Dissociative Ionization and Thermal Decomposition of Cyclopentanone Johani.M.Pastoors,[A] Andras Bodi,[B] Patrickhemberger,[B] and Jordy Bouwman*[A, C]

DOI:10.1002/chem.201702376 Full Paper

& Reaction Mechanisms |Hot Paper| Dissociative Ionization and Thermal Decomposition of Cyclopentanone JohanI.M.Pastoors,[a] Andras Bodi,[b] PatrickHemberger,[b] and Jordy Bouwman*[a, c]

Abstract: Despitethe growing use of renewable and sus- out areversebarrieropen up to oxopropenyl and cyclopro- tainable biofuels in transportation,their combustion chemis- panonecations by ethyl-radical loss and asecond ethene- try is poorly understood, limiting our efforts to reduce harm- loss channel, respectively.Astatistical Rice–Ramsperger– ful emissions. Here we report on the (dissociative) ionization Kassel–Marcus model is employed to test the viability of this and the thermaldecomposition mechanism of cyclopenta- mechanism. Thepyrolysis of cyclopentanone is studied at none, studied using imaging photoelectron photoion coinci- temperatures ranging from about 800 to 1100 K. Closed- dence spectroscopy.The fragmentation of the ions is domi- shell pyrolysis products,namely 1,3-butadiene, ketene, pro-

nated by loss of CO, C2H4,and C2H5,leadingtodaughter pyne, allene, and ethene, are identified based on their pho- + ions at m/z 56 and 55. Exploring the C5H8OC potential toion mass-selected threshold photoelectron spectrum.Fur- energy surface reveals hydrogen tunneling to play an impor- thermore, reactiveradical species such as allyl, propargyl, tant role in low-energydecarbonylation and probably also in and methyl are found. Areaction mechanism is derived in- the ethene-loss processes, yielding 1-butene and methylke- corporating both stable and reactive species, which were tene cations, respectively.Athigher energies,pathways with- not predicted in prior computational studies.

Introduction tion that can be produced from furfural,[3,4] which in turn is produced industriallybyacidic hydrolysis of hemicellulose, a In general, biofuelsare renewable and often sustainable alter- major constituent of plant matter,[5] or,alternatively,through natives to fossil fuels, whichcan also mitigate harmful emis- pyrolysis of biomass.[6] Cyclopentanone also serves as aprecur- sions.[1] In particular, more and more research focusesonfuels sor for other, high-density,fuels,[7] which are suitable for use in derived from waste biomass.While fossil fuels are typically airplanes. Athorough understanding of the unimolecular dis- composed of mostlycarbon and hydrogen, biofuels also con- sociation of cyclopentanone is crucial to accurately model its tain oxygen in the form of hydroxyl (e.g.,ethanol), carboxyl, al- combustion. dehyde, or ketone functional groups.[2] Their varying composi- The pyrolysis of cyclic ketoneshas been studied in some tion leads to differing combustion chemistry,whichneeds to detail. Employing fractional product condensation and IR spec- be thoroughly understood. troscopic analysis, Johnson and Walters[8] found 2-cyclopenten- One of numerouspossible classes of biofuel molecules are 1-one,1-butene, ethene, carbon monoxide, and hydrogen as [9] ketones. Cyclopentanone is acyclic ketone of C5H8Ocomposi- cyclopentanone pyrolysis products. Delles et al. performed similar experiments and reported the formation of 4-pentenal [a] J. I. M. Pastoors, Dr.J.Bouwman in addition to products reported by Johnson and Walters.[8] Radboud University,Institute for Molecules and Materials Several low-molecular-weight species were observed in agas FELIX Laboratory,Toernooiveld 7c, 6525 ED Nijmegen (The Netherlands) E-mail:[email protected] chromatogram of the reactionmixture, but could not be as- [10] [b] Dr.A.Bodi,Dr. P. Hemberger signed. Porterfield et al. studied the pyrolysis of another cy- Laboratoryfor Synchrotron Radiation and Femtochemistry cloketone, cyclohexanone (C6H10O), using flash pyrolysis in Paul Scherrer Institute, 5232 Villigen (Switzerland) combination with photoionization mass spectrometry and [c] Dr.J.Bouwman matrixisolation Fouriertransform infrared (FTIR) spectroscopy. Presentaddress:Sackler Laboratoryfor Astrophysics They reported that methylvinyl ketone and ethene are formed Leiden Observatory,Leiden University,P.O. Box 9513 2300 RA Leiden (The Netherlands) through keto–enol tautomerization followed by aretro-Diels– Supporting information and the ORCID identification numbers for the Alder reaction. Furthermore, they argued that the other detect- authors of this article can be found under https://doi.org/10.1002/ ed products, among which ketene (C2H2O), are formed by chem.201702376. meansofa, b,org C Cbond cleavage followed by rearrange- À 2017 The Authors. Published by Wiley-VCH Verlag GmbH&Co. KGaA. mentsofthe thus formed diradicals. This is an open access article under the terms of Creative Commons Attri- Cyclopentanone pyrolysis has also been addressed computa- bution NonCommercialLicense, whichpermitsuse, distributionand repro- [11] duction in any medium, provided the original work is properlycited and is tionally.Zaras et al. characterized the energetics of the uni- not used for commercial purposes. molecular dissociation of cyclopentanone using the composite

Chem. Eur.J.2017, 23,13131 –13140 13131 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Full Paper

G3B3 method. They considered C Cscissionreactions, among Threshold photoelectron spectrum À other possible channels,and only found products formed from reactions initiatedbya-cleavage, despite the fact that b-cleav- The cyclopentanone cation has C2 symmetry and the highest age was determinedtobelower in energy.The computational occupied molecular orbital(HOMO), shown togetherwith the results on the neutralcyclopentanone potential energy surface mass-selected threshold photoelectronspectrum in Figure 1, (PES) were employed to calculate rate constants of the initial isomerization steps over the 800–2000 Ktemperature range by using Rice–Ramsperger–Kassel–Marcus(RRKM) modeling. They concludedthat the main products are C2H4 +CO and the rate constantssuggest that keto–enol isomerization plays apivotal role in the formation of other products, namely cyclopentenone, 1,3-butadiene, and acetylene. Wang et al.[12] performedacombined theoretical and experimental study on the dissociative photoionization of cyclopen- + tanone, providing adetailed C5H8OC PES. Experiments were performed using afemtosecond laser to induce dissociative ionization and time-of-flight(TOF) mass spectrometry for detection of fragments. Based on quantumchemical computations, rates and branching fractions were derived by RRKM modeling. Figure 1. Threshold photoelectron spectrum of cyclopentanone (connected Literature data to validate the theoretical computations on blue circles)displayed togetherwith aFranck–Condon simulation of the cyclopentanone dissociation are scarce. In prior experimental spectrum (green). The inset showsthe highest occupied molecular orbital on cyclopentanone. pyrolysis studies, product specieswere assigned offline after condensation, introducing the possibility of secondary reac-

tions and inherently imposing uncertainties on the actual un- has antibondingcharacter along the C=Obond and bonding derlyingformation mechanism of the detected products.More- character along the! two (O=)C Cbonds. Upon ionization À over,thesestudies were not sensitive to reactive intermediates, through the X˜ + 2B X˜ 1Atransition, electron densityisremoved which could provide important clues to the overall chemistry. from the HOMO,leadingtoa0.02 contraction of the C=O Furthermore, single-photon dissociative ionization data em- bond and a0.03 elongation of the two (O=)C Cbonds, as À ploying parent-ion internal energy selection are not available. predicted by B3LYP/6–311 ++G(d,p) geometry optimizationof However,such data can be used to quantify bond dissociation the neutraland ionic ground state. The measured spectrum re- energiesonthe cationic surface accurately,and provide details veals astrong resonance at (9.27 0.02) eV,whichisinagree- Æ on the underlying dissociation mechanism,[13] relevant for un- ment with the ionization potentialfor cyclopentanone of derstanding and modeling, for example, mass spectra.[14] (9.25 0.02) eV reportedbyChadwick et al.[16] The simulated Æ Here we report on an experimentalstudy of the (dissocia- spectrum matches the resonances in the measured spectrum tive) ionization and pyrolysis of cyclopentanone, using thresh- very well. Three major features at 9.30, 9.36 and 9.43 eV can be 1 old photoelectron photoion coincidence spectroscopy (TPEPI- assigned to aCH2 rocking motion (260 cmÀ ), aCC(=O) C [15] 1 À À CO). The pyrolysis products are sampled directly from the stretching vibration (776 cmÀ )and acombination of C=Oand 1 pyrolysis reactor and are identified by comparing their thresh- aC C(=O) Cstretch (1268 cmÀ ). À À old photoelectron spectra to data availablefrom the literature. Apyrolysis mechanism is proposed and discussed in light of Dissociative ionization the recent findings on cyclohexanonepyrolysis. The experimental findings on the dissociativeionization are rationalized The dissociative photoionization of cyclopentanone has been by hydrogen tunneling and amodel is constructed to test the studied to distinguish dissociative ionization products from viability of this fragmentation mechanism. thermaldecomposition products.[17] Furthermore, it is interesting to compare and contrastthe fragmentation processes in the neutralmolecule, induced by heat in the microreactor,and in the ion, broughtabout by the excess energyofanionizing Results vacuum ultraviolet (VUV) photon.Inthe breakdowndiagram in Figure 2A,the fractional abundance of parentand fragment The ionization, dissociative ionization, and pyrolysisofcyclo- ions detected in threshold photoionization of the room-tem- pentanone have been studied and the results are presented in peraturesample is plotted as afunction of photonenergy.The the next sections. Areaction mechanism is proposed for the lowest energychannel sets in at around 10.4 eV in the room- dissociative ionization and experimental resultsonpyrolysis temperature sample, and involves the loss of 28 amu, that is, are compared to acomputational studyfrom the literature. CO or C2H4.The CBS-QB3 calculated dissociative photoioniza- Supplementary pyrolysis reactionmechanismsare proposed to tion energies to form the most stable fragment ions, the 1- explain someofthe observed products. butene and methylketene cations, are 10.46 and 10.76 eV,re-

Chem. Eur.J.2017, 23,13131 –13140 www.chemeurj.org 13132 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Full Paper

m/z 55 is quite symmetric, suggesting absence of akinetic

shift. Therefore, the rate-determining step in the CO- or C2H4- loss channel appears to be atight transition state, and loss of

C2H5,associatedwith alooser transition state, quicklyoutcom- petes 28 amu loss once it is energetically allowed.

The potential energy surfaceofthe C5H8Ocation is explored considering these insights, and dissociation pathways are proposed in Figure 3. Dissociation is preceded by a-bond breaking

Figure 3. Potential energy diagram displaying transition states and minima in the dissociativephotoionization of cyclopentanone.Energy levels ob- tainedusing the CBS-QB3 methodare shownrelativetothe neutral and are italicized for transition states.Wavy arrowsindicatetunneling contributions Figure 2. A) The cyclopentanone breakdown diagram showing threshold in hydrogen-transfer isomerization steps. The dotted line in methylketene photoionization fractional parent and daughterion abundances as afunc- formation represents several omitted transition states with increasing tion of photon energy.Opensymbolsrepresentexperimentallymeasured energybut all below the energy of the products. values and continuous lines correspond to the RRKM model result. Loss of both CO and C2H4 may contribute to the 28 amu product loss at high photon energies, and the individual modelcontributions are shown (purple in the parent ion, yielding astraight-chain intermediate at and green) together with their sum (black). B) Illustrative threshold photo- 10.00 eV,whichplays acentral role in all three fragmentation ionization TOF distributionsbetween10.55 and 11.14 eV.Dots correspond to processes,that is, the loss of C2H5,C2H4,and CO. Ethene loss to experimental data, continuous lines representthe statistical model fits. form the cyclopropanone cation comprises the highest energy product channel, but can take place without an energy maxi- spectively.Aswill be discussed later,the formation of the cy- mum,along an attractive reactionenergy curve. 2,5-Hydrogen clopropanone cation is already possible at its thermochemical transfer results in aslightly more stable isomer overamoder- onset at 11.26 eV over aloose transition state, but its contribu- ate barrierat10.37 eV,which maybesubjecttotunneling, tion can be ruled out at low photonenergies. On theother after which an ethyl radicalmay be expelled. The moststable hand, loss of 29 amu corresponds unequivocally to the loss of dissociation products,the 1-butene cation and carbon monox-

C2H5,ofwhich the thermochemical threshold lies at 11.14 eV, ide, are only accessible after surmounting an energetic 3,5-hy- more than 0.5 eV lower than HCO loss (11.69 eV). drogen-transfer transition state at 11.18 eV.However,these hy- We note in passing that, upon ionization, the calculated OC- drogen-transfer processes may be enhanced by tunneling, sim- [18] CH2-CH2 bond angle increases from 57.68 to 106.38 in cyclopro- ilar to the less dominant methane-lossprocess in acetone, panone, that is, the cationic minimum is severely distorted. which could also explain the low and slowly rising rate con- The cyclic structure corresponds to atransition state, which stant associated with 28 amuloss. The first, a-bond breaking connects the two structures with the carbonyl group coordi- step in the parention mayalso be accompanied by simultane- nated to the each of the CH2 groups. For sake of simplicity,we ous 3,2-hydrogen-atom transfer over atransition state continuecalling this ion the cyclopropanone cation. (11.17 eV)that is virtually isoenergetic with the 3,5-hydrogen- The time-of-flight distributions, shown in Figure 2B,shed transfer transition state mentioned earlier.Although afurther, light on the competition between the 28 and 29 amu loss lower energy hydrogen-transferstep may yield the CO-loss channels. The m/z 56 peak exhibits aquasi-exponentialdecay precursor isomer at abarrier of 10.84 eV,this step is unlikely to to higher TOF,which is indicative of fragmentation in the ac- competewith the loss of ethene, yielding the methylketene celerationregion of the mass spectrometer,and—in the exper- cation already at 10.76 eV.This process involves CO-transfer imentalconfiguration used in this work—aunimolecularrate and C2H4-rotation transition states at progressively higher barri- 4 7 1 constantof10–10 sÀ .Incontrast, the 29 amu loss peak at ers (omitted in Figure 3, see dotted line), butnone is higher in

Chem. Eur.J.2017, 23,13131 –13140 www.chemeurj.org 13133 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Full Paper energy than the products. It is likely that the pure hydrogen- Pyrolysis transfer transition state leading to loss of CO is more enhanced by tunneling than the simultaneous openingofthe carbon The pyrolysis of cyclopentanone has been characterized by ring and vicinal hydrogen transfer,therefore the product photoionization mass spectrometry.Pyrolysis products are formed from loss of 28 amu is probably mostlythe 1-butene identifiedbymeans of mass-selected threshold photoelectron cation, but methylketenecontributionsbyloss of ethene spectroscopy (ms-TPES) and chemical pathways leading to the cannotberuled out at energies exceeding 10.76 eV. detected products are identified. In addition to the 28 and 29 amu loss fragment ions, the center-of-gravity analysis[19] of the parention peak revealed trace amountsofhydrogen-loss fragments at around10.8 eV Mass spectrometry photon energy.This is much lower than the direct C Hbond À Atypical all-electron time-of-flightmass spectrum of the prod- breakingdissociativephotoionization energy (11.42 eV), which ucts and intermediates formed from the pyrolysis of cyclopen- implies isomerizationtothe enolicparent ion, which subse- tanone at about 1100 Kand ionized at 10.5 eV is shown in quently may lose ahydrogen atom already at 10.32 eV,making Figure 4. Cyclopentanone is heavily diluted (ca. 0.1%)inargon this the lowest energy dissociative photoionization process dis- and species with m/z 85 have not been observed, so contri- cussed herein. The reasonfor its almost negligible abundance butionsbybimolecular chemistry are deemed unlikely.The lies in the enolization transition state, which, at 11.54 eV,isalso the highest energy structure that appears to playarole in the studied energy range. Similartothe low-energy CO and C2H4 losses, enol formation followed by hydrogen loss maytake place predominantly by tunneling, and the small hydrogen- loss abundance confirms negligible enol formation in dissociative photoionization. The dissociative photoionization mechanism shown in Figure 3isput to the test by constructing astatisticalRRKM model based on the computed energetics, assuming that the rate-determining step is the formation of the products for the

C2H5-and C2H4-loss channels, the latter yieldingthe cyclopro- panonecation, and passing over or throughthe hydrogen- transfer transition state for the CO-losschannel. Since the CO- loss and the low-energy C2H4-loss channel are expected to behaverather similarly with the formerbeing the more dominant one, we have not explicitly considered the latter in the Figure 4. Time-of-flight mass spectrum of cyclopentanone pyrolysis products formed at about1100 Kand ionized at 10.5 eV constructed from coinciden- model.The hydrogen-atom tunneling rate constantsinCOloss ces of all photoelectrons regardless of their position on the detector (i.e., [20] were calculatedusing an Eckart barrier, and only the transi- energy) withphotoions. tion state transitional mode frequencies were fitted in the high-energyC2H4 and C2H5 loss, together with the critical tunneling frequency in CO-loss.Asshown in Figures 2A and B, the peaks at 55 and 56 amu are attributed to dissociative photo- model reproduces the observed ion abundances and dissocia- ionization of thermally excited cyclopentanone as described in tion rates quite well, thereby validating the dissociative photo- the previoussection, while all other products are formed by ionization mechanism. Onlythe fitted criticalfrequency of H pyrolysis. tunneling is significantly higher than the calculated one, at 1 1 3500 cmÀ vs. 510 cmÀ ,which suggestsastrongly anharmonic Product identifications barrierand possibly asignificant contribution of the low- energy C2H4-loss process. The mass spectra also establish the Photoion mass-selected threshold photoelectronspectra were 55 and 56 amu ion signals as the sole dominant low-energy recorded to identify products and intermediates formed upon dissociative photoionization products. While the 56 amu signal pyrolysis of cyclopentanone. The ms-TPESofthe main prod- is also apotential thermaldecomposition product of cyclopen- ucts at m/z 84, 54, 42, 41, 40, 39, 28 and 15 are shown in tanone (vide infra), mass-selected threshold photoelectron Figure 5. The remaining product signals at m/z 27, 52, 80 and spectroscopy (ms-TPES) is used to rule out significant 1-butene 82 that are apparent in Figure 4are not intenseenough for neutralformation upon pyrolysis. Therefore, dissociative pho- constructing an ms-TPES with sufficient signal-to-noise, and toionization products do not interfere with the neutral frag- these products are discussed later in light of the overall chemi- ments generated through pyrolysis. cal mechanism. The most intense peak in Figure 4corresponds to the precursor molecule as confirmed by the spectra in Figures 1and 5. The main discrepancies between the two spectra are caused by the difference in spectral resolution,compounded by the

Chem. Eur.J.2017, 23,13131 –13140 www.chemeurj.org 13134 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Full Paper

The product detected at m/z 42 can have contributions by

isomersofC3H6 and C2H2Ocomposition. The ms-TPES of this product is compared with photoelectron spectroscopic data of species that may contribute (Supporting Information).[26–28] The photoelectron spectrum of ketenetaken from reference [28] fits our ms-TPES at m/z 42 best (see Figure5). Asmall contribution of propene to the overall m/z 42 ms-TPEScannot be ruled out. An additional resonance is observed at around10eV, which could be avibronic band of ketene,but there is alarge intensity mismatch. Alternatively,this band could be the result

of avibronic transitionofthe C2H2Otautomer ethynol. In absence of an experimental photoelectron spectrum, the ionization potential is computed at the CBS-QB3 level of theory to be 9.75 eV.Notransition is observed around9.75 eV,rendering contributions from ethynol very unlikely. The ms-TPES at m/z 41 (Figure 5) exhibits astrong resonance at around 8.2 eV and is in good agreementwith the spectrum of the allyl radical.[29] Cyclopropyl contributions[30] can be ruled out. Allene, propyne, and cyclopropene can be responsible for the fragment detected at m/z 40. While cyclopropene can be excluded, due to the absence of aresonance at 9.66 eV,[31] con- tributionsfrom both allene and propyne areclearly observed, indicating either two differentformationchannels or arear- rangementbetweenthe two isomers. The resonantly stabilized propargyl radicalisthe only speciesresponsible for the signal observed at m/z 39. The photoelectronspectrum of propargyl is also displayed in Figure5.[32,33] Indeed, its photoelectron spectrum resembles that of the product detectedhere very well. Due to the strong transitionat10.5 eV,the m/z 28 is assigned to ethylene.[34] Carbon monoxide could contributeto this mass channel as well, but its ionization threshold is 14.01 eV,[35] beyondthe photon energy range of this study.The bottom trace in Figure 5(m/z 15) is attributed to the methyl radical. The ionization thresholds of the weak signals observed in the mass spectrum (Figure 4) at m/z 52, and m/z 27 correspond to vinylalcetylene[36] and vinyl radicals, respectively.

Figure 5. Normalized threshold photoelectron spectra(TPES) of fragments m/z 84, 54, 42, 41, 40, 39, 28, and 15 recordedat920 Kover aphoton energyrange of 8to11eVinsteps of 0.05 eV.Plotted in green are the (threshold) photoelectron spectraofthe assigned species taken from the literature. broadening effect of the higher temperature of the neutrals in the pyrolysis experiment. There is no clear sign of cyclopente- nol that may be formed from keto–enol tautomerization and has an ionization threshold at 8.4 eV,[21] nor is there aclear spectroscopic signatureofalinear C5H8Oisomerthat may form upon ring opening. The pyrolysis product detectedatm/z 54 can have acompo- sition of either C4H6 or C3H2O. The spectrum of 1,3-butadiene[22]—displayed in green on top of the m/z 54 TPES in Figure 5—matches the measurement very well, leading us to concludethat 1,3-butadiene is the only species contributing to this mass channel. Significant contributions by otherspecies are ruled out based on the absence of spectroscopic signa- [22–25] tures (see Supporting Information for more details). Figure 6. Normalized integratedion signals as afunction of pyrolysis tube temperature for the parent and prominent closed-shell pyrolysis products.

Chem. Eur.J.2017, 23,13131 –13140 www.chemeurj.org 13135 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Full Paper

Temperature dependence c-C H O 2C H CO 1 5 8 ! 2 4 þ ð Þ

Figure 6shows peak areas as taken from mass spectraat The barrier involved in this reactionislocated at 10.5 eV photon energy as afunctionofpyrolysis tube tempera- 1 [11] 325.1 kJmolÀ . Alternatively, a-cleavage of cyclopentanone ture. The products at m/z 82 and 28 appear at relatively low followed by an isomerization was calculated to result in the temperature, whilethe other channels open up when the tem- same products,albeit at alower rate. Etheneisconfirmedto peratureisfurtherincreased. Note that the signal strengths be one of the main pyrolysis products and its formation sets in may not reflect the absolute abundances, as the intensities are at the lowest temperatures (see Figure 6). Carbon monoxide not scaled by absolute ionization cross sections. However,the does not ionize in the studied photon energy range, but it is apparent trend does hold clues to the underlying chemical fair to assumethat CO is formed asco-product. mechanism. The product path to yield 1,3-butadiene and CO was com- The ms-TPESdata presented in Figure5 are found to be puted to have the lowest overall rate-limiting transition state largely insensitivetothe pyrolysis temperature, with the ex- and proceeds through dehydrogenation to yield 3-cyclopent- ception of the spectrumtaken at m/z 40. This peak is dominat- en-1-one, which subsequently decarbonylates [Eqs. (2) and ed by allene at low temperatures, while atransition/rearrange- (3)].[11] ment to the most stable isomer propyne is observed at higher temperatures (see Supporting Information, Figure S3). c-C H O c-C H O H 2 5 8 ! 5 6 þ 2 ð Þ c-C H O 1,3-C H CO 3 5 6 ! 4 6 þ ð Þ

Decomposition mechanism of cyclopentanone The dehydrogenation to yield 3-cyclopenten-1-one may pro- ceed directly from cyclopentanone through loss of H ,but the Cyclopentanone pyrolysis products are compared to those pre- 2 multi-step process involving keto–enoltautomerization to 1- dicted in apreviouscomputational study to pin down the un- cyclopenten-1-ol shown in Figure7 has alower rate-limiting derlyingdissociation mechanism. Asummary of the relevant barrier. The tautomerization comprises the rate limiting step of reactionmechanismsleadingtodetected species and showing 1 303.8 kJmolÀ .Next, molecular hydrogen is lost to form 1,4-cy- only the rate-limiting transition states as computed by Zaras clopentadien-1-ol. Asubsequentrearrangement via 2-cyclo- et al.[11] is presented in Figure 7. penten-1-one produces 3-cyclopenten-1-one, from which CO The fastestchemical reactionaccording to RRKM calculations can be eliminated to form 1,3-butadiene. Indeed asignal is ob- is asingle-step decomposition leading to CO accompanied by served at m/z 82 and could point to the stabilization of afrac- two equivalents of ethene[Eq. (1)].[11] tion of the formed c-C5H6Ospecies. This signal grows in togetherwith ethene (See Figure6), but gets depleted quickly as the 1,3-butadiene signalgains in intensity,confirming this reaction pathway.Further dehydrogenation of 2-cyclopenten-1-one may yield 2,4-cyclopentadien-1-one. In this study,ketene, allene, and propyne as wellasthe radicals allyl, propargyl,vinyl and methyl have also been identified. However, their formationhas neither been predicted computa- tionally previously,nor were these speciesdetected experimen- tally before.[9,11] Ketene and allene may be formed from diradicals that form through the a-, b-, or g-cleavage of cyclopentanone as shown in Figure 8, analogous to the unimolecular decomposition of cyclohexanone.[10] Both a-and b-cleavage may

yield ketene. The accompanying CCH2CH2CH2C diradical further dissociates by means of hydrogen-atom loss to allyl and even- tually allene.The detection of allene as aproduct indicates that ketene, allene, and allyl can be direct decomposition products of cyclopentanone. The mechanism in Figure 8sug- gests that a and g scission may also yield propadienal. Howev- er,noclear signature is found in the ms-TPES, suggesting that alternative decompositionprocessesproceed at ahigher rate. Further unimolecular dissociation and isomerization of products formed from the initial pyrolysis of cyclopentanone also Figure 7. Main low-energycyclopentanone decomposition pathways report- need to be considered as possible contributors to signals ob- ed by Zaras et al.[11] and confirmed in this work. Arrow labelsindicatethe energyofthe rate-limiting transition state and minimum energies of the pic- served in the mass spectrum that have not been assignedso 1 + tured structures are also showninkJmolÀ relativetocyclopentanone. far.First, C3H3 can form throughdissociativeionization of allyl

Chem. Eur.J.2017, 23,13131–13140 www.chemeurj.org 13136 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Full Paper

tial of acetylene, see Supporting Information) reveals asignal at m/z 26, confirming that subsequent dissociation of cyclopentenone indeed contributes to the detected products, as no alternative pathway can explain the formation of acetylene. Pyrolysis of 1,3-butadiene as formed from dissociation of cyclopentanone could also contribute to the detected products. Its pyrolysis has been studied in the past[37] and was found to proceedmostly by meansofthe reaction shown in Equa- tion (7).

C H CH C C H C 7 4 6 ! 3 þ 3 3 ð Þ In the photoionization mass spectra reported by Chambreau et al.,[37] products are also seen in the mass channels corre-

spondingtovinylacetylene (C4H4)and vinyl (C2H3). Thus, thermal decomposition of 1,3-butadiene could also explain the formation of these products, although the temperaturesreported for the decomposition of butadiene exceed 1000 Kand are thus rather high compared to the temperatures used herein. The formation of vinyl and methylradicals can alternatively proceedbydiradical driven disproportionation reactions, which are summarized in the Supporting Information. Atomic hydrogen is abundantlyformedaccording to the proposed cyclopentanone dissociation mechanism and,although the density of the parent molecule in the reactor is kept low (ca. 0.1%), secondary reactions initiated by atomichydrogen may have asmall contribution to the formationofthe observed products,orisomerization thereof. The m/z 40 peak is the only one to exhibit aclearly temperature-dependent isomer composition. Allene is the dominant isomer at low temperatures and is convertedtopropyne at the highest temperature. Theformation of allene is expected from the mechanisms proposed in Figure 8. Isomerization of allene to propyne has been studied extensively in the past and propyne is the moststable isomer.[38,39] Alternatively,propyne Figure 8. Decomposition pathways of cyclopentanone and cyclopentenone formation can already be initiated in the thermally excited through a-, b-, and g-cleavage. Computedbond dissociation energies (BDE) C H Obythe increased hydrogen mobility. These two mecha- are taken from Zaras et al.[11] 5 8 nisms cannotbediscerned in the current experiments.

Discussion and Conclusions [Eq. (4)] or,alternatively,through thermal dissociation of allyl followed by ionization [Eqs. (5) and (6)].[17,29] The dissociative ionization and pyrolysisofcyclopentanone have been studied on the iPEPICO (imaging photoion photo- hn C H C C H þ H eÀ 4 electron coincidence) instrument at the VUV beamline of the 3 5 ! 3 3 þ 2 þ ð Þ Swiss Light Source. The dissociative ionization between 10.3 DT C3H5C C3H3C H2 5 and 11.7 eV yields product ions at m/z 56 and 55 that are ! þ ð Þ formed through loss of CO, C2H4 and C2H5C.The potential hn C3H3C C3H3þ eÀ 6 energy surfacehas been characterized and it is found that the ! þ ð Þ formation of the m/z 56 product proceeds by means of hydro- Another possible channel to allyl proceeds via cyclopenta- gen tunneling and resultsinthe 1-butene cation and CO and none. The formation of cyclopentenone+H2 was computed to also the methylketene cation plus ethylene for energies ex- be one of the lowest energy pathways for dissociation of cy- ceeding 10.76 eV.The high-energyethene loss yielding the cyclopentanone (Figure7)and its subsequent dissociation clopropanone cation starts to contribute to this mass channel through a-, b-, and g-cleavage is summarized in Figure 8. The at energies exceeding 11.26 eV.The m/z 55 product is formed detection of propargyl could point to cyclopentenone also through loss of C2H5 with abarrierat11.14 eV.AnRRKM model being an intermediate speciesinketene formation. ATOF mass is employed to support the validity of the proposed hydrogen- spectrum recorded at 11.5 eV (i.e.,above the ionization poten- tunneling mechanism.

Chem. Eur.J.2017, 23,13131–13140 www.chemeurj.org 13137 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Full Paper

The potentialenergy surfaceofthe dissociative ionization of Keteneand allene hadnot been detected in previousexperi-

C5H8Opresented here differs slightly from that reportedprevi- mental studies, nor had they been predicted from computa- ously.[12] First, the lowest energy pathway leading to the tions. Zaras et al.[11] did not identify pathways through b-cleav- [12] C3H4O+C2H4 reported by Wang et al. yields cyclopropanone age, although it was established that this is the weakest C C 1 À at an energy of 170.7 kJmolÀ (1.77 eV) relative to the cation. bond in the molecule. The unambiguous identification of In this work, we find that this channel is significantly higher keteneand allene from this study clearly points to the impor- (by 0.2 eV)inenergy at 1.97 eV with respecttothe cation tance of b-cleavage in the pyrolysis of cyclopentanone and a (11.26 eV w.r.t. to the neutral). In addition, the calculations pre- more detailedcomputational study needs to be undertakento sented here provide an additional pathway to an alternative characterize the energetics and kinetics of this reaction path-

C3H4Oproduct, methylketene.This channelrequires0.5 eV less way.Itisinteresting to note that the decomposition in our py- energy than the cyclopropanone pathway with arate-limiting rolytic microreactor andupon dissociative ionization share sim- barrierlocated at 11.17 eV and all other barriers below the ilar reactionproducts, such as CO, C2H4 and C2H5C.Thus, under- thermochemical threshold. Moreover,this rate-limiting barrier standing the dissociative photoionization mechanism is impor- is subject to tunneling,making this aviable contributor to dis- tant, because the onset of dissociative ionization can be shift- sociationfor energies exceeding10.76 eV.Lastly,anoticeable ed towards lower photon energies at elevated temperatures, difference lies in the CO-losschannel. In addition to the con- which may be wrongly assigned to decomposition upon pyro- certed ring opening and hydrogenshift at abarrier of 11.17 eV lyis.[40,41] In addition, similar fragmentation connects both the reportedhere and by Wang et al.,[12] we also located apracti- neutral molecules’ and ions’ potential energy surface, which cally isoenergetic transition state (at 11.18 eV)that involves can be used to derive heats of formation of reactive intermedi- ring openingfollowedbymigration of ahydrogen (i.e.,anon- ates in the absence of areverse barrier.[42] It is also interesting concerted channel), making it somewhat more likely to domi- to note an important differencebetween dissociative photo- nate the CO-loss mechanism.The detection of avery weak ionization process and pyrolysis mechanisms. Keto–enol tauto- signal at m/z 83 indicates that keto–enoltautomerization only merization appears to play acrucial role in the formationof plays aminor role in dissociative photoionization. butadiene on the neutralPES, whiletautomerization is out- Cyclopentanone pyrolysis products formed atatemperature competed by other product channels on the ionic surface. ranging from 800 to 1100 Khave been identified by means of However,even on the neutralsurface, tautomerization involves ms-TPESofthe speciessampled directly from the pyrolysis ahigher barrier than subsequenthydrogen abstraction, which tube reactor.The main identified product speciesare ethene, means that the enol only acts as ashort-lived intermediate 1,3-butadiene, ketene, allene,propyne, allyl, propargyl, and and dissociates promptly. methyl.Vinylacetylene, acetylene, and vinyl radicals have also The findings reportedhere can also be compared to experi- been observed, but only as trace species. Allene and propyne mental work reported on cyclohexanone. Porterfield et al.[10] are found to be the only isobaric products that are sensitive to found that keto–enol tautomerization followed by aretro the reactortemperature, with allene being the dominantC3H4 Diels–Alder fragmentation is responsible for the formationof isomer at lowand propyne at high temperature. CH =C(OH) CH=CH + C H .Ring cleavage reactions andsub- 2 À 2 2 4 Previousexperimentalstudies on cyclopentanone pyrolysis sequent decomposition reactions had been proposed to yield identified 2-cyclopenten-1-one, 1-butene, ethene, carbon mon- the other observed products,among whichspeciesalso ob- oxide, 4-pentenal, and hydrogen as products. Of these species, served in the pyrolysis of cyclopentanone reported here, such only C2H4 is positivelyidentifiedand CO and H2 are likely pres- as 1,3-butadiene, ketene andallene, allyl. In contrastwith their ent, but evade detection at the photonenergies used. The work, enols have not been detected in our study.Furthermore, non-detection of 1-butene in our experiments can be caused analogoustothe retro Diels–Alderreaction proposed by Por- by the fact that it dissociates further at the temperatures ap- terfield, aretro [2+2] cycloadditionreactionyielding ketene plied in our study,orasmall fraction of 1-butenecould be and allene could occur for cyclopentanone after reorganization hidden in the large signal at m/z 56 caused by dissociative ion- to afour-membered ring;however,this processes is energeti- ization.However,close inspection of this signal does not reveal cally unfavorable in cyclopentanone and it is more likely that indications of asignificant fraction of 1-butene. The non-detec- keteneand allene are formed upon a-orb-ring cleavage reaction of 1-buteneinthis study could be due to 1-butene’s tions instead. prompt dissociation by C Cfission [Eq. (8)]. Our resultsexemplify the power of isomer-resolvedproduct À detection in flash pyrolysis using ms-TPESrecorded by iPEPICO. C H CH C C H C 8 Reactive product species, such as keteneand anumber of radi- 4 8 ! 3 þ 3 5 ð Þ cals, have been sampled directly from the reactor and have An alternative and more likely explanation is that in the pre- been identified unambiguously.The multiplexed nature of ms- vious pyrolysis experiments 1-butene is formed from bimolecu- TPES offers aclearadvantage over the use of FTIR matrix isola- lar reactions in the pyrolysis tube, ratherthan through direct tion spectroscopy to assign products;spectralinformation is pyrolysis of cyclopentanone. The fact that 1-butene formation obtained mass selectively and isomer specific identification of has been computed to have arate-limiting barrier of the products are readily made.[22] The detection of the first 1 462.8 kJmolÀ andthat an RRKM modelpredicts low formation generation of species from the reactor reveals the true under- rate[11] supports this explanation.

Chem. Eur.J.2017, 23,13131 –13140 www.chemeurj.org 13138 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Full Paper lying chemical mechanism and supplies crucial information hence mass, were used. Internal energy selection was also achiev- that is needed for modeling the combustion of biofuels. ed by threshold photoionization, because the photon energy minus the adiabatic ionization energy was quantitatively converted to the internal energy of the parent ion. This excess energy and the original thermal energy of the sample was available for frag- Experimental Section mentation. When studying dissociative photoionization processes, the threshold ionization fractional parent and daughter ion abun- The experiments were performed on the iPEPICO instrument at the dances were plotted in the breakdown diagram, which often con- vacuum ultraviolet (VUV) beamline of the Swiss Light Source. A tains clues as to the fragmentation mechanism,[46] and can be mod- short description of the experimental setup is presented here and eled quantitatively using statistical thermodynamics approaches to amore extensive description of the apparatus can be found else- determine accurate dissociative photoionization thresholds.[47] Fur- where.[43] Synchrotron VUV radiation produced by abending thermore, owing to the long acceleration region and low extrac- magnet was collimated, dispersed in grazing incidence by agra- tion field, it took several microseconds for the photoions to reach ting, and focused at the exit slit in adifferentially pumped noble their terminal velocity in the mass spectrometer.Ifadissociation gas filter,which serves to remove higher order radiation. The ion- process occurs during this time, the fragment ion will have alower ization and detection chamber was flanged onto the exit port of than nominal kinetic energy,and ahigher time of flight than the gas filter,and the photon beam size was about 42mminthe promptly formed ions of the same mass. This results in TOFpeak interaction region. Dissociative photoionization measurements 1 profiles that are broadened towards higher TOF,and were modeled were recorded using a600 groovesmmÀ grating. Agrating with 4 1 1 to extract unimolecular dissociation rate constants in the 10 sÀ < 150 groovesmmÀ was used for recording mass-selected threshold 7 1 k(E)<10 sÀ range. The experimental rate curves were also mod- photoelectron spectra (ms-TPES) of the pyrolysis products. This eled to take kinetic shifts into account. grating was selected to enhance the VUV photon flux in the interaction region and, thus, signal levels by an order of magnitude at areduced photon energy resolution. The photon energy was tuned from 8.0 to 11.7 eV in steps of 0.05 eV for pyrolysis measure- Computational methods ments and in steps of 0.025 eV for the dissociative photoionization Quantum chemical calculations using the Gaussian 09[48] suite of experiments. + programs were performed to explore the C5H8OC PES to yield in- Cyclopentanone (Sigma–Aldrich, >99 %) was seeded in argon carri- sights into the chemical pathways leading to the products formed er gas at aconcentration of 0.1%and abacking pressure of from dissociative ionization. Reaction coordinates were scanned at 1.1 bar.The mixture entered aresistively heated microtubular reac- the B3LYP/6-311++G(d,p) level of theory to locate transition tor of about 4cminlength and 1mminner diameter through a states and intermediates. Subsequent re-optimizations at the CBS- 100 mmorifice. The temperature on the outer surface of the reactor QB3 level of theory were performed to compute accurate energies. is monitored by means of a“type C” thermocouple. Based on the These energies and the B3LYP computed vibrational frequencies [44] extensive simulations by Guan et al., the temperature on the were used as input for astatistical model to fit both the break- centerline of the reactor may be 25%lower than that measured down diagram and the time-of-flight profiles using the PEPICO pro- on the outside. The pressure at the entrance of the microreactor gram.[47] was only afew mbar and dropped in the heated zone towards the The B3LYP/6-311++G(d,p) optimized geometries and vibrational exit of the tube. The sample expanded into the source chamber, 4 normal modes of the electronic ground state of the neutral and which was maintained at about 110À mbar.The residence time cation were used as input for Franck–Condon simulations of the was estimated to be of the order of 10 to 100 ms. photoionization spectrum. The program ezSpectrum[49] was em- The gas mixture containing the pyrolysis products exited the reac- ployed to compute the spectral intensities of the vibronic transi- tor tube and was skimmed by a1mm diameter skimmer,allowing tions and the resulting stick diagram was convoluted with aGaus- 1 only the central part of the expansion to reach the ionization sian profile with afull-width-at-half-maximum (FWHM) of 100 cmÀ volume in the iPEPICO detection chamber,inwhich the molecular to account for the rotational envelope and to facilitate comparison beam intersects the monochromatic ionizing radiation from the with the experimental data. 1 VUV beamline. Aconstant 120 VcmÀ electric field accelerated the photoelectrons towards aRoentDek delay line detector,where they were velocity map imaged. The positions of the electrons on the detector contained information on their kinetic energy and Acknowledgements their arrival time served as the start signals for the TOF measurements of the cations. Threshold photoelectron spectra were plot- J.B. acknowledges the Netherlands Organisation for Scientific ted by selecting only the central spot on the electron detector and Research(NWO) for aVENI grant (Grant number 722.013.014). subtracting the “hot” electron background.[45] Cations were acceler- This work was furthersupported by NWO Exacte Wetenschap- ated in the opposite direction in aWiley-McLaren type TOF config- uration and detected with amicrochannel plate (MCP) detector. pen (Physical Sciences) for the use of the supercomputer facili- The mass resolution achieved in the experimental system was ties at SurfSara. A.B. and P.H. acknowledge funding by the m/Dm=125. Swiss Federal Office for Energy(BFE Contractnumber SI/ Coincidences between electrons and cations were plotted as a 501269-01). function of delay time to yield the TOFmass spectra. Those obtained by considering coincidences between all electrons and ions, regardless of the electron energies, are referred to as all-electron mass spectra. Isomer selective product identification is based on Conflict of interest ms-TPES, for which only coincidences between threshold electrons (kinetic energy <10 meV) and ions of aspecific time of flight, and The authors declare no conflict of interest.

Chem. Eur.J.2017, 23,13131 –13140 www.chemeurj.org 13139 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Full Paper

[30] J. Dyke, A. Ellis, N. Jonathan, A. Morris, J. Chem.Soc. Faraday Trans. 2 Keywords: flash pyrolysis · high-temperature chemistry · 1985, 81,1573 –1586. photoelectron spectroscopy · reactionmechanisms · [31] K. Vasilatou,J.M.Michaud, D. Baykusheva, G. Grassi, F. Merkt, J. Chem. renewable resources Phys. 2014, 141,064317. [32] H. Gao, Z. Lu, L. Yang, J. G. Zhou, C. Y. Ng, J. Chem. Phys. 2012, 137, 161101. [1] C. K. Westbrook, Annu.Rev.Phys. Chem. 2013, 64,201 –219. [33] P. Hemberger,M.Lang,B.Noller,I.Fischer,C.Alcaraz,B.K.C.deMiran- [2] S. K. Hoekman,A.Broch, C.Robbins, E. Ceniceros, M. Natarajan, Renewa- da, G. A. Garcia, H. Soldi-Lose, J. Phys. Chem. A 2011, 115,2225 –2230. ble Sustainable Energy Rev. 2012, 16,143 –169. [34] R. Stockbauer,M.G.Inghram, J. Electron Spectrosc. Relat. Phenom. 1975, [3] M. Hronec, K. Fulajtrova, T. Sotk, Appl. Catal. B 2014, 154–155,294 – 7,492 –496. 300. [35] P. Erman,A.Karawajczyk, E. Rachlewkallne, C. Stromholm, J. Larsson,A. [4] M. Winfough, K. Voronova, G. Muller, G. Laguisma, B. Sztray,A.Bodi, G. Persson,R.Zerne, Chem. Phys. Lett. 1993, 215,173–178. Meloni, J. Phys. Chem.A2017, 121,3401 –3410. [36] T. A. Cool, J. Wang, K. Nakajima, C.A.Taatjes, A. McIlroy, Int. J. Mass [5] P. McKendry, Bioresour. Technol. 2002, 83,37–46. Spectrom. 2005, 247,18–27. [6] A. Demirbas, Fuel Process. Technol. 2007, 88,591–597. [37] S. D. Chambreau, J. Lemieux, L. Wang,J.Zhang, J. Phys. Chem. A 2005, [7] J. Yang, N. Li, G. Li, W. Wang,A.Wang, X. Wang, Y. Cong, T. Zhang, 109,2190–2196. Chem. Commun. 2014, 50,2572 –2574. [38] A. Lifshitz, M. Frenklach, A. Burcat, J. Phys. Chem. 1976, 80,2437 –2443. [8] E. R. Johnson,W.D.Walters, J. Am. Chem. Soc. 1954, 76,6266 –6271. [39] M. Yoshimine, J. Pacansky,N.Honjou, J. Am. Chem. Soc. 1989, 111, [9] F. M. Delles, L. T. Dodd, L. F. Lowden, F. J. Romano, L. G. Daignaul, J. Am. 4198 –4209. Chem. Soc. 1969, 91,7645 –7647. [40] S. Liang, P. Hemberger,N.M.Neisius, A. Bodi, H. Grützmacher,J.Leva- [10] J. P. Porterfield, N. ThanhLam, J. H. Baraban, G. T. Buckingham,T.P.Troy, lois-Grützmacher, S. Gaan, Chem. Eur.J.2015, 21,1073 –1080. O. Kostko, M. Ahmed, J. F. Stanton, J. W. Daily,G.B.Ellison, J. Phys. [41] S. Liang, P. Hemberger,J.Levalois-Grützmacher,H.Grützmacher,S. Chem. A 2015, 119,12635 –12647. Gaan, Chem.Eur.J.2017, 23,5595–5601. [11] A. M. Zaras, S. Thion, P. Dagaut, Int. J. Chem. Kinet. 2015, 47,439 –446. [42] F. Holzmeier,I.Wagner,I.Fischer,A.Bodi, P. Hemberger, J. Phys. Chem. A [12] Q. Wang,D.Wu, M. Jin, F. Liu, F. Hu, X. Cheng, H. Liu, Z. Hu, D. Ding, H. 2016, 120,4702 –4710. Mineo,Y.A.Dyakov, A. M. Mebel, S. D. Chao, S. H. Lin, J. Chem. Phys. [43] A. Bodi, M. Johnson,T.Gerber,Z.Gengeliczki, B. Sztaray,T.Baer, Rev.Sci. 2008, 129,204302. Instrum. 2009, 80,034101. [13] O. Kostko, B. Bandyopadhyay,M.Ahmed, Annu.Rev.Phys. Chem. 2016, [44] Q. Guan, K. N. Urness, T. K. Ormond, D. E. David, G. B. Ellison, J. W. Daily, 67,19–40. Int. Rev.Phys. Chem. 2014, 33,447–487. [14] C. A. Bauer, S. Grimme, J. Phys. Chem. A 2016, 120,3755 –3766. [45] B. Sztµray,T.Baer, Rev.Sci. Instrum. 2003, 74,3763–3768. [15] T. Baer,R.P.Tuckett, Phys. Chem. Chem. Phys. 2017, 19,9698 –9723. [46] P. Hemberger,A.Bodi, T. Gerber,M.Wuertemberger,U.Radius, Chem. [16] L. Weiler,D.Chadwick, D. C. Frost, J. Am. Chem. Soc. 1971, 93,4320 – Eur.J.2013, 19,7090 –7099. 4321. [47] B. Sztµray,A.Bodi, T. Baer, J. Mass Spectrom. 2010, 45,1233 –1245. [17] J. Bouwman, A. Bodi, J. Oomens, P. Hemberger, Phys. Chem. Chem.Phys. [48] Gaussian 09, Revision D.01, M. J. Frisch,G.W.Trucks, H. B. Schlegel, G. E. 2015, 17,20508 –20514. Scuseria,M.A.Robb, J. R. Cheeseman, G. Scalmani, V. Barone, B. Men- [18] A. Bodi, T. Baer,N.K.Wells, D. Fakhoury,D.Klecyngier, J. P. Kercher, Phys. nucci, G. A. Petersson, H. Nakatsuji, M. Caricato, X. Li, H. P. Hratchian, Chem. Chem. Phys. 2015, 17,28505 –28509. A. F. Izmaylov,J.Bloino,G.Zheng, J. L. Sonnenberg, M. Hada, M. Ehara, [19] B. West, C. Joblin, V. Blanchet, A. Bodi, B. Sztaray,P.M.Mayer, J. Phys. K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Chem. A 2012, 116,10999 –11007. Kitao, H. Nakai, T. Vreven,J.A.Montgomery,Jr. ,J.E.Peralta, F. Ogliaro, [20] A. Bodi, B. Sztaray,T.Baer, Phys. Chem. Chem. Phys. 2006, 8,613–623. M. Bearpark, J. J. Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov, T. Keith, [21] G. Meloni, T. M. Selby, D. L. Osborn, C. A. Taatjes, J. Phys. Chem. A 2008, R. Kobayashi, J. Normand, K. Raghavachari, A. Rendell, J. C. Burant, S. S. 112,13444 –13451. Iyengar, J. Tomasi,M.Cossi, N. Rega, J. M. Millam, M. Klene, J. E. Knox, [22] A. Bodi, P. Hemberger,D.L.Osborn,B.Sztaray, J. Phys. Chem. Lett. 2013, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Strat- 4,2948 –2952. mann, O. Yazyev,A.J.Austin,R.Cammi, C. Pomelli, J. W. Ochterski, R. L. [23] K. B. Wiberg, G. B. Ellison, J. J. Wendoloski, C. R. Brundle,N.A.Kuebler, J. Martin,K.Morokuma, V. G. Zakrzewski, G. A. Voth, P. Salvador,J.J.Dan- Am. Chem. Soc. 1976, 98,7179–7182. nenberg, S. Dapprich, A. D. Daniels, O. Farkas, J. B. Foresman, J. V. Ortiz, [24] J. D. Savee, S. Borkar,O.Welz, B. Sztaray, C. A. Taatjes, D. L. Osborn, J. J. Cioslowski, D. J. Fox, Gaussian, Inc.,Wallingford CT, 2013. Phys. Chem.A2015, 119,7388–7403. [49] V. Mozhayskiy,A.I.Krylov,ezSpectrum: http://iopenshell.usc.edu/down- [25] Wr.Harshbarger,N.A.Kuebler,M.B.Robin, J. Chem.Phys. 1974, 60, loads. 345–350. [26] A. Katrib, J. W. Rabalais, J. Phys. Chem. 1973, 77,2358 –2363. [27] P. Rademacher, Chem. Rev. 2003, 103,933 –975. [28] C. Baker,D.W.Turner, J. Chem. Soc. D 1969,480 –481. Manuscript received:May 24, 2017 [29] I. Fischer,T.Schler,H.-J. Deyerl, M. Elhanine, C.Alcaraz, Int. J. Mass Spec- Accepted manuscript online:July 10, 2017 trom. 2007, 261,227 –233. Version of record online:August 31, 2017

Chem. Eur.J.2017, 23,13131 –13140 www.chemeurj.org 13140 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim