Edited by Neta A. Bahcall, Princeton University, Princeton, NJ, and approved November 2, 2011 (received for review August 31, 2011)
Buckminsterfullerene (C60) was recently confirmed as the largest molecule identified in space. However, it remains unclear how and where this molecule is formed. It is generally believed that C60 is formed from the buildup of small carbonaceous compounds in the hot and dense envelopes of evolved stars. Analyzing infrared observations, obtained by Spitzer and Herschel, we found that C60 is efficiently formed in the tenuous and cold environment of an interstellar cloud illuminated by strong ultraviolet (UV) radiation fields. This implies that another formation pathway, efficient at low densities, must exist. Based on recent laboratory and theoretical studies, we argue that polycyclic aromatic hydrocarbons are converted into graphene, and subsequently C60, under UV irradiation from massive stars. This shows that alternative—top-down—routes are key to understanding the organic inventory in space.
The midinfrared spectra of a variety of astrophysical objects are dominated by band emission (strongest at 3.3, 6.2, 7.7, 8.6, and 11.2 μm) attributed to carbonaceous macromolecules [i.e., polycyclic aromatic hydrocarbons (PAHs)] (1). These molecules are large (30–100 C atoms), abundant (approximately 5% of the elemental carbon), and their ionization plays a key role in the energy balance of gas in the interstellar medium (ISM) and in protoplanetary disk. In addition to PAH bands, infrared signatures observed at 7.0, 8.5, 17.4, and 19.0 μm have been reported recently (2, 3) and found to coincide precisely with the emission of buckminsterfullerene (C60) (4), a cage-like carbon molecule. This detection heralds the presence of a rich organic inventory and chemistry in space. However, observed abundances of C60 challenge the standard ion-molecule or grain-surface chemistry formation routes, which build up molecules from small to large in the ISM. For that reason, it has been suggested that C60 is formed in the hot and dense envelopes of evolved stars (5–7) in processes similar to those found in sooty environments (8–11), and eventually, is ejected in space. Yet, this scenario faces the problem that it has a limited efficiency (6). PAHs and C60 are known to coexist in the ISM (3); however, so far, the connection between PAHs and C60—and in particular the possibility to go from one compound to the other in space—has not been investigated. In this paper, we present a study of PAH and C60 chemical evolution in the NGC 7023 nebula, using Spitzer (12) and Herschel (13) infrared observations.
Earlier Spitzer observations of the NGC 7023 reflection nebula have revealed a chemical evolution of PAHs: Deep in the cloud, emission is dominated by PAH clusters, which evaporate into free-flying PAHs when exposed to the UV radiation from the star (14–16). There, gaseous PAHs are, in turn, ionized. While the neutral PAHs are dominated by zig-zag edges—as demonstrated by the strong C-H solo out-of-plane modes—the ions have an armchair molecular structure, characterized by strong duo out-of-plane modes (17). In regions closest to the star, the presence of C60 in the neutral state is evidenced by Spitzer observations (Fig. 1). New Herschel observations provide a measurement of dust emission in the same region, at high angular resolution (Fig. 1). This measurement can be used to derive the integrated intensity radiated by the nebula, which can be used as a calibrator, to convert the Spitzer observations of the PAHs and C60 bands into absolute chemical abundances of these species, allowing a quantitive study of PAH and C60 chemical evolution.
Overview of the NGC 7023 nebula. (A) Multiwavelength color-coded view of the nebula in the infrared. Red is the emission at 70 μm observed with the Photodetector Array Camera end Spectrometer (PACS) onboard Herschel. This emission is dominated by dust. Green is the Spitzer-Infrared Array Camera (IRAC) 8-μm emission tracing the PAH C-C mode, and blue is the IRAC 3.6-μm emission, tracing the PAH C-H mode and stellar emission. The position of the intermediate mass young star HD 200775 illuminating the nebula is indicated. (B) Color-coded image of the spatial distribution of different compounds in NGC 7023 : Red is the emission of dust observed with Herschel-PACS, green shows the emission integrated in the 6.2 μm C-C band of PAHs observed with Spitzer-Infrared Spectrograph (IRS), and blue is the C60 emission observed with Spitzer-IRS integrated in the 19.0-μm band. The white rectangle shows the region on which the extraction of the PAH and C60 abundance (Fig. 2) was performed. (C) Spitzer IRS midinfrared spectra taken at positions (1) (Upper) and (2) (Lower) in B. The distance from the star at positions (1) and (2) are, respectively, approximately 35′′ and 15′′. The bands of PAHs and C60 are labeled in the spectra.
Abundances of C60 (A) and PAHs (B) in NGC 7023 as a function of distance from the star in the cut shown in Fig. 1. The red curves give the 1 sigma uncertainty, obtained from the propagation of instrumental uncertainty on the extraction of integrated intensities of PAHs and C60 bands.
The far infrared integrated intensity IFIR was extracted by fitting the spectral energy distribution (SED) at each position in the cross-cut shown in Fig. 1. For these positions, we have used the brightnesses as measured by Herschel Photodetector Array Camera end Spectrometer (PACS) (18) and SPIRE (19) photometers. We have used the 70- and 160-μm channels of PACS and the 250-μm channel of SPIRE. This data is presented in detail in ref. 20. The modified blackbody function fit to these SEDs containing three spectral points is defined by 
[1]where K is a scaling parameter, λ is the spectral index, and B(λ,T) is the Planck function with λ the wavelength and T the temperature. We have used a constant value of 1.8 for β so that only T and K are free parameters. The results of the fits to the observations are shown in Fig. S1. The temperatures we have derived from the fit of the data range between 25 and 30 K. These values are in agreement with those of ref. 20 for the same region. The peaking position of the modified blackbody function moves to shorter wavelengths (i.e., the grain temperature increases) when getting closer to the star, implying that we are indeed tracing matter inside the cavity and not material behind on the line of sight. The far infrared integrated intensity, IFIR, is then derived by integrating I(λ,T) over frequencies.
The integrated intensity of PAH emission, IPAH, is measured by fitting a PAH emission model to the observed Spitzer midinfrared low-resolution spectrum. This data covers the 5- to 14-μm range where most of the emission occurs. Because this spectral range contains the emission due to the vibration of both C-C and C-H bonds, it is insensitive to ionization of PAHs. An example of this fitting model is presented in ref. 21. This tool provides the integrated intensity in the PAH bands as an output and takes care of continuum subtraction and extinction correction. The C60 integrated intensity (IC60) extraction needs to be done separately for each band. We first measure the integrated intensity in the 19- and 17.4-μm band. The first step consists in extracting the intensity in the 19.0-μm band (I19.0) by fitting a Gaussian and subtracting a local linear continuum. The 17.4-μm band is contaminated by PAH emission, but this can be removed effectively. The contribution of PAHs to the 17.4-μm band can be estimated in this way: In the outer regions of the nebula, the 17.4-μm band is 100% due to PAHs (because no C60 is detected there) and so is the 16.4-μm band. The 16.4- and 17.4-μm band of PAHs are known to correlate. Indeed, in the outer regions of the nebula we find 
. Therefore, over the whole nebula, we can estimate the intensity of the 17.4-μm band due to C60 by 
. Fig. S2 shows the result of this process on the map. The 7.0-μm band is faint and usually hard to detect. In the regions closest to the star where both the 19- and 7.0-μm features are observed, we can calibrate the ratio of I19.0/I7.0. It is found to be relatively stable (at least in this small zone close to the star) and of the order of 0.4. Therefore, we use I7.0 = 0.4 × I19.0. The 8.5-μm band is undetectable because of the strong PAH band present at 8.6 μm, so we use the ratio provided in ref. 3 for 5 eV photons; that is, I8.5 = 0.4 × I19.0. The total C60 IR emission is hence given by IC60 = I19.0 + I17.4 - I16.4 × 0.35 + I19.0 × 0.4 + I19.0 × 0.4.
The method to derive the abundance of carbon locked in PAHs and C60 from IPAH, IC60, and IFIR is presented in detail in ref. 22. The fraction of carbon locked in PAHs and C60 (respectively, 
and 
) per atom of interstellar hydrogen are given by 
[2]and 
[3]where 
[4]and 
[5]
and 
are the UV absorption cross-sections of PAHs and C60. Following ref. 22, we adopt 
per C atom. There are no detailed measurements of the UV absorption cross-section of C60; following ref. 3, we adopt the same value as for PAHs.
The results of the abundance variations within the nebula as derived using the above method are shown in Fig. 2. The C60 abundance in the nebula is seen to increase from 1.4 × 10-4 to 1.7 × 10-2% of the elemental carbon abundance when approaching the star (Fig. 2). On the other hand, the abundance of PAHs is seen to decrease, from 7.0 to 1.8% of the carbon (Fig. 2). This shows that C60 is being formed in the nebula while PAHs are being destroyed or processed. The correlation of these variations with the increasing UV field strongly suggests that UV photons control C60 formation and PAH processing and destruction.
The formation of C60 in the ISM is unexpected and has several implications in our understanding of the formation process of this molecule. In the laboratory, the main formation route invoked for C60 is the buildup from atomic carbon, small carbon clusters or rings (8–11). In space, such processes are efficient in the hot (1,500 K) and dense (nH > 1011 cm-3) envelopes of evolved stars (6). In NGC 7023, the gas is approximately seven orders of magnitude less dense (see ref. 23 and SI Text), making this aggregation processes inefficient. Instead, we invoke photochemical processing of PAHs as an important route to form C60. Upon UV irradiation, several channels for fragmentation can be open depending on excitation energy; e.g., H loss and C2H2 loss (24). However, experiments on small PAHs have shown that H loss is by far the dominant channel (25) and that complete dehydrogenation [i.e., graphene (26) formation] is then the outcome of the UV photolysis process (27, 22, 28, 29). In a second step, carbon loss followed by pentagon formation initiates the curling of the molecule (30). We envision that this is followed by migration of the pentagons within the molecule, leading to the zipping-up of the open edges forming the closed fullerene (31) (Fig. 3 and Movie S1). Graphene formation through PAH photolysis can give rise to a rich chemistry. Besides the route toward fullerenes outlined above, fragmentation toward small cages, rings, and chains may also result. The relative importance of isomerization and fragmentation will determine the carbon inventory delivered by the photochemical evolution of PAHs in space (Fig. 3).
Schematic representation of top-down interstellar carbon chemistry. (A) The chemical evolution of PAHs in the ISM under the influence of UV photons combines the effects of dehydrogenation and fragmentation with those of isomerization. Fully hydrogenated PAHs—injected by stars into the ISM—are at the top right side. Near bright stars, UV photolysis will preferentially lead to complete H loss (e.g., the “weakest link”) and the formation of graphene. Further fragmentation may lead to the formation of flats, rings, and chains. However, this process competes with isomerization to various types of stable intermediaries such as cages and fullerenes. (B) Schematic illustration of conversion of graphene into C60 in 13 steps. Dehydrogenated PAHs (i.e., graphene sheets) loose carbon atoms under UV irradiation, giving rise to pentagonal defects (represented with orange C atoms) at the edges of the sheet. These defects in the hexagonal network induce curvature of the sheet, the migration of the pentagons allows the molecule to close. Image courtesy of L. Cadars (www.laurecadars.com).
Schematically, the fragmentation process can be written as 
[6]where PAH∗ is the excited species that can stabilize through emission of IR photons or through fragmentation and PAH-H is a dehydrogenated PAH radical. There are various ways to evaluate the unimolecular dissociation rate constant for this process (compare ref. 22, section 6.4). We will follow ref. 32 and write the rate constant in Arrhenius form, 
[7]where Te is an effective excitation temperature, Eo is the Arrhenius energy describing the process, and the preexponential factor ko depends on the interaction potential (in the reverse reaction). For PAHs, the internal excitation of the vibrational modes after the absorption of a UV photon can be well described by 
[8]where Nc is the number of C atoms and E is internal energy in eV (ref. 22, p. 184). Because typically the energy involved in these reactions is a fair fraction of the total energy in the system, a correction has to be made to this excitation temperature. The finite heat bath correction results in (22, 32) 
[9]The preexponential factor can be set equal to 
[10]with ΔS the entropy change for which we will adopt 5 cal/K (28). ko is then, typically, ≃3 × 1016 s-1. The Arrhenius energy parameter, Eo, cannot be easily evaluated from theoretical calculations (33). Here, we use a fit to the experimental fragmentation studies on small PAHs [< 24 C atoms; (25)], which results in Eo = 3.3 eV (ref. 22, p. 204). The probability for dissociation depends then on the competition between fragmentation and IR photon emission, 
[11]where kir(E) is the IR emission rate at an internal energy E. For a highly excited PAH, kir(E) is approximately 1 s-1. The total fragmentation rate is then 
[12]where kuv(E) is the absorption rate of UV photons with energy, E.
The photochemically driven H loss is balanced by reactions of atomic hydrogen with dehydrogenated PAHs; namely, 
[13]The rate of this reaction has been measured to be ka = 1.5 × 10-10 cm3 s-1 for a number of small PAHs (28, 34). We can define the dissociation parameter, 
. With kuv = 7 × 10-10NcGo, where Go is the flux of UV photons in units of the interstellar Habing field (35), we have ψ ≃ 0.2nH/NcGopd(E). The hydrogen coverage of interstellar PAHs is a very sensitive function of ψ, and a small increase of ψ can change PAHs from fully hydrogenated to graphene (22, 34). This is illustrated for the circumovalene, C60H20, in Fig. S3. Thus, PAHs are fully hydrogenated if ψ is much less than 1 and fully dehydrogenated if ψ is much larger than 1. Hence, 
[14]provide a critical relation for the transformation of PAHs into graphene. We have evaluated this relation assuming an absorbed UV photon energy of 10 eV (Fig. 4).
Photochemistry of PAHs in NGC 7023 (A) The red curve shows the evolution of the critical value when PAHs have lost half of their H atoms, as a function of the physical conditions (G0/nH where G0 is the radiation field in Habing units and nH is the density of H atoms in the gas in cm-3; see SI Text for details) and the number of carbon atoms in the PAH molecule NC. Above this line, PAHs are stable against dehydrogenation. Below this curve, PAHs rapidly loose all their H atoms. The value of G0/nH for NGC 7023 is shown; dark blue represents the “narrow” range, and clear blue represents the “broad” range of values for this parameter (SI Text). (B) Timescale for the loss of a carbon atom by UV photon absorption as a function of graphene sheet size.
After formation of graphene, UV photoabsorption can lead to loss of carbon from the skeleton. This fragmentation process competes with stabilization through IR photon emission. The reaction rate is again given by Eq. 7. We adopt ΔS = 5 cal/K, resulting in ko ≃ 3 × 10-16 s-1, but the exact value is not critical. For Eo, we have adopted the Arrhenius energy [3.65 eV (36)] derived from experiments on the C loss from small catacondensed PAHs (25). The calculated cohesive energy of carbon in graphene is much larger, 7.4 eV per C atom (37) but that refers to typically carbon inside the skeleton, and carbon at the edge will be less strongly bound. Ref. 31 finds a theoretical value of 5.4 eV for the C atoms at a zig-zag edge. Moreover, theoretical cohesive energies are not a good measure for the Arrhenius energy in unimolecular dissociation experiments (33). Given the low abundance of gas phase carbon, the reverse reaction is unimportant and the chemical lifetime [
] has to be compared to the dynamical lifetime of the region. We have evaluated this chemical lifetime for a typical internal energy of 10 eV as a function of the size of the graphene sheet, Nc, and the results are shown in Fig. 4.
To evaluate graphene formation in NGC 7023, we look at PAH stability. The physical conditions in the nebula, characterized by parameters G0 and nH, can be obtained (SI Text). As shown in Fig. 4, H loss is very efficient in NGC 7023 for PAHs up to 70 C atoms. We then evaluate the timescale for the loss of C atoms by the graphene flakes and compare this to the dynamical age of the nebula for a ≃70 C atom PAH (Fig. 4). It appears that if we adopt Eo = 3.65 eV, C loss is rapid compared to the age of the nebula.
Our models reveal that 70 C atoms PAH become unstable to graphene formation in the studied region of NGC 7023. Much larger PAHs will survive closer to the star, whereas smaller PAHs will rapidly be completely dehydrogenated further away, but—because of size—this will not lead to the formation of fullerenes. As shown in laboratory experiments (31), graphene sheets larger than about 70 C atoms can be transformed into fullerene, but in space this is driven by UV photons rather than energetic electrons. We surmise that fullerene formation is initiated by single C atom loss (38) at the zig-zag edge of the graphene flake. Quantum chemical calculations indeed show that the armchair structure of graphene clusters stabilizes through in plane π-bond formation, whereas the open-shell orbitals associated with the dangling bonds are located at the zig-zag edge (37), making these atoms more labile (31). A detailed study of graphene edge structure (39), based on experimental results (38), has indeed shown that single C atom loss leads to the formation of reconstructed edges bearing pentagons. These defects dramatically stress graphene flakes, inducing significant positive curvature in its topology (30). Eventually, the barrierless migration of the pentagons within the hexagonal network will lead to the zipping-up of the flake into C60 (31) (Fig. 3). The above mechanism is severely limited by the physical conditions, controlling dehydrogenation, and the size of the precursors, which need to be above approximately 70 C atoms. As shown by our observations, only about 1% of the available PAHs are converted into C60 by this process (Fig. 2). The activation of this “top-down” chemistry requires the high UV fields available near massive stars. As a corollary, high abundance of C60 in the diffuse ISM (40) reflects processing by massive stars, which are capable of dehydrogenating PAHs in the relevant size range (60–70 C atoms). PAHs much larger than than 70 C atoms will survive even in close encounters with massive stars. Smaller-size PAHs (< 50 C atoms) cannot reach the fullerene island of stability, hence their photoprocessing (Fig. 2) must be a source of various carbon nanocompounds (45) (e.g., rings, chains, cages, bowls, tubes, etc.) (Fig. 3). Laboratory experiments have indeed shown that processing of small graphene constrictions can lead to carbon rings and chains (41). Finally, we note that recent studies suggest that PAH molecules do not provide a satisfactory explanation to the diffuse interstellar bands (42, 43). The compounds described in this top-down chemistry may be relevant to understand these features of the UV extinction curve, in the Milky Way and in other galaxies (44).
Analyzing infrared observations of the NGC 7023 nebula, we have found evidence that C60 is being formed in the ISM. Classical bottom-up formation routes fail to explain these observations, so we propose a chemical route in which C60 is formed directly by the photochemical processing of large PAH molecules. Other carbonaceous compounds (cages, tubes, bowls) are also the product of this photoprocessing. This route must be relevant in the ISM, but may also be important in the inner regions of protoplanetary disks around solar-type stars where accretion on the young star generates intense UV fields and PAHs are known to be present. There, they can be converted, on timescale of approximately 1 million years, into fullerenes and hydrocarbon fragments. This source of organic compounds remains to be considered in models studying the organic photochemistry of the ISM and regions of terrestrial planet formation. It is clear that such studies may benefit greatly from the progress achieved in the field of graphene stability.
The authors thank the referees for their constructive comments. Laure Cadars (www.laurecadars.com) is acknowledged for producing the C60 formation sketch. Studies of interstellar PAHs at Leiden Observatory are supported through advanced European Research Council Grant 246976.
Author contributions: O.B. performed research; A.G.G.M.T. performed modeling; A.G.G.M.T. wrote section on dehydrogenation; O.B. analyzed data; and O.B. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1114207108/-/DCSupplemental.
