Figure S5) also supports this model where electronic repulsion further destabilizes the cluster. This also indicates that the emission centers need to be situated at the surface of the individual species. Fluorescence quenching using methyl viologen in our study (Figure S7) as well as in previous studies13 strongly supports this claim. A recent report shows the fluorescent spectrum of a single CND is asymmetric and broadened at the lower energy region.26 This clearly indicates an energy relaxation process. The particles are also found to show similar spectra when excited at two different wavelengths. However, the particles in that study are immobilized. In a polar solvent a single CND should certainly show broadening effect and excitation dependence. We emphasize, in our study, the flash chromatographic purification, which produced a fairly pure and homogeneous sample. This pure sample shows even stronger excitation-dependent fluorescence suggesting that the excitation dependence does not arise from the particle heterogeneity. This typical continuous shifting is invariably observed throughout the entire spectral range, i.e., from violet to red.27,28 Like many other reports, the study of ref 25 also used a heterogeneous sample, where four different random particles show different spectra. So this does not necessarily prove that the heterogeneity is the origin of excitation-dependent multicolor spectrum.
at the onset of their rotation, ensuring a larger anisotropy and requiring longer time to complete its rotation. A number of environmental factors affect the overall photoluminescence of CND.1−3 The intensity of CND fluorescence generally shows maxima at neutral pH and drops at both acidic and basic environment.1,22 Additionally, we observed a large spectral shift toward higher energy at pH > 10 (Figure S5). In time-resolved fluorescence experiment, the effect of pH is shown in Figure 4D−F. A longer range (∼70 nm) spectral shift is distinctly visible at higher pH compared to the lower pH values. Time resolved anisotropy decay at variable emission position shows the dependence of rotational time at different pH (Table S2). Protonation at higher pH or deprotonation at lower pH induces a higher degree of charge separation within the fluorophore, which in turn generates a stronger dipole moment within the molecule. At pH > 10, emission starts from higher energy states created by the complete ionization of both −COOH and −OH groups. The long-range migration of time-resolved spectra can originate from various complex mechanisms. Excited state proton transfer (ESPT)23,24 is one of the possible routes of this significant relaxation. In that case, protons must be transferred from other groups of CND (intramolecular) with a very high dissociation constant, as transfer of proton from the solvent at pH > 10 would be rare. Besides, given the large number of lone pairs of oxygen at high pH, solvent relaxation can itself give rise to a significant energy relaxation without any transfer of protons. Though ESPT might be associated with CND, it does not completely rely on that as CND is bright enough in the aprotic solvents (e.g., DMF) and shows usual spectral migration and edge effect (Figure 2C,D). The concentration of the CND is also an important parameter to consider. Throughout our experiments we used much diluted solution (
