Towards a Comprehensive Characterization of the Low-Temperature Autoxidation of Di-n-Butyl Ether

In the present study, we investigated the oxidation of 2500 ppm of di-n-butyl ether under fuel-rich conditions (φ = 2) at low temperatures (460–780 K), a residence time of 1 s, and 10 atm. The experiments were carried out in a fused silica jet-stirred reactor. Oxidation products were identified and quantified in gas samples by gas chromatography and Fourier transform infrared spectrometry. Samples were also trapped through bubbling in cool acetonitrile for high-pressure liquid chromatography (HPLC) analyses. 2,4-dinitro-phenylhydrazine was used to derivatize carbonyl products and distinguish them from other isomers. HPLC coupled to high resolution mass spectrometry (Orbitrap Q-Exactive®) allowed for the detection of oxygenated species never observed before, i.e., low-temperature oxidation products (C8H12O4,6, C8H16O3,5,7, and C8H18O2,5) and species that are more specific products of atmospheric oxidation, i.e., C16H34O4, C11H24O3, C11H22O3, and C10H22O3. Flow injection analyses indicated the presence of high molecular weight oxygenated products (m/z > 550). These results highlight the strong similitude in terms of classes of oxidation products of combustion and atmospheric oxidation, and through autoxidation processes. A kinetic modeling of the present experiments indicated some discrepancies with the present data.


Introduction
The use of fossil fuels contributes to anthropogenic emissions of CO 2 and the formation of particles in the atmosphere, which are harmful to the ecosystem and cause health problems [1]. In addition, the increase in global energy needs and the depletion of fossil resources create socio-economic problems [2] which lead to the exploration of other energy sources such as nuclear, hydro, wind, solar, and renewable fuels from biomass which do not compete with food resources [3]. The chemical composition of a fuel induces specific physico-chemical characteristics, e.g., viscosity, density, boiling point, etc., as well as combustion and oxidation specificities. These characters influence the performance of internal combustion engines and the formation of pollutants and particles [4]. The development of future oxygenated lignocellulosic biofuels has put di-n-butyl ether (DBE) at the ranking of the most promising lignocellulosic biofuel and fuel additive [5]. Various experimental studies have highlighted the combustion characteristics of di-n-butyl ether and its effect on the performance of engines and injection systems [6][7][8], particle emissions [9], soot formation [10,11], and flame characteristics [12]. The combustion of di-n-butyl ether has been explored in different experimental systems and over various temperature and pressure ranges. Fan et al. [13] investigated its pyrolysis kinetics using a flow reactor experiment at low and atmospheric pressures. They identified and quantified several intermediates and products using photoionization molecular-beam mass spectrometry (PI-MBMS). Thion et al. [14] studied the oxidation of 1000 ppm of di-n-butyl ether at low to high temperatures (470 to 1250 K) in a fused silica jet-stirred reactor (JSR) at 1 and 10 atm. Oxidized species have been identified by gas chromatography-mass spectrometry (GC-MS) and Fourier transform infrared spectrometry (FTIR). GC analyses have also been performed using mass spectrometry, flame ionization, and thermal conductivity detectors. The authors observed two negative temperature coefficient regions (NTC). They also proposed a detailed chemical kinetic reaction mechanism to simulate their JSR experiments and model literature data. Tran et al. [15] studied the oxidation of di-n-butyl ether at 400-1100 K and the nearly atmospheric pressure using different experimental systems, i.e., a plug flow reactor (PFR) combined with electron ionization molecular-beam mass spectrometry (EI-MBMS) as well as two different jet-stirred reactors with either online gas chromatography or synchrotron vacuum ultraviolet photoionization molecular-beam mass spectrometry (SVUV-PI-MBMS). They confirmed two negative temperature coefficient zones around 500-550 K and 650-750 K, as reported earlier by Thion et al. [14]. About 40 species have been reported, among which 20 low-temperature products were not previously observed, such as hydroperoxides (CH 4 [15]. They also proposed a detailed chemical kinetic reaction mechanism to simulate their experiments [15].
More recently, an experimental study has been carried out in our research team [16]. A jet-stirred reactor has been used to study the cool flame oxidation of 5000 ppm of di-nbutyl ether (i.e., 520 K and 480 to 670 K at 1 and 10 atm, respectively). A rapid compression machine (RCM) has also been used to perform ignition experiments of DBE/an air mixture at 550-630 K and 5 bar. High resolution mass spectrometry (HRMS) analyses of JSR and RCM chemical samples have revealed the formation of hydroperoxides (C 8 Thus, a more comprehensive study is necessary to clarify the oxidation processes occurring during the cool flame oxidation of di-n-butyl ether and to assess the occurrence of pathways other than extended oxidation routes observed earlier [17][18][19]. Moreover, whereas several studies have shown a similarity between combustion processes and those involved in atmospheric chemistry, this has not been addressed for DBE. The aim of the present work is to better characterize the low-temperature oxidation chemistry of DBE and to probe the possible formation of products of importance in an atmospheric chemistry such as highly oxidized molecules (HOMs), products of RO • isomerization and oxidation, and accretion products. To this end, the oxidation of 2500 ppm of di-n-butyl ether was performed at low temperature (460-780 K), a residence time of 1 s, and high pressure (10 atm) in a fused silica jet-stirred reactor. Products of DBE oxidation through both 'combustion' and 'atmospheric' oxidation routes were tracked using gas chromatography, Fourier transform infrared spectrometry, and ultra-high pressure/ high-pressure liquid chromatography (UHPLC/HPLC) coupled to HRMS. An existing kinetic reaction mechanism [14] was used to model the present experiments.
The present study focuses on characterizing low-temperature intermediates of oxidation such as DBE-yl-hydroperoxides (C 8 Figure 1 presents the variation of the mole fraction of di-n-butyl ether as a function of temperature. As can be seen from that figure, two NTCs were observed under the present conditions. This is in line with what was reported earlier [14,15]. We observed maximal fuel conversion at~540 and~660 K. The kinetic model tends to represent the GC data fairly well, whereas those from Orbitrap analyses were scaled to the GC mole fraction at the lowest oxidation temperature common to the two types of experiments (480 K), i.e., GC, FTIR, and HPLC-MS. The normalized rate of consumption analysis was performed with PSR [20] at 540 K and 660 K. It indicated that DBE is essentially consumed by H-atom abstraction by hydroxyl radicals at both temperatures. However, the formation of hydroxyl radicals proceeds via different routes when temperature increases from 540 to 660 K. In the first cool flame, the production of hydroxyl radicals is predominantly due to the decomposition of C 8 • OOQOOH radicals deriving from the fuel (54.3%) and, to some extent, to C 8 KHPs' decomposition (29.1%). Minor contributions of the decomposition of C 3 H 7 OOH (1.6%) and C 4 • OOQOOH (3.7%) were predicted. At 660K, 25% of the hydroxyl radical production was due to the decomposition of C 1 -C 3 ROOH: C 3 H 7 OOH (12%), C 2 H 5 OOH (7.1%), and CH 3 OOH (5.8%). At 660 K, the decomposition of C 8 • OOQOOH radicals and C 8 KHPs represents only 13.1% and 11.7%, respectively. A small contribution to • OH formation also comes from the successive decomposition of C 4 and C 3 intermediates: C 4 • OOQOOH → • OH + KHP (3%) → • OH + carbonyls (2.6%) and C 3 KHPs → carbonyls + • OH (4.4%).
This behavior is responsible for the observed two-stage cool flame oxidation of DBE as initially reported by Thion et al. [14], notably similar to what was reported recently for the cool flame oxidation of di-n-propyl ether in a JSR at 700 Torr [ can produce carbonyl species (e.g., 2-butenal, acrolein, acetaldehyde, and formaldehyde). These products were identified here. Mole fraction profiles and simulations for acrolein, acetaldehyde, and formaldehyde are presented in the Supplementary Materials ( Figures S2 and S3). Additionally, 2-butenal was identified in di-n-butyl ether oxidation samples through the formation and detection of the DNPH derivative (crotonaldehyde + DNPH) standard. Unsaturated di-hydroperoxides (C 8 H 16 O 5 ) and unsaturated tri-hydroperoxides (C 8 H 16 O 7 ) can also be formed during the low-temperature oxidation of di-n-butyl ether. Different formation routes can be proposed, as illustrated in Scheme 3a-c.   As mentioned in the introduction, the combustion of fuels is based on the reaction sequences initiated by the abstraction of a hydrogen atom from the fuel. In the case of di-n-butyl ether, the oxidation proceeds via the initial formation of a • C 8 H 17 O radical: Proposed pathways leading to the formation of C 3 , C 4 , C 5 , and C 6 KHPs are presented in Scheme 4. These ketohydroperoxides (C 3 H 6 O 3 , C 4 H 8 O 3 , C 5 H 10 O 4 , and C 6 H 12 O 4 ) can be formed from different low-temperature oxidation intermediates, i.e., from the decomposition of C 8 -ketohydroperoxides (Scheme 4a-c) and C 8 -peroxy radicals' reactions (Scheme 4d). Their decomposition produces hydroxyl radicals contributing to the observed second cool flame (620-720 K), as shown in Figure 1. Due to the use of liquid chromatography (HPLC and UHPLC) coupled to high resolution mass spectrometry and positive or negative APCI modes, signals corresponding to C 3 Figure S4). Their signal profiles were plotted and compared to the modeling results ( Figure 3). One should note that only the formation of C 3 H 6 O 3 and C 8 H 16 O 4 was been considered in the model of Thion et al. [14]. For convenience, LC-HRMS data for C 3 and C 8 KHPs were scaled to their maximum simulated mole fractions. For the other KHPs, we present the variation of the signal of their positive ions as a function of temperature. Considering the low signals obtained for C 5 and C 6 KHPs, the data were multiplied by 100. As can be seen from Figure 3, the kinetic model predicts a maximum mole fraction of C 8 Figure S5, presenting a C 8 H 15 O 3 + UHPLC-HRMS chromatogram). Only butyric anhydride was identified. The unavailability of diketo-ethers standards made it difficult to identify them. The UHPLC-HRMS experimental profiles for butyric anhydride and the sum of the other diones are presented in Figure 4a.   Figure S5). 2,4-DNPH derivatization was applied to distinguish cyclic ethers and butoxy-carbonyl C 8 H 16 O 2 isomers (4-butoxybutanal, 4-butoxybutan-2-one, and 1-butoxybutan-2-one). Cyclic ethers do not react with 2,4-DNPH, whereas carbonyls do. Therefore, the intensity of their chromatographic peaks decreases. Computed and experimental (GC, UHPLC) mole fraction profiles of the sum of cyclic ethers and n-butyl butanoate, and the sum of butoxy-carbonyls isomers are presented in Figure 4b-d.
One should note that unlike the modeling and UHPLC-HRMS data, which reveal the formation of several cyclic ethers (ten species in the kinetic model, i.e., 4-ethyl-2-propyl-1,3-dioxolane, 4-methyl-2-propyl-1,3-dioxane, 2-butoxytetrahydrofuran, etc., and at least five in the UHPLC-HRMS analyses), GC analyses allowed for the detection of only two cyclic ethers. Their maxima of concentration are not reached at the same temperature (640 K in GC, 680 K in UHPLC, and 720 K in the simulation), which could be explained by the absence of the detection of some cyclic ethers in the GC analyses. The kinetic model underestimates the formation of cyclic ethers by about 50%.
Concerning butyl butyrate profiles (Figure 4c), we noticed a very good qualitative agreement between the GC and UHPLC results. In addition, it is noticeable that butyl butyrate reached maxima at 520 K and 640 K. This could be due to its formation via different intermediate species. Figure 4d shows that the butoxy-carbonyl profile (sum of 4-butoxybutanal, 4-butoxybutan-2-one, and 1-butoxybutan-2-one) peaks at 540 K.
One should note that KHPs are isomeric forms of hydroperoxyl-cyclic ethers. We have considered that the chemical formulae detected in this work for these isomers correspond mostly to KHPs. This assumption is based on results of simulations performed for n-hexane under conditions close to those of the present study [23] using a literature mechanism [18], including the formation of functionalized cyclic ethers (hydroperoxy and di-hydroperoxycyclic ethers). The model predicted 16 and 160 times less hydroperoxy and di-hydroperoxycyclic ethers, respectively, than KHPs. This result seems to validate our assumption. Scheme 5 gives a simplified view of the reaction pathways forming these products.

Oxygenated Products Formed via Atmospheric Oxidation Routes (RO 2 • and RO • )
Additional oxidation channels not taken into account in the combustion were considered in the atmospheric chemistry, apart from the usual oxidation routes (Reactions (4) and (5)).
Among them, one can find reactions which could also occur during combustion but are generally ignored (Reactions (7)-(12)): Besides, ROH can also be formed through H-atom abstraction by an alkoxy radical: Through Reactions (4) O 3,5,7 species obtained before and after DNPH derivatization revealed a decrease of the intensity of some chromatographic peaks, which indicates the presence of carbonyl compounds. The other peaks showing no decreased intensity should be attributed to species having no carbonyl function, e.g., unsaturated hydroperoxides. The chromatograms are given in the Supplementary Materials, Figure S7.
UHPLC-HRMS analyses with positive APCI ionization allowed for the detection of molecules corresponding to ROOR' species. Among them, one could detect C 16    Both Reactions (12) and (13) can form ROH molecules. In the case of di-n-butyl ether, butoxy-butanol species (C 8 H 18 O 2 ) were detected using UHPLC HRMS analyses (positive APCI, C 8 H 19 O 2 + , m/z 147.1378; Figure 8a). We can note that the C 8 H 18 O 2 signal reached a maximum at 540 K. Based on simulated RO • and ROO • mole fractions (Figure 8b) showing RO • radicals dominating, one can expect that the production of ROH is mainly due to H-atom abstraction reactions by RO • radicals (Reaction (13)). The formation of ROH and R' -H =O, involving ROO • radicals, should be a minor channel under the present conditions as it seems to be based on the observed minor formation of butoxy-carbonyls (Figure 8a).

Further Characterization of Oxygenated Products
To further characterize the products of the oxidation of DBE, additional analyses and data processing were performed. A DBE oxidation sample collected at 540 K was analyzed using the flow injection analysis technique (FIA). The analyses were performed using the negative APCI-HRMS parameters given in Table S1 of the Supplementary Materials. As can be seen from the mass spectrum shown in Figure 9, the mass range extended to m/z > 550. This observation is somewhat unexpected for the oxidation of DBE in a cool flame, whereas it would likely be less surprising if DBE had been oxidized under simulated atmospheric conditions where accretion products have been reported for a range of organics [24]. Unfortunately, no data are available for the atmospheric oxidation of DBE to be compared to the present results. However, in a previous study of ours concerning limonene oxidation in a JSR at atmospheric pressure and 590 K [25], we showed that products of combustion and atmospheric oxidation presented strong similitude in terms of the chemical formulae detected and the variation of the oxidation state of carbon atoms as a function of the number of carbon atoms in the detected formulae. The oxidation state of carbon atoms (OSc ≈ 2O/C-H/C) allows for evaluating the degree of oxidation of a large range of organic species (alcohols, carbonyls, carboxylic acids, esters, and ethers, but not peroxides) [26].
We computed the variation of OSc as a function of the number of carbon atoms, namely nC, for presently detected molecular formulae obtained by HRMS ( Figure 10). One can clearly distinguish several regions corresponding to low-volatility oxygenated organic aerosols (LV-OOA), semi-volatile oxygenated organic aerosols (SV-OOA), and water-soluble organic aerosols (WSOC), as defined by Kroll et al. [26]. Since di-n-butyl ether is composed of eight carbon atoms (C 8 H 18 O), for nC < 8, the observed products result from oxidative fragmentation. For 8 < nC < 16, we observed products of addition or condensation, and for nC > 16, we observed products of oligomerization. A comparison of Figure 10 with results published earlier for the oxidation of limonene in a JSR and with a range of atmospheric oxidation studies [25] indicates unexpected similar trends regarding the presence of LV-OOA, SV-OOA, and WSOC for the two fuels, namely DBE and limonene (see Supplementary Material, Figure S8).

JSR Experiments
The oxidation of DBE was performed in a jet-stirred reactor, which has 4 injectors (nozzles of 1 mm I.D.) providing stirring [27][28][29]. Di-n-butyl ether (>99% purity, Sigma-Aldrich ® , Saint-Louis, MO, USA) was atomized by a nitrogen (N 2 ) gas flow and vaporized in a heated chamber. Diluted oxygen and a fuel-nitrogen mixture were sent separately to the injectors to avoid oxidation before reaching the JSR. The liquid fuel was pumped using a Shimadzu ® LC10 AD-VP HPLC pump equipped with a Shimadzu ® DGU-20A3 online degasser. N 2 and O 2 flow rates were controlled by mass flow controllers (Brooks Instruments, Hatfield, PA, USA,). A Pt-Pt/Rh-10% (0.1 mm O.D., located inside a thin wall jacket) was used to verify good thermal homogeneity (gradients of <1 K/cm) along the vertical axis of the JSR. The experiments were performed twice. First, we collected gas samples for online FTIR and gas phase analysis (GC). Secondly, we collected samples by bubbling gas from the sampling sonic probe into cooled acetonitrile (0 • C). The liquid samples were stored in a freezer (−15 • C) and analyzed by both HPLC and UHPLC coupled to high resolution mass spectrometry (Orbitrap Q-Exactive ® , Thermo Fisher Scientific, Waltham, MA, USA). This analytical protocol has been used in previous works of ours where its effectiveness was demonstrated [16,23]. The JSR experimental conditions are shown in Table 1. Excellent stability was observed for the fuel, oxygen, and nitrogen flowrates, temperature, and pressure in the JSR during the experiments.

Analytical Procedures
Different complementary analytical methods were developed to analyze di-n-butyl ether oxidation products. Fourier transform infrared spectroscopy was used to detect formic and acetic acids, namely H 2 O, CO 2 , CO, and CH 2 O. Gas chromatography (GC) with a flame ionization detector (FID) to quantify hydrocarbons and oxygenates up to C 8 was employed. A GC with a thermal conductivity detector (TCD) was employed to quantify H 2 and O 2 , and a GC-quadrupole mass spectrometer (q-MS) was used for ≤C 8 products' identification.
Liquid chromatography analyses were mostly performed in the reversed-phase (RP-C 18 Luna Omega column, Phenomenex ® , Torrance, CA, USA). Additional high-performance liquid chromatography (HPLC)−HRMS analyses were performed with a Supelco ® Ascentis Silica column. To this end, a Vanquish™ system from Thermo Fisher Scientific ® , Waltham, MA, USA was used. Soft ionization using atmospheric pressure chemical ionization (APCI) in positive ([M+H] + ) and negative ([M-H] − ) modes was used to ionize products after chromatographic separation. Direct infusion was also used (see Supplementary Materials, Table S1). Then, the sample was injected into the ionization chamber of the mass spectrometer using a syringe pump. FTIR, GC, and HPLC/UHPLC analytical settings are detailed in the Supplementary Materials, Table S1. The HPLC-Orbitrap data are only qualitative because no calibration could be done in this work. The other data obtained by GC and FTIR are quantitative because the chemical species (reactants and products) were calibrated using standards. Uncertainties on the mole fractions are of the order of 10% in GC and 15% in FTIR. HOMs are minor products (in the ppm-range); the carbon balance could not be impacted by the fact they are not quantified in this work.

Kinetic Modeling
The present experiments were simulated using the PSR computer code [20]. The detailed kinetic reaction mechanism of Thion et al. [14] was used. It included 2768 reactions (most of them reversible) and 467 chemical species describing both low and high-temperature oxidation of DBE. This mechanism is limited to two additions of O 2 to fuel radicals ( • C 8 H 17 O), which yield ketohydroperoxides (C 8 H 16 O 4 ).

Conclusions
The low-temperature (460-780 K) and high-pressure (10 atm) oxidation of 2500 ppm of di-n-butyl ether was studied in a jet-stirred reactor. Due to the use of high-resolution mass spectrometry (Orbitrap Q-Exactive) coupled to liquid chromatography (HPLC and UHPLC) and atmospheric pressure chemical ionization (APCI), a wide range of previously observed and newly observed low-temperature oxidation products were detected. They include, besides light ubiquitous hydrocarbons' oxidation products, species resulting from the one to four O 2 additions on the fuel's radicals, i.e., hydroperoxides and di-hydroperoxides (C 8 O 3 , were also detected. The present results allowed for highlighting the finding that common reaction pathways pertaining to combustion and atmospheric autoxidation chemistry occur under the present JSR experimental conditions. Further studies involving complementary techniques, e.g., SVUV-PI-MBMS, would be useful for confirming the proposed identification of isomeric products such as cyclic ethers vs. carbonyls. FIA indicated the formation of high molecular weight oxygenated products (m/z > 550) under the present experimental conditions. Accretion products are of high importance for the formation of atmospheric particulates. Then, it would be interesting to investigate the formation of the newly detected species in this study using a smog chamber or flow reactors to further clarify similarities in terms of oxidation pathways in combustion and atmospheric oxidation conditions. Supplementary Materials: Figure S1: Evolution of O 2 , H 2 O, CO, CO 2 , and H 2 , and formic acid experimental and computed mole fraction profiles obtained for the low-temperature oxidation of 2500 ppm of di-n-butyl ether at 10 atm, ϕ = 2, and a residence time of 1 s; Figure S2: Comparison between experimental mole fractions of organic species and simulations for the oxidation of 2500 ppm of di-n-butyl ether in JSR (pressure of 10 atm, low-temperature 460-780 K, ϕ = 2, and a residence time of 1 s); Figure S3: Mole fraction profiles of aldehydes (formaldehyde, acetaldehyde, propionaldehyde, and butyraldehyde), ketones (acetone and 2-butanone), carboxylic acids (acetic acid and butyric acid), and 2,3-DHF formed during the low-temperature oxidation of 2500 ppm of di-n-butyl ether at 10 atm; Figure S4 Bottom: chromatogram showing the formation of cyclic ethers, butyl butyrate, and butoxy-carbonyls (C 8 H 16 O 2 ) during the oxidation of di-n-butyl ether; Figure S6: Chromatograms showing HOMs formed during the low-temperature oxidation of 2500 ppm of di-n-butyl ether at 10 atm; Figure S7: Chromatograms showing C 8 H 16 O 3,5,7 obtained before and after DNPH derivatization; Figure S8: Variation of the OSc as a function of the number of carbon atoms in molecular formulae detected by the negative FIA-HRMS of a limonene oxidation sample collected at 590 K and 1 atm; and Table S1: Conditions for GC, normal, and reverse-phase liquid chromatography and FTIR analyses. Funding: This research study was funded by the Labex CAPRYSSES (grant number ANR-11-LABX-0006-01) from the Labex VOLTAIRE (grant number ANR-10-LABX-100-01) and from the Région Centre-Val-de-Loire, EFRD, and CPER (projects PROMESTOCK and APROPOR-E).
Institutional Review Board Statement: Not applicable.