Resonance-stabilized hydrocarbon-radical chain reactions may explain soot inception and growth

See allHide authors and affiliations

Science  07 Sep 2018:
Vol. 361, Issue 6406, pp. 997-1000
DOI: 10.1126/science.aat3417

A radical route to soot

The chemical origin of soot is a persistent puzzle. It is clear that small hydrocarbon fragments formed in flames must aggregate into larger particles, but the initial driving force for aggregation remains a mystery. Johansson et al. combined theory and mass spectrometry to suggest a solution based on resonance-stabilized radicals (see the Perspective by Thomson and Mitra). Aromatics such as cyclopentadiene have a characteristically weak C–H bond because their cleavage produces radicals with extended spans of π-electron conjugation. Clusters thus build up through successive coupling reactions that extend conjugation in stabilized radicals of larger and larger size.

Science, this issue p. 997; see also p. 978


Mystery surrounds the transition from gas-phase hydrocarbon precursors to terrestrial soot and interstellar dust, which are carbonaceous particles formed under similar conditions. Although polycyclic aromatic hydrocarbons (PAHs) are known precursors to high-temperature carbonaceous-particle formation, the molecular pathways that initiate particle formation are unknown. We present experimental and theoretical evidence for rapid molecular clustering–reaction pathways involving radicals with extended conjugation. These radicals react with other hydrocarbon species to form covalently bound complexes that promote further growth and clustering by regenerating resonance-stabilized radicals through low-barrier hydrogen-abstraction and hydrogen-ejection reactions. Such radical–chain reaction pathways may lead to covalently bound clusters of PAHs and other hydrocarbons that would otherwise be too small to condense at high temperatures, thus providing the key mechanistic steps for rapid particle formation and surface growth by hydrocarbon chemisorption.

Soot is produced during incomplete combustion of hydrocarbon fuels. Carbonaceous interstellar particles, containing ~70% of the carbon in interstellar space, are formed under similar chemical conditions (15). Terrestrial soot has enormous impact on human health and the environment (6); it plays a major role in deaths attributed to air pollution worldwide (7) and is an important contributor to global warming (8). Notably, the identity of the chemical mechanisms for soot and interstellar dust formation is a scientific puzzle analogous to a former long-standing challenge of understanding particle nucleation in Earth’s atmosphere. The key pathway to new-particle formation and growth from biogenic vapors was only recently discovered (9, 10).

There is a large body of evidence linking polycyclic aromatic hydrocarbon (PAH) species to soot formation (1113). Key missing steps in the understanding of soot formation include inception (i.e., the production of hydrocarbon clusters that mark a transition to the condensed phase) and the continuation of rapid growth by hydrocarbon addition to these incipient particles (i.e., hydrocarbon clusters). There are two main classes of mechanisms for soot inception under active research: physical bonding of PAHs by dispersive or van der Waals forces into stacked clusters and formation of covalently bound clusters (12). At temperatures where soot inception occurs (≥1450 ± 250 K), the PAHs typically present in appreciable quantity contain two to five aromatic rings and are far too volatile to nucleate by van der Waals forces (12); a long history of research indicates a need for a chemical mechanism that rapidly bonds PAHs covalently at these temperatures. No detailed chemical pathways have been posited, however, and current chemical models for covalent-cluster formation generally include proxy reactions to quickly and irreversibly bond highly stable PAHs in violation of the second law of thermodynamics (12). Such reactions tend to necessitate repeated activation of stable PAHs through H abstraction via substantial barriers (forcing rates to be too slow), yield nearly uniform growth among all PAHs instead of cluster formation, and quench the radical pool (12). The proposed mechanisms and associated issues are similar for particle-surface growth. These issues are also relevant to the understanding of stardust formation near carbon-rich stars because the chemistry and precursor species of interstellar-dust formation are thought to be very similar to those of terrestrial soot (14).

Here we present theoretical and experimental evidence for a radical-driven hydrocarbon-clustering mechanism that could provide a physically viable route to particle formation and growth. This mechanism involves chain reactions of resonance-stabilized radical (RSR) species that can form covalent bonds with a wide variety of hydrocarbons to produce a clustered product with RSR character. The mechanism is initiated and propagated by RSRs that require an unpaired electron for extended conjugation. Products of reactions of these RSRs with other radical and closed-shell hydrocarbons readily regenerate RSRs that maintain extended conjugation via an unpaired electron. Hence, loss and gain of extended conjugation promote these RSR-driven radical-chain reactions. This radical-regeneration process maintains a pool of radicals that can undergo molecular growth and act as clustering nuclei for particle formation. We refer to this mechanism as clustering of hydrocarbons by radical-chain reactions (CHRCR). Because larger carbonaceous particles are likely to have surface-radical sites with chemical traits similar to those of the RSRs, we hypothesize that these pathways may also contribute to particle-surface growth under high-temperature conditions.

Figure 1 provides an overview of the CHRCR mechanism in three stages. The first stage (Fig. 1A) increases the size of an RSR through acetylene (C2H2) or vinyl (C2H3) addition, via radical-chain reactions. We chose to start this stage with cyclopentadienyl, but propargyl (C3H3 at 39 u) is also an RSR, which generates cyclopentadienyl through reaction with acetylene (14), and thus precedes cyclopentadienyl in the sequence. The second stage (Fig. 1B) involves hydrocarbon clustering via radical-chain reactions initiated by the RSRs from the first stage. The third stage (Fig. 1C) is an extension of the second and might promote particle surface growth by RSR-driven radical-chain reactions at sites on the particle surface.

Fig. 1 Schematic overview of the clustering of hydrocarbons by radical-chain reactions (CHRCR) mechanism.

(A) The CHRCR mechanism is initiated and propagated by resonantly stabilized radicals sequentially generated through radical-chain reactions involving acetylene or vinyl. (B) To form an incipient particle, these RSRs can cluster a wide range of hydrocarbons, including radicals, stable PAHs, and unsaturated aliphatic species, through radical-chain reactions fueled by loss and gain of extended conjugation. (C) Cyclopentadienyl-type moieties on cluster surfaces are posited to further propagate growth via the CHRCR mechanism.

The first stage (Fig. 1A) involves molecular growth of the pool of RSRs that act as seeds for initiating hydrocarbon clustering. Our electronic-structure calculations predict a sequence of RSRs readily generated through radical-chain reactions. Each reaction adds acetylene (C2H2) or vinyl (C2H3) to an RSR and generates a new RSR (see supplementary materials). Hence, these radicals are not consumed when reacting with C2H2 and C2H3. Instead they grow to larger species with similar chemical traits and could, therefore, survive long enough to act as clustering seeds for other hydrocarbons. The species generated by these reactions, starting from 65 u, have masses of 91, 115, 141, 165, 189, 215, 239, and 263 u and, likely, higher values. These masses are frequently observed experimentally in condensed samples extracted from flames (13, 1518). Acetylene is generally the most abundant hydrocarbon under high-temperature conditions in space and on Earth (3, 11). This stage of the mechanism thus has strong theoretical and experimental support.

We used a vacuum ultraviolet aerosol mass spectrometer (VUV-AMS) to measure these species on particles extracted from flames. A mass spectrum highlighting the RSR sequence is shown in Fig. 2. The sample was extracted from an atmospheric-pressure laminar premixed ethylene-oxygen flame. Laminar premixed flames provide stable and reproducible flame conditions with controllable fuel to oxidizer ratios. Fuel-rich premixed combustion is relevant to the first stages of soot formation under diesel-engine conditions (19), and ethylene is a standard fuel for soot-formation studies because of its simple structure and high propensity to produce soot. Most natural flames are formed under diffusion-controlled conditions. In the supplementary materials, we present similar results from both diffusion and premixed flames with several different fuels. Concentrations suggested by peak intensities for the RSRs are biased low relative to heavier closed-shell species because many radical-quenching reactions preclude detection or yield signal at different masses.

Fig. 2 VUV-AMS spectrum demonstrating a sequence of radicals.

This VUV-AMS spectrum was recorded on a sample extracted from an atmospheric-pressure laminar premixed ethylene-oxygen flame using a photon energy of 9.4 eV. Masses are indicated for species posited to drive the CHRCR mechanism. Our best estimates of the predominant structures are shown. Four hypothesized isomers are shown for mass 165 u. There are many more potential isomers for higher radical masses not shown.

Whereas prior work has assumed that the mass peak at 91 u (R91) stems from benzyl, we identified R91 as vinylcyclopentadienyl from its ionization energy in both ethylene- and acetylene-fueled flames (see supplementary materials). Likely formation pathways include vinyl or acetylene addition to cyclopentadienyl (Fig. 1A). The large numbers of isomers for species heavier than R91 preclude experimental identification; however, the mass sequence highlighted in Fig. 2 is consistent with the predicted RSRs.

The second stage of the mechanism (Fig. 1B) describes clustering of available hydrocarbons by the initiator radicals produced in the first stage. Figure 3 shows an example of a pathway commencing with reaction between indenyl and phenyl to form a covalently bound closed-shell dimer (i.e., σ-dimer). Indenyl (C9H7) is a resonantly stabilized π-radical whose extended conjugation depends on maintaining an unpaired π electron. Hence, indenyl undergoes partial loss of conjugation during σ-dimerization (the site of phenyl attack changes hybridization from sp2 to sp3). The dimerization process is barrierless, but because indenyl loses conjugation during dimerization, the energy is lowered substantially less than a typical C-C bond energy. The loss of conjugation means that the barrier to subsequent H abstraction from the α carbon of the five-membered ring, i.e., the carbon atom to which phenyl attached (using R=H), can be as low as ~4.5 kcal/mol (Fig. 3 shows two abstraction sites) because H loss leads to extended conjugation. This energy landscape also promotes direct H ejection from the chemically activated dimer. H-atom abstraction or ejection generates a new RSR, similar in character to indenyl, whose extended conjugation stabilizes its unpaired electron and reduces the bond length between the dimerized entities in Fig. 3 from ~1.51 to ~1.45 Å. Several of the CHRCR reaction pathways could proceed without radical consumption or with net radical production (see supplementary materials).

Fig. 3 Representative potential energy surface for hydrocarbon clustering.

The potential energy diagram shows formation of an indenyl-phenyl σ-dimer and subsequent H abstraction, which stabilizes the bond and regenerates an RSR whose extended conjugation requires an unpaired electron (see supplementary materials for details on the calculations). These reactions are examples of possible steps in the hydrocarbon-clustering scheme proposed. Phenyl represents an arbitrary PAH radical; the cluster is shown for illustrative purposes. The y axis displays differences in the zero-point–corrected electronic energies. Abstraction steps, energy changes, and barriers are evaluated with R=H. The total number of H atoms was identical for all of the calculated energies, but, in the figure, species involved during the H-abstraction reactions are only shown when active.

The barrierless dimerization process and low H-abstraction barrier provide the energy landscape required for fast reactions. Dimerization and H abstraction together release more than 100 kcal/mol (zero-point corrected electronic energies), minimizing reversibility. In addition to reacting with a PAH radical, indenyl can attack an unsaturated aliphatic species (see supplementary materials) or a nonradical aromatic ring. The latter reaction, however, leads to loss of conjugation in both indenyl and the aromatic ring, yielding a weakly bound dimer that requires fast stabilization via H loss.

These reactions represent a repeatable class of pathways for cluster formation. The clustering is accelerated by adding RSRs whose extended conjugation requires an unpaired electron and benefits from previously proposed gas-phase growth mechanisms, which can expand the sizes of clustered species before and after clustering. Each clustered species adds edge sites that can become activated through H abstraction. This scheme concentrates clustering on a limited number of nucleation seeds and readily attaches species too volatile to condense via dispersive forces at high temperatures.

The reactants for this second stage of the mechanism are commonly observed under flame conditions. In addition, many experimental studies have demonstrated the formation of small (1 to 6 nm) incipient particles with carbon-to-hydrogen ratios of ~2 composed of randomly ordered aromatic structures with some aliphatic character (20), which are consistent with the clusters predicted by the CHRCR mechanism. This stage of the mechanism is thus supported by theoretical and experimental results. In addition, the CHRCR mechanism clusters both aromatic and aliphatic species and may explain the aliphatic and aromatic content observed spectroscopically in carbonaceous particles formed in interstellar space and in laboratory experiments that mimic the interstellar and circumstellar conditions where stardust forms (1, 2, 4, 5).

Soot particles that have grown and chemically evolved at high temperature are composed of aggregates of quasi-spherical primary particles with diameters of 10 to 50 nm. These primary particles often have disordered carbon cores surrounded by more ordered shells of graphite crystallites (21, 22). The particle core is 1 to 4 nm in diameter (21, 22), which is consistent with the size range of clusters likely formed by the CHRCR mechanism, indicating that such clusters could seed further particle growth.

The third stage of the mechanism involves growth beyond the initial cluster (Fig. 1C). We hypothesize that, once clusters have formed, RSR moieties at particle surfaces could provide sites for chemisorption of small hydrocarbons that are abundant but too volatile to condense on the surface via dispersive forces. Thus, CHRCR pathways could potentially explain recent results showing that the surfaces of soot particles may have less aromatic and more aliphatic character than the rest of the graphitic shell (12, 23).

The CHRCR mechanism overcomes the major problem of insufficient abundances of viable PAHs because it can account for clustering of a wide range of hydrocarbon sizes and structures. This mechanism thus relies on abundant small hydrocarbons observed under conditions where carbonaceous particles are formed. Because the CHRCR mechanism regenerates radicals and is likely to eject free H atoms as it proceeds, it concentrates clustering on certain species and minimizes radical quenching, which slows other mechanisms involving radical reactions.

The CHRCR mechanism is relevant to rich hydrocarbon combustion where soot is formed, the outflow of carbon-rich stars where carbonaceous dust is formed, and hydrocarbon-pyrolysis conditions; it is generally applicable for conditions under which hydrocarbon radicals can form and react by high-temperature or photolytic processes. Future studies are needed to investigate reaction rates and yields of these RSRs, rates and identities of nucleated particles, and rates and mechanisms of particle-surface growth under relevant conditions to verify these results.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S11

References (2436)

Data File S1

References and Notes

Acknowledgments: We thank B. Rude and D. Taube for assistance at the ALS and R. Bambha for help improving the paper’s clarity. Funding: This work was funded by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences (BES), Chemical Sciences, Geosciences, and Biosciences Division, Gas Phase Chemical Physics Program. M.P.H.-G. and K.R.W. are supported by this program under Contract no. DE-AC02-05CH11231. The ALS at LBNL is supported by the Director, DOE BES, under Contract no. DE-AC02-05CH11231. SNL is a multimission laboratory, managed and operated by National Technology and Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International, Inc., for the DOE’s National Nuclear Security Administration under Contract DE-NA0003525. The views expressed in this article do not necessarily represent those of the DOE or U.S. Government. Author contributions: K.O.J., H.A.M., M.P.H.-G., and K.R.W. wrote the paper. K.O.J. and H.A.M. conceived of the project and wrote the original draft of the paper. K.O.J., P.E.S., K.R.W., and H.A.M. performed the experimental work. K.O.J. analyzed the mass spectra and performed the calculations. M.P.H.-G. helped guide the calculations. K.R.W. built the aerosol mass spectrometer. H.A.M. managed the project. Competing interests: The authors declare no competing interests. Data and materials availability: The mass spectra shown in Fig. 2 and figs. S1 to S3 are provided in a separate file (Data File S1).
View Abstract

Navigate This Article