Shock Tube and Flame Speed Measurements of 2,4,4-Trimethyl-1-Pentene: A Co-Optima Biofuel

Real fuels such as gasoline, diesel, and aviation fuels can consist of hundreds of individual components whose compositions change based on the origin of the harvested crude oil, the refinery process, and even the season in which the oil was refined. For gasoline, the primary chemical classes are n-alkanes, iso-alkanes, aromatics, naphthenes, and olefins, which can be represented by n-heptane, isooctane, toluene, cyclohexane, and diisobutylene, respectively [1]. Because of concerns about climate change and CO₂ emissions, many researchers, including the present authors, have recently helped the development of alternate energy sources, fuels, and power cycles [2–14]. Diisobutylene (DIB) is also a candidate fuel under the Department of Energy’s Co-Optimization of Fuels and Engines (Co-Optima) initiative 2,3, aiming to optimize new fuels and engine designs simultaneously [11,12]. One of the objectives of this initiative is to present advanced biofuels for high-efficiency spark and compression ignition engines. There have been recent studies of the combustion behavior of several of these Co-Optima fuels using a variety of laboratory techniques [15–18]. DIB is composed of three parts 2,4,4-trimethyl-1-pentene (DIB1) and 1 part 2,4,4-trimethyl-2-pentene (DIB2) and is preferred to represent the olefin category because its branched structure is more representative of the branched olefins typically found in gasoline, and it has a similar carbon length to isooctane. However, since DIB is primarily DIB1, this is often times the only species considered in reduced surrogate mechanisms to represent the olefin class [19,20].

The literature on neat DIB kinetics is rather scarce though several studies exist for lower alkenes (e.g., ethylene, propene, butene) [21–28]. Metcalfe et al. [29] examined the ignition delay times (IDTs) of DIB1, DIB2, and DIB between 1200 and 1550 K, at 1 and 4 atm, and equivalence ratios from 0.25 to 1.0. It was found that DIB2 is more reactive than DIB1 under the conditions investigated, and DIB responds in a linear fashion to the ratio that the constituent isomers are combined. A chemical kinetic model was developed and was successful at reproducing ignition delay times for the conditions studied with slight underprediction of times at higher temperatures. The authors also found that DIB1 decomposition is highly dependent on isobutene. Mittal and Sung [30] used a rapid compression machine to investigate the ignition characteristics of binary blends of DIB1, toluene, and isooctane. It was found that within the range of 740–1060 K, 15–45 bar, and at an equivalence ratio of 0.75, DIB1 has a higher reactivity among the fuels at temperatures greater than 820 K and lies between isooctane and toluene for lower temperatures. The addition of DIB1 or isooctane to toluene resulted in a much greater reactivity than just neat toluene at similar conditions, with the effect of DIB1 addition being much greater than that of isooctane. Hu et al. [31] measured ignition delay times of DIB1 and extended the previously measured range of Metcalfe et al. [29] to higher pressures and equivalence ratios. Ignition delay times were measured from 2 to 10 atm and at equivalence ratios of 0.5–2.0. Realizing a higher reactivity from the mechanism presented by Metcalfe et al. to the experimental IDT times, the authors performed rate adjustments to reactions sensitive to the IDT, yielding great agreement to the collected data. Zhang et al. [32] measured the laminar burning speed (LBS) of DIB1 with the constant pressure method at 400 K and 450 K. The constant pressure method needs extrapolation to calculate unstretched laminar burning speed, and the authors observed that the nonlinear method performed better than the linear method to calculate the laminar burning speed.

As it is clear, most of the mentioned literature is the only one concerned with the ignition characteristics of neat DIB or one of its isomers; the majority of the literature available focuses on surrogate mechanisms containing some portion of DIB [1,19,33,34]. A recent publication from Li et al. [1] compares experimental IDTs of three DIB-containing surrogates (0.0465–1.0072 mol%) to a quinary surrogate mechanism with a modified DIB submechanism from Metcalfe et al. [29]. The newly constructed model demonstrated superior performance over the previous DIB-containing surrogate mechanisms at conditions of 10, 15, and 20 bar, equivalence ratios of 0.5–2.0, temperatures of 950–1300 K, and exhaust gas recirculation (EGR) loadings ranging from 0% to 60%.

To this point, all chemical kinetic measurements in the literature for DIB or its isomers consist only of IDTs or LBS; however, many other parameters such as turbulent burning speeds, species time–histories, and heat release rate are essential for model validation. Therefore, in the present study, we also obtained measurements of carbon monoxide time–histories using laser absorption spectroscopy in the mid-infrared during the oxidation of DIB1 for model validation and improvement. Laminar burning speeds were measured at 1 atm and 428 K between equivalence ratios of 0.91–1.52 with the constant volume method. Time–histories of CO and ignition delay times were collected on a stoichiometric mixture of DIB1 between 8.5 and 10.9 atm and temperatures of 1186–1414 K.

Mixtures were prepared manometrically with 100 and 10,000 Torr capacitance manometers (MKS E27, error 0.1% of reading and MKS 628D, error 0.25% of reading, respectively). Argon and oxygen gases were supplied by Nexair with purities of 99.999%.

Laminar burning speeds were collected in UCF’s constant volume, spherical combustion rig, Fig. 1. Our earlier works can provide specific details about the subsections of the combustion chamber, such as the ignition circuit and mixture preparation process [39–41].

In this study, synthetic air was used as the oxidant and was created with 21.0 ± 0.01% mole of O₂ (99.999%, Praxair) and 79.0 ± 0.01% mole N₂ (99.999+%, Air Liquide) in the mixing tank. The prepared synthetic air was held for a minimum of 2 hr to achieve uniformity before conducting experiments. DIB1 from Acros Organics (Lot# A0370437, 99% pure) was used for both LBS and shock tube measurements. Table 1 presents detailed information on the studied compound.

In a vacuumed tank of less than 0.20 Torr, synthetic air was created to a total pressure of 19,760 Torr. Before injecting fuel, a furnace, which houses the combustion chamber, was set and maintained at 428 K to retain the fuel in a vapor phase. To check the condensation status of injected fuel, the vapor pressure of the fuel and the mass of injected fuel were compared. There was a linear relationship between the injected mass and vapor pressure, which indicates no condensation in the chamber. After checking the condensation status, synthetic air was introduced into the chamber until the desired initial experimental pressure was reached. The equivalence ratio (ϕ) was calculated by partial pressures of the fuel and synthetic air. After obtaining a stable and homogenous mixture in the chamber, the mixture was ignited by a modified NGK-BR4HS spark plug. A ceramic tube was installed on the modified spark plug to ensure the spark started at the top of the electrode. Schlieren imaging was used to capture flame propagation with a high-speed camera (v12.1, Phantom) with a resolution of 512 × 512 pixels at 20,000 fps. The pressure trace during the combustion event was recorded by a dynamic pressure transducer (603B1, Kistler) and charge amplifier (Type 5010, Kistler).

The UCF shock tube was used to measure temporally resolved speciation measurements of CO and ignition delay times of a 0.15%DIB1/1.8%O₂/Ar mixture over a temperature range of 1186–1414 K and pressures of 8.5–10.9 atm. Experimental and model ignition delay times are given in Fig. 2. The mechanisms demonstrated by Hu et al. [31] and Metcalfe et al. [29] are in great agreement at lower temperatures; however, a higher reactivity at elevated temperatures is predicted by Metcalfe et al.’s model. The rate constant adjustments to H-atom abstraction reactions of isobutene and DIB1, performed by Hu on Metcalfe’s mechanism, adjusted the reactivity at higher temperatures, improving ignition delay time predictions. The quinary, DIB1 containing gasoline surrogate mechanism from Li et al. [1] predicts a much longer ignition delay time over all temperatures in the displayed region. However, as shown in the study by Li et al. [1], there is good agreement with ignition data for a 0.75% DIB1 mixture performed at various equivalence ratios (0.5–1.0) and pressures (1–4 atm) throughout the entire temperature range (1200–1600 K). In addition, with an increase in pressure from 1 to 4 atm, the data also show an improvement in agreement. To gain insights into the reaction chemistry of this fuel, the discrepancy in modeled IDT compared to the current study’s data (10 atm) will be further discussed, along with a comparison to the mechanism of Hu. Emphasis will be placed on the mechanism of Li et al. due to the practicality of a five-component surrogate model, and if properly validated, it can be immensely beneficial in the design of advanced combustion technologies.

Measured CO profiles, compared to mechanisms several literature, are shown in Fig. 3. Speciation profiles are truncated at either the end of the test time when a compression or expansion wave is forced into the test section by the interaction of the reflected shock wave and the contact surface or at the time of ignition, where the pressure and temperature rise are significant. Nevertheless, there are often several milliseconds to compare model predictions against experimental data. Figure 3(a) shows excellent agreement with the mechanism presented by Hu et al. [31]. Because the mechanism presented Li et al. [1] overpredicts the ignition delay time, it underpredicts the rate of formation of CO. As the temperature is increased, specifically from Figs. 3(b) and 3(c), the predictive capabilities of the mechanism presented by Hu et al. deteriorate slightly, while the mechanism presented by Li et al. improves mildly, indicating a shift in the dominant chemistry for both models as temperature shifts.

From this pathway analysis, the relationship between allene and the critical radicals isobutenyl and 1-propenyl can be used to understand the governing chemistry responsible for the discrepancies observed with the mechanism of Li et al. [1] Concentration time–histories for isobutenyl and allene were output using the mechanisms presented by Hu et al. and Li et al. to understand their limitation or availability. Figures 6(a) and 6(b) show that there is a vast disparity between the predictions, with the mechanism presented by Hu et al. forming nearly four times more isobutenyl radicals and three times more allene. It should be noted that the differences in the overall trend/shape of allene formation indicate that allene chemistry in the mechanism by Li et al. is drastically different and likely a key factor. To better understand the limited formation of iC₄H₇ and aC₃H₄ in the mechanism by Li et al., pathway analyses were performed at times where the maximum formation of each intermediate was reached.

Figure 7 shows a pathway analysis of allene formation using both the mechanisms presented by Hu et al. and Li et al., with the directional rate of production (ROP) starting from DIB1. The ROPs (percentage) for outputs of the studied by Hu et al. and Li et al. are marked in black and blue, respectively. From this, one can observe a glaring difference in the amount of fuel decomposition between the two mechanisms, with 68.8% and 85.7% in the studies by in Li et al. and Hu et al. This difference in fuel decomposition may partially explain the lack of iC₄H₇ radicals that, in turn, limit the radical involvement via critical pathways (e.g., iC₄H₇ + O₂ = aC₃H₄ + CH₂O + OH). By limiting iC₄H₇ radical concentration, aC₃H₄ formation is hindered, which is necessary for reaction pathways to proceed for ignition and CO formation. In addition, the relationship between iC₄H₈ and iC₄H₇ from the mechanism by Li et al. shows discrepancies in ROP; moreover, the brute force sensitivity analysis shows the opposite effect on overall reactivity when compared to the Hu mechanism, specifically the reactions iC₄H₈ + H → iC₄H₇ + H₂ and iC₄H₇­ + H → iC₄H₈. From Fig. 5, it is also clear that consumption of aC₃H₄ resulting in the C₃H₃ radical is critical in CO formation. To better understand this chemistry, the pathway analysis of aC₃H₄ consumption leading to important intermediates necessary for CO production was performed, shown in Fig. 8. These pathway analyses for each mechanism were performed at the time near the max formation of aC₃H₄, and as shown in Fig. 6(b), this time is rather early in the mechanism by Li et al.

It is clear that a reduced iC₄H₇ radical pool, when compared to the mechanism of Hu et al., ultimately results in insufficient aC₃H₄ formation that is required to participate in the reactions presented in Fig. 8. The ROP of C₃H₃ in reactions of aC₃H₄ + OH → H₂O + C₃H₃ and aC₃H₄ + HO₂ → H₂O₂ + C₃H₃ shows a major difference between the mechanisms presented by Hu et al. and Li et al.; the former reaction also being sensitive to promoting IDT (from Fig. 4). This discrepancy in C₃H₃ and aC₃H₄ ROP between the two mechanisms explains the insufficient sensitivity to IDT as shown from the brute force analysis and ultimately describes how this chemistry is connected to the formation of CO.

Shown in Fig. 10 are CO time–history comparisons of the updated mechanism presented by Li et al. with the experimental data gathered in the current study. Very good agreement is achieved at the lower temperature condition of 1186 K; however, as the temperature increases, the predictability of the updated mechanism deteriorates, with vast disparity at the higher temperature of 1413 K. Although the poor agreement exists at elevated temperatures, the sensitive early low-temperature chemistry is well captured. For example, in Fig. 10(a), there seems to be very early CO chemistry occurring in the experimental profile (i.e., the early “hump”), which the updated mechanism captures with good agreement. Interestingly, this early CO chemistry is not prominent at elevated temperatures, although it is still mildly present. The early stages of CO formation are in very good agreement throughout the temperature spread; however, as time proceeds, this rate of formation is expedited, which is indicative of increased global reactivity (IDT) and is observed in data presented in Fig. 11. As stated previously, several of the changed reactions required input for reverse reaction rates. Even still, many (hundreds) reactions do not include reverse reaction rates, which may be a major contributing factor to the significant disagreement in the updated mechanism at the later stages of CO formation. Moreover, a multitude of isobutene and isobutenyl radical reactions are likely missing and/or do not match the most up-to-date rates found in the mechanism presented by Hu et al. Implementing these changes will require extensive modifications to Li et al.’s current quinary surrogate mechanism.

Furthermore, from the comparison shown in Fig. 3, the mechanisms of Li et al. and Hu et al. exhibit a shift in chemistry at elevated temperatures, where the predictions from Li et al. are in better agreement, and predictions from Hu et al. indicate increased reactivity. In general, all the mechanism predictions (with the original mechanism of Li et al.) in Fig. 3 show overreactivity at elevated temperatures. With sensitive allene and 1-propenyl chemistry matching the Hu et al.’s mechanism in the update, this seems to further support deficiencies in the isobutene submechanism as a limiting factor. It is hypothesized that the isobutene chemistry (C₀—C₄ kinetic mechanism) incorporated in the mechanisms compared in this study is not accurately described, resulting in poor predictions of experimental targets for DIB1 oxidation. The consumption and formation captured in isobutene chemistry are critical in accurately predicting these various experimental combustion targets of higher-order hydrocarbons, and continued efforts dedicated to elucidating these issues are necessary.

Uncertainty in reported CO mole fractions is calculated as a time-dependent quantity via the root-mean-square (RMS) of the uncertainties present in the parameters of Beer’s law, similar to the method presented in Ref. [39] except reported concentrations are not carried through the ignition event due to large changes in temperature and pressure in the current study. Root-mean-square uncertainties calculated in this manner for CO profiles range between 11.5% and 19.6% and are displayed as transparent regions in Fig. 3. Uncertainties in ignition delay times were calculated via Hu et al.’s mechanism [31] and Chemkin-PRO [43]. This was done by calculating the IDTs with the mechanism at T₅ and P₅ conditions and then calculating the IDTs at T₅ + UT₅, P₅ + UP5, and T₅-UT₅, P₅-UP5, where UT₅ (0.75–0.81% of T₅) and UP₅ (1.47–1.50% or P₅) are the uncertainties in temperature and pressure of state 5 conditions calculated from normal shock relations. By this means, the RMS uncertainty in IDT is found to be 12%.

The constant volume method with a multizone model was used to calculate LBS. For a more detailed explanation of this multizone model, the reader is directed to our earlier works [41,44,45]; this article provides only a brief explanation of the model. Two zones, burned and unburned, are considered in the multizone model. Equilibrium of thermodynamic properties in the burned zone was calculated for many discrete regions within the zone, [46], while the unburned mixture was considered as a single region. The cantera software [47], with Metcalfe et al.’s mechanism [29], was used to calculate the thermodynamic properties to obtain the burned mass fraction of species in each time-step.

O’Donovan and Rallis [48] and Bradley and Morley [49] derived an expression for the flame radius (Eq. (3)). Equation (3) is a function of the vessel radius (R_o), mass fraction (X), initial pressure (P_i), and instantaneous pressure (P) from a pressure sensor. Assumptions for this equation include: the flame is smooth and has a spherical shape, all gases are ideal, and the buoyancy of the fluids is negligible during the combustion process. Hill and Hung [50] and Takizawa et al. [51] derived Eq. (4), which is a theoretical expression of LBS with R_o, P, and P_i, flame radius (R_f), and burned mass fraction rate (dx/dt).

Figure 12 shows the DIB1 flame propagation image for an equivalence ratio of 1.1, which showed the highest LBS as captured with the high-speed camera. In Fig. 12, the left, center, and right images represent times at 2 ms, 4 ms, and 6 ms, respectively. The flame shape maintains its smooth and spherical shape as a laminar flame until it is no longer within the experimental viewing window on the combustion chamber.

Figure 13 presents LBS of DIB1 at 1 atm and 428 K in different equivalence ratios. The maximum LBS of DIB, 0.70 m/s, occurred at an equivalence ratio of 1.08. The present DIB1 LBS was compared with that presented in the study by Zheng et al. [32], which was conducted at an initial temperature and pressure of 450 K and 1 atm, respectively. Since higher unburned temperature results in higher adiabatic temperature, higher unburned temperature results in higher LBS [55]. Even though the initial temperatures were different, the present study and the previous findings were in good agreement, notably up to an equivalence ratio of 1.3. The present work was also compared with burning speed calculations of ethanol under the same initial conditions since ethanol is one of the standards of biofuels. DIB1 LBS is slower than ethanol LBS in the whole equivalence ratio range. For the simulations, the mechanisms presented by Metcalfe et al. [29], Hu et al. [31], and Li et al. [1] were used in Chemkin-PRO [43]. In the fuel-rich region, the present work was in the best agreement with the mechanism by Hu et al., whereas in the lean region, the present data were in best agreement with the mechanism by Metcalfe et al.

Carbon monoxide time–histories and ignition delay times were collected behind reflected shock waves during the combustion of 2,4,4-trimethyl-1-pentene (diisobutylene). Measurements were compared to the predictions of several chemical kinetic models in the literature, and the model of Hu et al. was found to perform very well over most conditions investigated. Discrepancies between model and experiment are present in the mechanism by Li et al. throughout the temperature range, although a better agreement is apparent at higher temperatures. Laminar burning speeds of 2,4,4-trimethyl-1-pentene were collected in a constant volume, spherical combustion chamber over a range of equivalence ratios, and a burning speed sensitivity analysis was conducted at different equivalence ratios. Burning speeds were found to be slightly less than ethanol velocities over the same conditions. The maximum laminar burning speed was recorded at an equivalence ratio of 1.08 with a velocity of 0.7 m/s. A burning speed sensitivity analysis was also conducted at different equivalence ratios.

Sensitivity and pathway analyses were used to discuss the differences in CO formation and overall reactivity (IDT) between the mechanisms presented by Hu et al., and Li et al. The sensitive chemistry of the quinary gasoline surrogate mechanism by Li et al. was compared to that by Hu et al. The disparity in chemistry involving isobutene and allene, in addition to isobutenyl and 1-propenyl radicals, was found. From this, sensitive reactions to IDT and CO formation were targeted and changed in the mechanism by Li et al. to match with the mechanism by Hu et al. This showed improved CO predictions during low-temperature chemistry, as well as early formation stages. However, the updated mechanism showed poor predictions of overall CO formation at higher temperatures. Deficiencies in the isobutene kinetics are likely to present in the mechanisms, which will need to be further investigated and updated in the future work. These effects will become exacerbated as the temperature increases. The multitude of reactions should be compared between the mechanisms presented by Hu et al. and Li et al. to elucidate the complexities of the quinary surrogate mechanism. As modifications are necessary to the surrogate mechanism, an emphasis on capturing low-temperature chemistry is paramount for its practical application in simulating the combustion of a real fuel.

This contribution was identified by Gaurav Agrawal (Exxon Mobil) as the Best Presentation in the session Developments in Alternative Fuels and Enabling Technologies I of the 2019 AIChE Annual Meeting in Orlando. The authors acknowledge help from Owen Pryor (presently at Southwest Research Institute, San Antonio, TX) for his assistance with experiments. This material is mainly based upon work supported by the U.S. Department of Energy’s Office of Energy Efficiency and Renewable Energy (EERE) under the Co-Optima initiative (Award Numbers DE-EE0007982, DE-EE0007984).

This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

There are no conflicts of interest.