Experimental and modeling study of benzaldehyde oxidation

Abstract Benzaldehyde is an aromatic aldehyde commonly considered in bio-oil surrogate formulation, and an important intermediate in the oxidation of other aromatic reference fuels such as toluene. However, its oxidation has never been previously investigated experimentally and no product formation profiles were reported in the very limited pyrolysis studies available in the literature. In this study, the gas-phase oxidation of benzaldehyde was investigated in a jet-stirred reactor. 48 species were detected using gas chromatography, mainly CO, CO2 and phenol. The important formation of CO and phenol indicates a rapid formation of phenyl radicals. This was confirmed by a kinetic analysis performed using the current version of the CRECK kinetic model, in which the reactions of phenyl radicals and oxygenated aromatic compounds have been updated.


Introduction
Benzaldehyde is the lightest aromatic aldehyde and is one of the major intermediates in the combustion and atmospheric oxidation of benzyl radical [1]. For this reason, its formation and consumption pathways are accounted for in detailed kinetic models for the combustion of toluene and heavier aromatic compounds (e.g. [2][3][4]). Furthermore, many oxygenated aromatic compounds are formed during the decomposition of biomass [5,6] and aromatic aldehydes (e.g. furfural and derivatives or hydroxybenzaldehydes (see Figure 1a) and methoxybenzaldehydes) are a non-negligible part of these products [5,6]. a) b) c) Figure 1. Structures of a) 2-hydroxybenzaldehyde, b) benzaldehyde and c) acetaldehyde. Bond dissociation energies are given at 298 K and in kcal mol−1 (references: a) et b) [7] and c) [8]).
Oxygenated aromatic compounds (phenol, anisole, catechol, guaiacol, vanillin) are gaining academic and industrial interest due to their presence in bio-oils produced from biomass fast pyrolysis [2,7,9]. Therefore, the influence of different functional groups on their combustion properties should be systematically addressed. Moreover, the impact of oxygenated aromatics on PAHs and soot growth should be better analyzed, as the high aromatic content of bio-oils could lead to undesired increased formation of particulate matter [10]. The purpose of this study is to better assess the combustion kinetics of benzaldehyde (see Figure 1b), and also to propose a first overview towards an improved understanding of the oxidation of oxygenated aromatics with multiple substitutions of interest as bio-oil surrogates (e.g. vanillin) [7].
Benzaldehyde pyrolysis has already been studied in batch reactors by Hurd and Bennet [11] and by Solly and Benson [12] covering a temperature range from 573 K to 963 K at atmospheric pressure. Pyrolysis experiments have also been carried out in flow reactors by Grela et al. [13] at high temperatures (1005-1270 K) and low-pressure (P = 0.01 kPa), by Bruinsma et al. [14] at atmospheric pressure and over a temperature range from 773 K to 1173 K, and by Vasiliou et al. [8] between 1200 K and 1800 K and pressures ranging from 0.1 to 0.2 kPa. These studies showed that benzaldehyde decomposes to give phenyl radical plus H atom and CO: C6H5CHO → C6H5CO + H → C6H5 + CO + H, contrary to the usual aliphatic aldehydes where the alkyl acyl bond is the weakest (R-CHO) (see Figure 1c). However none of these studies provided mole fraction profiles for the fuel or the decomposition products for benzaldehyde pyrolysis. Also to our knowledge no oxidation study has been reported in the literature at present.
The aim of this study is to propose a first detailed set of data for the oxidation of benzaldehyde in a jet-stirred reactor (JSR) coupled with gas chromatographs (GCs). Using different type of detectors coupled to gas chromatography, mole fractions profiles were obtained for benzaldehyde and 48 of its products. These profiles were then used to update the benzaldehyde oxidation subset in the current version (1902) of the CRECK kinetic model.

Experimental methods
The JSR used for the experiments has often been used for kinetic studies of pyrolysis [15] and combustion [16]. It consists of a fused silica sphere (volume 92 cm 3 ) equipped with four injection nozzles positioned in a cross located at the center of the sphere. This injection method ensures high turbulence in the reactor and leads to homogeneity in gas phase concentration. A quartz annular preheating zone, in which the temperature of the gas is progressively increased up to the reactor temperature, precedes the reactor inlet and is responsible for the homogeneity in gas phase temperature.
The gas residence time inside the annular preheater is very short compared to its residence time inside the reactor (a few percent), so as to limit undesired reactivity in this section. The heating is ensured by resistances (Thermocoax) rolled around the reactor and around the preheating zone, which allows flexibility and swiftness in the heating of each area. Temperatures are measured by K-type thermocouples located in a glass finger inside the inlet cross (for the reaction temperature measurement) and between the resistances and the external wall of the reactor (for the temperature control of each zone). The reaction temperature is assumed to be equal to that measured in the inlet cross according to the isothermal reactor hypothesis, with a maximum gradient of ±5 K.
Benzaldehyde was provided by Sigma Aldrich with a purity of 99%. He (99.999% pure) and O2 (99.999%) were provided by Messer. Gas flow rates are controlled by mass flow controllers and the liquid flow rate by a Coriolis flow controller followed by a vaporization chamber maintained 10 K above benzaldehyde boiling temperature. The relative uncertainty in the flow measurements is around 0.5% for each controller, thus about 2% on the residence time.
The outlet gas leaving the reactor is then transported by a short heated line to GCs. The first GC, equipped with a Carbosphere packed column, a thermal conductivity detector (TCD) and a flame ionization detector (FID), is used for the quantification of light-weight compounds like methane, carbon monoxide and carbon dioxide. The second GC, fitted with a Q-Bond capillary column and a FID preceded by a methanizer, is used for the quantification of compounds containing from 2 carbon atoms, like acetylene or ethylene, to compounds with 5 carbon atoms (e.g. cyclopentadiene). The methanizer (nickel catalyst for hydrogenation) allows to detect species like CO and CH3CHO with a better sensitivity. A third GC equipped with a HP-5 capillary column and an FID is used for the detection of the heaviest compounds. Products like benzene, benzaldehyde, phenol, naphthalene or biphenyl are quantified with this apparatus. The identification of reaction products is performed using a GC equipped with both types of capillary columns and coupled to a mass spectrometer (quadrupole) with ionization at 70 eV. Calibrations are performed by injecting standards when it is possible with a maximum relative error in mole fractions around ±5% and ±10% for the other species calibrated using the effective carbon number method. Concerning fuel calibration, as benzaldehyde properties are at the limit of the experimental set-up abilities, the error in the fuel measurements can be estimated to ±10%, despite the direct calibration. Experimental mole fraction data in a spreadsheet are reported in the Supplementary Material attached to this study.
Benzaldehyde oxidation was studied between 700 K and 1100 K at 107 kPa, at three equivalence ratios: φ = 0.5-1-2. The residence time in the reactor was fixed at 2 s and the inlet mole fraction of fuel was 0.5%. Helium was used as the inert gas. For each experiment, the carbon mass balance has been calculated and is estimated to be ≈90% of the global inlet of carbon atoms.

Kinetic model
The CRECK model adopted in this work (466 species and 16,263 reactions) implements a C0-C3 core subset obtained by coupling the H2/O2 and C1/C2 from Metcalfe et al. [17], C3 from Burke et al. [18], and heavier fuels from Ranzi et al. [19,20]. It covers reactions from syngas to heavy diesel fuel pyrolysis and combustion, including PAH formation [2] as well as a new subset of reactions for oxygenated aromatics of interest as bio-oil surrogate components [7]. The thermochemical properties for species not specifically belonging to the benzaldehyde subset were adopted, when available, from the ATcT database of Ruscic [21] or from Burcat's database [22]. For the species involved in the benzaldehyde subset and not already implemented in the CRECK mechanism, the properties have been determined with the software THERGAS [23] based on Benson group additivity method [24]. Table 1 shows the primary pyrolysis and oxidation reactions of benzaldehyde, highlighting modifications and new reaction pathways implemented in this study (reactions whose parameters were estimated or taken from literature). It is composed of 18 reactions classed into different reaction types. The kinetic parameters given are: A, the pre-exponential factor in (cm −3 •mol −1 ) n-1 /s with n, the order of the reaction and Ea, the activation energy in cal/mol. Benzaldehyde can react through unimolecular initiations (reactions 1-2 in Table 1). The hypothesis has been made that only the aldehydic function is reactive and that the aromatic ring cannot react by its own. Due to the lack of specific kinetic information, kinetic parameters of initiation reactions have been determined based on bond dissociation energies (BDEs) as reported in Pelucchi et al. [7]. The enthalpies of formation of the radical involved in the BDE estimation are presented in SM.
Reactions 3-12 correspond to bimolecular initiations and H-abstraction reactions. Only the bimolecular reaction with oxygen has been taken into account. The kinetic parameters of the H-abstractions from the aldehydic position have been determined based on previous studies on linear aldehydes [25].
Ipso-addition reactions on the fuel are modeled by the reactions 13 to 17 and correspond to the additions of the main radicals onto the ring at the aldehydic position. The kinetic parameters of ipso-addition reactions have been determined based on the rate rules reported in [7]. Reactions with the aromatic hydrogen atoms were considered as negligible compared to those involving the aldehydic one.
Reaction 18 is the reaction of alpha-scission of the radical C6H5CO formed after the abstraction of the aldehydic hydrogen atom. The rate constant of reaction 18 has been adopted from Solly and Benson [12], as already proposed in the toluene model by Bounaceur et al. [26]. It should be noted that the value from Solly and Benson is the only high pressure limit determination available in the literature and most probably carries large uncertainties (i.e. one order of magnitude) mostly due to the very limited temperature range considered in the analysis (614-647 K). However, from a broader kinetic model perspective, the C6H5CO radical is very rapidly decomposed to phenyl radical and CO, without significant further interactions with the mixture. The addition of O2 to C6H5CO has also been investigated, based on acetaldehyde similar reactions. However, it was shown that these reactions were negligible and a decision was made to remove them from the mechanism.
The kinetic values of reactions 5 and 7 were already present in the mechanism, but their estimation needed a reevaluation due to deviations between the model computed results and the experimental data. Thus, their kinetic constants were decreased (within the estimated uncertainty) in order to better reproduce the fuel reactivity. The pre-exponential factors of reactions 5 and 7 have so been multiplied by factors of 0.5 and 0.3, respectively, with respect to the reference values for H-abstractions from the aldehydic moiety of linear aldehydes [25]. The adopted rate constant modifications are well within the uncertainty range of those parameters (a factor of 3). This uncertainty is established based on other parameters adopted in literature models [28]. In fact, Mendes et al. [29] declared an uncertainty of 2.5-3 based on the uncertainties of the single point energies. We considered a factor of 3 as an upper limit as uncertainties related to the treatment of anharmonicities and to the frequencies may also be present.
The OpenSMOKE++ package [30] has been used to perform all the kinetic simulations and analyses here presented.  It can be observed that the equivalence ratio has only a minor impact on the conversion of the fuel. The kinetic model reproduces the onset of reactivity well within the experimental uncertainty slightly overpredicting the fuel conversion between 850 and 950 K. It should be noted that the model exhibits a higher dependence on the equivalence ratio compared to the experimental measurements.     The 48 products have been identified during this study; they can be divided in three main categories (species in italic are only present under the form of traces, thus their profiles are not reported in Figures 3 and 4):
Due to our analytical set-up, the quantification of hydrogen and water is not possible. It can also be noticed that neither formaldehyde nor catechol were detected during the experiments. Co-elution problem for catechol and detection limits for formaldehyde could explain their absence amongst the detected products. Those products could have been expected due to the presence of methane and phenol in the gas phase with oxygen. Figure 5 presents the product selectivity at 850 K for the three studied equivalence ratios. Carbon monoxide and carbon dioxide are the main products at this temperature. Phenol, acrolein, acetaldehyde and ethylene are the other main products formed. Cresol is a specific product of the fuel rich conditions. The overall agreement between the experimental and modeling data presented in Figures 2, 3 and 4 is reasonable especially for equivalence ratios of 0.5 and 1. CO measured and estimates a production of catechol which was not observed experimentally. The profile of benzoquinone (C6H4O2) is also overpredicted by the model.
Steady state numerical solutions were not found for stoichiometric conditions above 1050 K. The model predicts an oscillating behavior as frequently observed for smaller species [31]. However, a stable behavior was observed in the experiments. Figure 6 shows a rate flux analysis of benzaldehyde oxidation at 850 K for stoichiometric conditions for a conversion of 50%. This conversion rate has been chosen in order to investigate the primary product decomposition pathways. For clarity, some species like CO and CO2 are not shown. Under these conditions, benzaldehyde mainly reacts by the abstractions of the aldehydic hydrogen atom as suggested by Vasiliou et al. [8] for pyrolysis (89% at 850 K). This H-abstractions on the fuel leads to the formation of the C6H5CO radical, which quickly decomposes by alphascission in phenyl radical and carbon monoxide. The phenyl radical then leads to the phenoxy radical (C6H5O in Figure 6). This radical is the precursor of almost all the products experimentally observed (e.g. acrolein, acetaldehyde, cresol…).
Benzaldehyde also reacts by ipso-addition reactions of hydrogen atom and C2H3 radical producing benzene and styrene. Styrene can then react through different H additions on different sites. An addition on the substituted unsaturation can lead to the formation of the radical C6H5CHCH3 or to the formation of a radical C6H5 and ethylene. The radical C6H5CHCH3 can also decompose into C6H5 radical and ethylene or can form the fuel back with the intervention of the radical HO2. Styrene can also react by ipso-addition reactions to form benzene.
Benzene is formed by ipso-addition of H atom on benzaldehyde, styrene and phenol. Benzene is almost exclusively transformed into C6H5O via two pathways: 1) O + C6H6 = H + C6H5O (97%) 2) R + C6H6 = RH+C6H5; C6H5+O2 = C6H5O+O (1.4%) Thus the formation and consumption of benzene is totally ruled by ipso-addition reactions on the other aromatic compounds present in the gas phase. As discussed in [27], the resonance stabilized phenoxy radical largely contributes to phenol formation by means of H-abstraction (C6H5O+RH=C6H5OH+R) on C6H5CHO and catechol and recombination with H atoms. Indeed 49% of the consumption flux of the radical C6H5 is dedicated to the formation of phenol. Phenol is then consumed in a significant fraction by H-abstraction reactions by O, OH, and HO2 to give C6H5O radicals (18%). Therefore, a cycle phenol/phenoxy is exhibited by the mechanism and could be important in the overall reactivity of the system. This observation is also supported by the sensitivity analysis on benzaldehyde performed at 770 K under stoichiometric conditions and presented in Figure 7. This temperature was chosen to investigate the reactions important at the start of the reactivity. Charts for other equivalence ratios are reported in the Supplementary Material attached to this study. Phenoxy radical (C6H5O) and the phenol/phenoxy cycle are controlling the reactivity at low temperatures. HO2 radical has also an important impact as it is involved in the phenol/phenoxy cycle via the reaction: HO2+C6H5OO2+C6H5OH. This analysis confirms that the C6H5O radical is the key intermediate in the reactivity of benzaldehyde at low-temperature. Thus its consumption, through the reactions HO2+C6H5O=>O2+C6H5OH and O + C6H5O=>2 CO+C4H5, inhibits the reactivity. Furthermore, these reactions are consuming radicals by forming stable species (CO, O2…) or by reducing the number of radical points in the system via termination reactions. Figure  8 presents a scheme of this analysis. The most important reactions, promoting the reactivity, are the H-abstractions on the fuel by HO2 and C6H5O to form the radical C6H5CO. This radical subsequently decomposes into C6H5 and C6H5O, also contributing to the phenol/phenoxy cycle. HO2 also reacts with the fuel to give H2O2, giving two OH radicals and therefore accelerating the reactivity through a multiplication of the number of radicals. The chemistry of other intermediates like acrolein (C2H3CHO) and catechol phenoxy-like radical (RCATEC) have also a promoting effect on the reactivity, but less dominant. Concerning the other equivalence ratios, the same analysis has been performed and the results are very similar. This observation seems to confirm the fact that the equivalence ratio has a minor impact on the overall reactivity.
Acetaldehyde and acrolein are produced by the decomposition of C6H5O to C4H5 and CO through the reaction: O + C6H5O=>2 CO+C4H5. Indeed the C4H5 radical reacts with oxygen to form HCO and acrolein: O2+C4H5=>HCO+C2H3CHO. Acrolein leads mainly to the formation of C2H3 radical (12.5%) via H-abstraction by OH, O, and H to give C2H3 and CO according to the lumped step R + C2H3CHO=>RH+CO+C2H3. This lumped step corresponds to the abstraction of the aldehydic H-atom followed by a decarbonylation of the radical (R-C*=O) as discussed in [25]. C2H3 then decomposes into formaldehyde in a large amount (30%) via the addition of dioxygen. Acetaldehyde is formed by the addition-elimination reaction of OH radical on acrolein: OH+C2H3CHO=>HCO+CH3CHO accounting for 3% of the consumption flux of acrolein. However the model is not able to predict the correct amount of acetaldehyde. This could also explain the overprediction of styrene mole fraction, due to an overproduction of C2H3 radicals from acrolein. Furthermore the model considers the formation of formaldehyde through the reaction: C2H3+O2=H+CO+CH2O. The non-detection of formaldehyde during the experiment may indicate that the transformation of acrolein into C2H3 and acetaldehyde needs a better assessment.
Concerning the simulated profiles of catechol, its formation is due to the ipso-addition reaction of an oxygen atom on phenol. This leads to the formation of the resonance stabilized phenoxy-like catechol radical, which can then react with other species in the system abstracting an H atom and forming catechol as shown in Figure 9a). a) b) Figure 9. (a) Flux rate between phenol and catechol at 850 K and (b) simulation data for catechol obtained with the kinetic model.
The importance of such pathways in oxygenated aromatics has been recently pointed out by Pelucchi et al. [7]. The model predicts a fraction of catechol between 60 ppm under fuel-rich conditions and 25 ppm under fuel-lean conditions. A chart in Figure 9b) presents the simulations results for catechol.
Despite this consideration the relative small amount of catechol predicted could lead to a more limited detection capability. Further analysis with gas chromatography and ad hoc analytical methods could help us to detect catechol in the outlet mixture.

Conclusion
In this work, benzaldehyde oxidation was investigated for the first time. Using a JSR coupled with gas chromatographs, 48 compounds were identified for various equivalence ratios covering from fuel lean to rich conditions. 28 mol fraction profiles were obtained for species from major ones (CO, CO2, phenol) to some minor products (acrolein, acetaldehyde, benzene, cresol, etc.), other species being present in the form of traces. These new data were compared with simulation results obtained with an updated version of the CRECK kinetic model. The model update allowed for a better description of benzaldehyde oxidation. The decomposition of benzaldehyde mainly leads to the formation of phenyl radical which then oxidize to give phenoxy radical (C6H5O), ruling the formation of almost all the observed products. Another relevant reaction class is that of ipsoaddition reactions leading to the formation of benzene and phenol, directly from benzaldehyde. Those compounds also react to give the C6H5O radical. The updated CRECK kinetic model provides a valuable starting point for further refinement towards a more systematic characterization of oxygenated functional groups effect on substituted aromatics kinetics of interest for bio-oils combustion applications.
This work confirms the importance of the C6H5O radical in the decomposition of oxygenated aromatic compounds and provides a first detailed assessment of benzaldehyde oxidation kinetics upon which more accurate kinetic subsets should be developed.