Buckminsterfullerene (C60), the most abundant fullerene as well as the most symmetric molecule in nature, has been studied extensively since its discovery in 19851. Among the outstanding properties of C60 are ability to accept up to six electrons, high intersystem crossing (ISC) quantum yield, and long-lived triplet states2, which have stimulated a thriving research effort for applications in photovoltaics3, photocatalysis4, and molecular probes5. Additionally, the latter two attributes make C60 an efficient singlet oxygen (1O2) sensitizer, particularly promising for photodynamic therapy and environmental remediation6,7,8.

However, C60 is subject to photodegradation in these applications at ambient temperature (while thermal oxidation of C60 occurs at 370 K and above)9,10. The initial work on photooxidation (PO) of C60 credited it to ozonation11,12,13, which however is limited to excitation wavelengths shorter than 240 nm for photogeneration of O314. Later, however, another reactive oxygen species became the suspect, 1O215,16,17. To this end, the most seminal findings have been: (i) unless photoexcited, C60 does not react with externally-produced 1O216; (ii) PO can occur under UV excitation (308 nm) in O2 ambient (albeit formation of 1O2 was not corroborated)17; and (iii) C60O is the major photoproduct16,17,18. Thus, the PO reaction is anticipated as \({\mathrm{C}}_{60}^{\ast} + {\,}^{1}{\mathrm{O}}_{2} \to {\mathrm{C}}_{60}{\mathrm{O}} + \frac{1}{2}{\mathrm{O}}_{2}\), where both C60 and O2 must be photoexcited. However, the mechanistic details of the photophysics and photochemistry remain unelucidated. A historical review of the C60 PO literature is provided in Supplementary Note 1.

C60 has long been known to be an efficient 1O2 sensitizer. Yet, no evidence has been shown that 1O2 reacts with its original C60 sensitizer, which we refer to as “oxidation with self-sensitized 1O2”. Here, we present experimental evidence for this phenomenon. Although the lifetime of 1O2, τ, in air is exceptionally long (i.e., 45 min)19, it shortens to microseconds to nanoseconds in solvents. Inspired by this broad range of τ, we investigated PO of C60 in hexane (C6H14), chloroform (CHCl3), and carbon tetrachloride (CCl4), where τ is 30, 207 and 87,000 μs, respectively20. Our kinetics study reveals C60 concentration decays exponentially under UV excitation and the decay rate increases with τ. We also show the decay dominantly occurs as a single-photon process above the photon energy (hυ) threshold of 3.7 eV, being the onset of 11Ag → 21Hu transition in C60.

Results and discussion

Absorption spectroscopy

Figure 1a, b shows time series optical absorption spectra for our slowest and fastest PO kinetics, which occur in C6H14 and CCl4, respectively. The major C60 absorption peaks at 256 and 328 nm are seen to decrease systematically, while the baseline rises indicative of a photoproduct, which is also evident from yellowing of the solution (Fig. 1b, inset). The spectrum of the excitation source (Fig. 1c) consists of a major narrow band peaking at 310 nm with no emissions below 250 nm. Hence, the possibility of O3 generation is ruled out14. In Fig. 1b, the noise below 255 nm is due to the high absorption of CCl4 attenuating the optical beam. However, it does not deteriorate the accuracy of the absorption peak of C60 at 260 nm in CCl4 (see Supplementary Discussion).

Fig. 1: Time series absorbance spectra of C60 under UV excitation of 3.74 mW/cm2.
figure 1

a In C6H14; b in CCl4. The inset of (b) shows photos of C60 solutions, unexposed and after 8 min of UV exposure. c The spectrum of the excitation source, UVP XX-15 UV lamp.

Phosphorescence spectroscopy

In Fig. 2a, the phosphorescence peak at 1273 nm substantiates photosensitization of 1O2 by C60 in the three solvents in our study. Figure 2b plots the measured phosphorescence peak intensity versus quenching rate constant (kq = 1/τ). Here, values of τ for the three solvents are borrowed from ref. 20 as listed above. The experimental data points match with the theoretical trend (Supplementary Eq. (48)). Hence, we confidently adopt the τ values from the literature.

Fig. 2: Phosphorescence of 1O2 sensitized by C60.
figure 2

a Spectra in different solvents (under 375 nm radiation of 9.5 mW/cm2 intensity). Here, [C60] in C6H14 is 3.92 times higher than the usual concentration. b Match of theoretical intensity (I) with experiment (in (a)) validating the kq values adopted from the literature. ks2 (0.4 s−1) is the sensitization rate by C60. The 2 in the subscript indicates \(h\upsilon\) < 3.7 eV (Supplementary Results).

Vibrational spectroscopy

PO of C60 is also characterized by the Fourier-transform infrared (FTIR) spectra in Fig. 3a, where C–O, C═O, and O–H stretching vibrations are indicative of C60 oxidation13. We anticipate the O–H groups result from the Norrish type II reaction21. The evolution of C–H vibrational peaks suggest fragmentation of the C60 cage subsequent to PO. The peak frequencies and important assignments are shown in Fig. 3b22,23. Detailed peak assignments are given in Supplementary Table 1.

Fig. 3: Time series FTIR spectra.
figure 3

a C60 in CHCl3 under the same UV exposure conditions as in Fig. 1. b Assignment of FTIR peaks after 6 h of UV exposure (ρ: rocking; δ: bending; υ: stretching).

Mechanism of C60 photooxidation in solvents

At first, we are inclined to explain the oxidation of C60 by its reaction with free 1O2. In this model, 1O2 is released to the solvent after photosensitization by a C60. Subsequently, it collides and reacts with a C60 unless quenched by the solvent. This straightforward model (detailed in Supplementary Information; Kinetics Model) is consistent with our observation that PO rate increases with τ. Additionally, its rate is quadratic in [C60] as well as radiation intensity. On the contrary, the kinetics of C60, as monitored from optical absorption (Fig. 4a) suggests exponential decay, i.e., \(\frac{\mathrm{d}}{{\mathrm{d}t}}\left[ {\mathrm{C}}_{60} \right] = - k_{\mathrm{pd}}\left[ {\mathrm{C}}_{60} \right]\), where kpd is the C60 photodecay rate. Additionally, Fig. 4b establishes a linear dependence of kpd on excitation intensity, and hence a single-photon process. Thus, we rule out “oxidation with free 1O2” as the dominant PO mechanism. Instead, consistent with the observed exponential decay, we propose “oxidation with self-sensitized 1O2”, where a C60 molecule photosensitizes a 1O2 and reacts with that same 1O2 in a collision complex:

$$ {\mathrm{C}}_{60} + {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{collision}}}{\mathrm{C}}_{60} \bullet {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{hv}}{\mathrm{C}}_{60}^ \ast \bullet {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{sensitization}}}\\ \hskip 50pt {\mathrm{C}}_{60}\bullet {\,}^{1}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{\mathrm{oxidation(?)}}{\mathrm{C}}_{60}{\mathrm{O}}$$
Fig. 4: C60 oxidation kinetics.
figure 4

a C60 time decay in different solvents. A is optical absorbance at 256 nm (peak), Ao being the original value (prior to exposure). Although the photoproduct baseline is not subtracted, the slopes represent exponential decay rates (kpd) with minimal error as corroborated in Supplementary Discussion. The UV exposures are same as in Fig. 1. b kpd in CCl4 as a function of irradiation intensity.

Yet, Scheme (1) has a flaw with the oxidation step, where C60 and 1O2 will not react, because C60 must be excited to \({\mathrm{C}}_{60}^{\ast}\). To meet this condition, a two-photon process could be proposed, where the first photon excites C60 to sensitize 1O2 and the second one excites C60 to a high energy singlet state, which thereafter reacts with 1O2:

$$ {\mathrm{C}}_{60} + {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{collision}}}{\mathrm{C}}_{60} \bullet {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{h\upsilon }}{\mathrm{C}}_{60}^ \ast \bullet {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{sensitization}}}\\ \hskip 25pt {\mathrm{C}}_{60}\bullet {\,}^1{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{h\upsilon }}{\mathrm{C}}_{60}^ \ast \bullet {\,}^1{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{oxidation}}}{\mathrm{C}}_{60}\mathrm{O}$$

However, a two-photon process is already excluded by our results (i.e., Fig. 4b). Additionally, Scheme (2) can be ruled out by fundamental considerations. Even the longest τ (0.087 s in CCl4) is significantly shorter than the period between two subsequent excitations of C60, being 3.7 s (see Supplementary Information; Kinetics Model). Therefore, before the second photon absorption occurs in Scheme (2), C601O2 will relax to C603O2 with a high probability. In other words, the excited O2 and excited singlet C60 will hardly coincide in time.

Accordingly, we are urged to consider a scheme, which allows simultaneous excitation of O2 and C60, after which they coexist and react. Scheme (1) considers the most basic photosensitization event, where C60 returns to its ground singlet state after imparting its energy to O2. On the other hand, it is possible that C60 returns to an excited singlet state, \({\mathrm{C}}_{60}^{\ast}\), (if it is photoexcited to a sufficiently high energy singlet state, \({\mathrm{C}}_{60}^{{\ast}{\ast}}\)). Accordingly, Scheme (1) may be modified to:

$$ {\mathrm{C}}_{60} + {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{collision}}}{\mathrm{C}}_{60} \bullet {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{h\upsilon }}{\mathrm{C}}_{60}^{ \ast \ast } \bullet {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{sensitization}}}\\ \hskip 50pt {\mathrm{C}}_{60}^{\ast}\bullet {\,}^{1}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{oxidation}}}{\mathrm{C}}_{60}{\mathrm{O}}$$

The lowest energy 1\({\mathrm{C}}_{60}^{\ast}\) is 2.33 eV above the ground state. Additionally, 1O2 sensitization requires 0.98 eV while 0.37 eV is lost to exchange interaction during singlet-to-triplet conversion24. Therefore, a minimum excitation energy (hυ) of 3.68 eV is needed for Scheme (3) (Supplementary Fig. 4) to succeed. Consistently, our investigation using 455 and 395 nm LED excitations (2.73 and 3.14 eV) with similar photon-count exposures as in Fig. 1 yielded no detectable PO, although we confirmed 1O2 sensitization for these excitations from the 1273 nm phosphorescence (Supplementary Fig. 1). In the below two paragraphs, we experimentally corroborate 11Ag → 21Hu is the major driver of C60 PO in the solvents. Interestingly, 11Ag → 21Hu starts at 3.72 eV25, being very close to the threshold energy for PO (Scheme (3)).

In Fig. 5a, excitation spectrum for 1O2 phosphorescence (sensitization) closely follows C60 absorption spectrum from 700 nm down to 370 nm. This trend is consistent with constant and near-unity 1O2 photosensitization quantum yield by C60, Φs, as established in the literature24. Figure 5b shows the deconvolution of the excitation spectrum to Gaussians below 400 nm. Each band marks an optical transition. Although these transitions may also be resolved from optical absorption, their deconvolution is more facile from our excitation spectrum.

Fig. 5: Excitation spectra.
figure 5

a Overlay of the absorption spectrum of C60 (black) and excitation spectrum for photosensitization of 1O2 by C60, monitored from 1O2 phosphorescence counts at 1270 nm (red). b Deconvolution of the phosphorescence excitation spectrum. c Overlay of normalized kpd at different excitation wavelengths (black) and the deconvoluted 21Hu band (blue). Confidence interval error bars are shown (red) after 3 independent measurements.

Below 370 nm, however, Φs diverges from the absorption spectrum and drops significantly. On the other hand, the excitation spectrum for oxidation (i.e., kpd) in Fig. 5c (“Methods”) exhibits a reverse trend. The PO rate, kpd, is essentially zero for the spectral range, where Φs is at its maximum value of unity, but it is activated at the threshold of about 335 nm (3.70 eV), at which Φs is reduced to 0.37. Hence, the excitation trends in Fig. 5a, c, being spectrally different, underscore the fact that sensitization of 1O2 is not sufficient for the oxidation of C60. Specifically, as seen in Fig. 5c, the kpd spectrum matches the 21Hu band. These findings validate Scheme (3) as well as 11Ag → 21Hu being the major driver of PO. Here, the normalized kpd values were derived from the 1O2 phosphorescence intensity (i.e., counts proportional to [C60]) kinetics (Supplementary Fig. 3).

While 11Ag → 21Hu is the major driver of C60 PO (Scheme (3)), 11Gu, 11T1u, and 21Gu states can also be excited to their vibronic levels higher than 3.7 eV (from 11Ag), as inferred from their deconvoluted bands in Fig. 5b. However, vibrational relaxation (VR) is the fastest process, quickly quenching 11Gu, 11T1u, and 21Gu to their ground vibrational levels at 3.12, 3.40, and 3.43 eV, respectively (Fig. 6). Hence, ISC from these singlet states at above 3.7 eV is expected to be outcompeted by VR. Alternatively, PO (Scheme (3)) is possible from 11Gu, 11T1u, and 21Gu vibronic states, if they transition to 21Hu by internal conversion (IC) before VR to below 3.7 eV. Because both IC and ISC are slower than VR by an order of magnitude or more, PO from 11Gu, 11T1u, and 21Gu vibronic states will be minor, but may not be negligible. In conclusion, the major PO is expected to be through direct excitation of 21Hu. Accordingly, kpd spectrum (data points) in Fig. 5c follows the 21Hu band. However, some deviation is seen, being highest for the 3.88 eV (320 nm) data point and over the 21Hu Gaussian, suggestive of additional excitations contributing, possibly through 11Gu, 11T1u, and 21Gu as discussed above.

Fig. 6: Jablonski diagram illustrating photooxidation of C60.
figure 6

To a first approximation, we adopt the energy structure of isolated C60 for the C60 of C60•O2. The gray curve represents the absorbance spectrum of C60.

We illustrate Scheme (3) with the gray arrows in the Jablonski diagram of Fig. 6. First, C60 is photoexcited through 11Ag(1C60) → 21Hu(1\({\mathrm{C}}_{60}^{{\ast}{\ast}}\)). Then, 1\({\mathrm{C}}_{60}^{{\ast}{\ast}}\) transitions to 3\({\mathrm{C}}_{60}^{{\ast}{\ast}}\) (21Hu → 23Hu) via ISC. Subsequently, triplet–triplet annihilation (TTA)26 occurs with sensitization of 1O2:\(2^{3}{\mathrm{H}}_{\mathrm{u}}\bullet{\,}^{3}{\mathrm{O}}_{2}\mathop{\rightarrow}1^{1}{\mathrm{T}}_{\mathrm{1g}}\bullet{\,}^{1}{\mathrm{O}}_{2}\). TTA also leaves C60 at an excited singlet state (11T1g), which can readily react with 1O2:\(1^{1}{\mathrm{T}}_{\mathrm{1g}}\bullet{\,}^{1}{\mathrm{O}}_{2}\mathop{\rightarrow}{\mathrm{C}}_{60}{\mathrm{O}}\). Hence, both 11T1g and 1O2, two energetic species, are created at the same time and same place (in collision complex) and have a higher chance to react. Another useful interpretation is that a single photon’s energy (\(h\upsilon\)) is partially utilized in sensitizing 1O2 while the excess energy leaves C60 at an excited state, which can react with 1O2. Scheme (3) may be expressed in more detail as:

$$ {\,}^{1}{\mathrm{C}}_{60} + {\,}^{3}{\mathrm{O}}_{2} \to {\,}^{1}{\mathrm{C}}_{60} \bullet {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{UV}}}{\,}^{1}{\mathrm{C}}_{60}^{ \ast \ast } \bullet {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{ISC}}}\\ \hskip 25pt {\,}^{3}{\mathrm{C}}_{60}^{ \ast \ast }\bullet {\,}^{3}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{\mathrm{TTA}}}{\,}^{1}{\mathrm{C}}_{60}^{\ast} \bullet {\,}^{1}{\mathrm{O}}_{2} \to {\mathrm{C}}_{60}{\mathrm{O}}$$

We provide a mathematical analysis of PO kinetics for Scheme (3), which considers all the steps as well as reverse/competing processes, such as 1O2 quenching, relaxation of \({\mathrm{C}}_{60}^{\ast}\), and complex dissociations (see Supplementary Information; Kinetics Model). Despite the full complexity of this model, it predicts simply an exponential decay for [C60], being consistent with the measured kinetics. The model allows us to write kpd as a function of 1O2 quenching rate, kq = 1/τ. Fitting of this function to experimental data (Fig. 7) reveals the rate constants, kox and kr, associated with \({\mathrm{C}}_{60}^{\ast} \bullet {\,}^{1}{\mathrm{O}}_{2}\mathop { \to }\limits^{k_{\mathrm{ox}}} {\mathrm{C}}_{60}{\mathrm{O}}\) and \({\mathrm{C}}_{60}^{\ast} \bullet {\,}^{1}{\mathrm{O}}_{2}\mathop{\longrightarrow}\limits^{{k_r}}{\mathrm{C}}_{60} \bullet {\,}^{1}{\mathrm{O}}_{2}\), respectively, which has an interesting implication, as discussed below.

Fig. 7: Fitting of theoretical kpd expression to experimental data.
figure 7

The fitted values of kox and kr are given in the inset.

Mechanism of C60 photooxidation in the atmosphere

PO of C60 is expected to be significantly accelerated in the atmosphere thanks to dramatically prolonged τ in the air (i.e., tens of minutes). Indeed, PO can dominantly occur as a two-photon process (Scheme (2), illustrated in Supplementary Fig. 5) driven by visible and UVA photons, being abundant in solar radiation. Unlike in solvents, it is challenging to monitor PO of C60 in air, since C60 being at detectable concentrations in air, would quickly undergo aggregation as well as adsorption to enclosure walls. However, kpd can be predicted from the rate constants, kox and kr, which are already captured in the present work from C60 dispersions in solvents. As such, we compute kpd = 0.011 s−1 for AM 1.5 solar radiation, suggesting a half-life of 63 s (see Supplementary Information; Kinetics Model). This rapid PO of C60 in the atmosphere potentially explains its scarceness in the environment27.

Two excited singlets by triplet–triplet annihilation

By its description in the literature, TTA involves Dexter energy transfer from a triplet to another, after which the acceptor transitions to a higher energy state (singlet), while the donor returns to its ground singlet state. On the other hand, Scheme (3) engages a unique TTA, which produces two excited singlets simultaneously and thereby enables an efficient photochemistry. This scheme is stimulating for the conception of novel efficient photochemical processes, implemented with C60 or other photosensitizers.


C60 solution preparation

A stock solution of C60 was first prepared by dissolving 1 mg of C60 (Thermo Fisher Scientific, >99.9%) in 10 mL of solvent via ultra-sonication for 30 min. Next, the solution was kept undisturbed in the dark for 30 min to let insoluble aggregates (e.g., C60O) settle down. Subsequently, the supernatant was transferred by a pipette to a spectrophotometer cell (optical path length of 10 mm) filled with the same solvent until the absorbance of C60 at 256 nm reaches 1.42 (as monitored by a spectrophotometer), corresponding to C60 concentration of 5.67 μM. Finally, the prepared solution was stored in a sealed glass vial and kept in the dark before use. High purity of C60 in the prepared C60 solution is confirmed by mass spectrometry (Supplementary Fig. 2).

Excitation spectroscopy for photooxidation

Photooxidation (photodecay) rate of C60, kpd, was measured as a function of excitation wavelength using the Fluorolog-3 spectrofluorometer. In a typical acquisition, 100 μL of C60 in CCl4 solution was placed in a standard microfluorescence cuvette (Science Outlet, 10 mm optical length, 0.35 mL capacity) and excited at selected wavelengths (from 250 to 420 nm, at 10 nm intervals) using a bandpass of 5 nm. The emission (1O2 phosphorescence) was parked at 1270 nm with a bandpass of 30 nm. At each excitation wavelength, the acquisition was performed 3 times and an unexposed sample was employed at each acquisition. The time-series phosphorescence intensity was collected in-situ at every 2 s with detector integration time of 2 s. Hence, the monochromatic optical beam of the spectrometer served as a dual probe simultaneously for measurement and exposure. For each excitation wavelength, kpd was derived from the exponential decay rate of the phosphorescence intensity, quantifying the decay rate of C60 concentration (see Supplementary Fig. 3).