A comprehensive experimental and modeling study of the ignition delay time characteristics of ternary and quaternary blends of methane, ethane, ethylene, and propane over a wide range of temperature, pressure, equivalence ratio, and dilution

A comprehensive experimental and modeling study of the ignition delay time characteristics of ternary and quaternary blends of methane


Introduction
The combustion of low-carbon fuels (C 1 -C 3 ) for energy and power generation is a very promising step towards future low-tozero emission energy production.Thus, gaining a deep understanding of the combustion chemistry of such fuels and their blends is important.Hence, the development of high-fidelity chemical mechanisms which can satisfactorily explain the oxidation and pyrolytic characteristics of low-carbon fuels is demanding.Together with speciation and laminar burning velocity techniques, ignition delay times (IDTs) are important in the validation of chemical kinetic  As these previous studies focused on the IDT studies of single and binary blended fuel mixtures, this paper intends to examine the IDT behaviours of ternary and quaternary blends of C 1 -C 3 hydrocarbons with relevance for engine and gas turbine applications.Furthermore, although there is enough data recorded in the literature for the IDT characteristics of C 1 -C 3 alkane blended fuels and natural gas mixtures (NG) [3][4][5][6][7][8][9][10][11] , evidence of the effect of adding an olefin to C 1 -C 3 alkane blends on IDTs has not been reported to date.Thus, towards developing a comprehensive IDT database, we have designed a set of IDT experiments to cover this void over a wide range of pressure, temperature, equivalence ratio, and dilution ( Fig. 1 ).The new experimental data sets are designed to explore the IDT characteristics of ternary and quaternary blends of CH 4 /C 2 H 4 /C 2 H 6 and CH 4 /C 2 H 4 /C 2 H 6 /C 3 H 8 mixtures.
Furthermore, providing stochastically and unbiasedly distributed experimental IDTs over a wide range of pressure, temperature, equivalence ratio, dilution with varying ethylene concentrations can help researchers develop more sophisticated and higher fidelity chemical kinetics.It can also lead to a valuable database capable of resembling the IDT characteristics of modified NG mixtures for a wide range of applications such as industrial furnaces and internal/external combustion engines working under homogenous charge compression ignition (HCCI), exhaust gas recirculation (EGR), and moderate or intense low oxygen dilution combustion regimes.
To achieve this, the current experimental and simulation study was defined, and the number of desired IDT experiments were op-timised, covering the target operating conditions discussed above using the Taguchi [5] design of experiments (DOE) approach as shown in Fig. 1 .However, high-pressure ( > 40 bar) tests were added to the Taguchi matrix to develop the database further.As shown in Fig. 1 (a), the composition of the blends is distributed diagonally to cover the target fuel compositions.Fig. 1 (b) demonstrates how using the Taguchi approach, we can reasonably populate the physical conditions in the desired cube so that a wide range of conditions can be covered without performing an overwhelming number of IDT experiments.The designed experiments encapsulate equivalence ratios of 0.5, 1.0, and 2.0 in 'air', at pressures ( p 5 and p C ) of 1, 20, 40, 90, and 135 bar, for diluent (N 2 and Ar) concentrations of 75%, 85%, and 90% of reactive mixtures in the temperature ( T 5 and T C ) range of ∼800 -20 0 0 K.The detailed kinetic mechanism NUIGMech1.2 is used to evaluate all of the newly measured experimental data, and the important reactions are identified to determine the synergistic/antagonistic effects of various blending effects on the IDTs.
The current study is organized in three stages, including (i) the design of new experiments over a wide range of operating conditions using the Taguchi approach; (ii) experimental measurements; and (iii) simulations using NUIGMech1.2.Comprehensive Supplementary material files containing non-reactive traces for RCM simulations, the original spreadsheets of experimental tests, L/HPST oscilloscope traces, and the combined figures of reactive, nonreactive, and modeling pressure traces are provided along with this paper.Moreover, the general information about the gasses (fuel/oxygen/argon/nitrogen), experimental facilities, and data acquisition systems are also accessible as Supplementary Material.

Design of experiments and experimental approach
As mentioned above, all of the new IDT experiments were designed using an L 9 Taguchi matrix to optimally reduce the number of experiments and time required.The Taguchi approach can tackle the issue using a specific design of orthogonal arrays which permits a comprehensive experimental investigation by doing a minimal number of experimental tests.In this regard, the minimum number of experiments is determined as follows: where, N Taguchi , NF , and L are the number of experiments, number of factors, and number of levels, respectively.According to the Taguchi approach, its performance is optimal when there are limited interactions between the desired variables.To use the Taguchi method, it is essential to define the controlling factors and levels.According to the factors and levels, several design of experiments (DOE) matrices are derived and are included as Supplementary material.The DOE process was followed for four parame-  ters of ternary and quaternary fuel combinations, pressure, equivalence ratio, and dilution at three levels, where the details are shown in Fig. 1 and Table 1 .As previously discussed by Baigmohammadi et al. [ 2 , 3 ], the new IDT experimental data presented in Table 1 were collected using low-and high-pressure shock tubes (L/HPST) and RCMs refers to the low pressure shock tube, the high pressure shock tube and the (red) rapid compression machine facilities at C 3 -NUI Galway, respectively, and PCFC refers to the RCM facility at PCFC RWTH Aachen University.
In the RCM facilities, the IDT of the normal studied mixtures (diluent concentration = 75%) and the pressure/time histories of their relevant non-reactive mixtures were recorded using a Kistler 6045A transducer mounted in the reaction chamber wall.However, the IDTs of mixtures at 85% and 90% dilution and at postcompression pressures of 20 and 40 bar were recorded using both the Kistler pressure record and light emission using a photomultiplier (PMT) equipped with a CH * filter due to the weak pressure signal observed at these diluted conditions.Therefore, as shown in the figure below, the IDT is defined as the maximum gradient in pressure or CH * after compressing the studied mixture.The following figure illustrates the IDT definition for experiments and simulations of the RCM data measured in the current work.More detailed information of the LPST, HPST, NUIG-RCM and PCFC-RCM data is included in Section 5 of the Supplementary material.

Uncertainty analysis
The details of the uncertainty analysis are provided as Supplementary material.However, a synopsis is presented here.The presented uncertainty analysis is adopted based on the methods ap-plied by Petersen et al. [6] and Weber et al. [7] .According to our analyses [ 2 , 3 ], the average uncertainties in the compressed mixture temperatures ( T 5 or T C ) and measured IDTs ( σ IDT %) in the NUIG STs and RCM are estimated to be approximately ± 20 K and ± 25% (in the H/LPSTs) and ± 5 -15 K and ± 20% (in the RCM), respectively over the entire range of cases studied.The uncertainty in the PCFC RCM is estimated using the methods described in Ramalingam et al. [8] , and for the compressed temperature, the uncertainty is estimated to be within ±5 K, with a measurement uncertainty of ± 0.15 bar for the compressed pressure and variation of ± 15% for the IDTs.

Computational modeling
NUIGMech1.2 is developed as a further development/refinement of NUIGMech1.1 with the addition and modification of several important reactions which are discussed in more detail in Section 4.2 below.This mechanism includes 2746 species and 11,279 reactions and reproduces similar good results as those previously published using NUIGMech1.1 for the oxidation of C 1 -C 6 species [9][10][11][12][13][14] .These experimental and theoretical studies include natural gas mixtures [10] , propane/propene blends [14] , propyne [11] , and the auto-ignition and pyrolysis studies of C 2 -C 6 alkenes [ 12 , 13 ].All of the simulations are conducted using a Python script based on the CANTERA [15] library for the ST simulation and CHEMKIN-Pro 18.2 [16] software for the RCM simulations.Details of these simulations have already been published [ 2 , 3 , 8 , 17-20 ].The effect of surface reactions on IDTs is ignored in our simulations [ 2 , 3 ].The definition of IDT is taken to be the maximum gradient of the CH * species dC H * dt | max or the maximum gradient of pressure  for the ST simulations.For the RCM simulations, the facility effects are included using the volume-time profiles derived from the non-reactive experimental pressure-time traces in which O 2 is replaced by N 2 in the mixture [ 18 , 21 ].
To identify the controlling chemistry both promoting and inhibiting the reactivity of the system and thus their effect on IDT predictions, we present sensitivity analyses based on the brute force definition, with the definition of sensitivity coefficient ( S) in Eqn.(2) [22] being: This sensitivity coefficient is calculated for every reaction included in the chemical kinetic mechanism of interest.The IDT ( τ ) is perturbed through direct changes in the pre-exponential factor in the Arrhenius equation.S can be either positive or negative, with a positive value corresponding to a reaction that inhibits reactivity giving longer IDTs and vice versa.Moreover, the flux analyses presented in this work are based on rate of production (ROP) analyses performed to track the consumption of the main components in the different mixture compositions as their intermediates species.
A global analysis of regressions implemented to correlate simulated IDTs assuming constant volume conditions using NUIG-Mech1.2 is presented in Section 4.4 .Moreover, the respective correlation equations and their coefficient values for specific conditions, based on various temperatures, pressure, equivalence ratios, and different fuel mixtures composition, are also provided.General correlations are intended to be a practical engineering tool to quickly and accurately calculate IDTs at the required condition.A more complete and fully detailed Tables S12 -S15 of correlation values, standard errors, and performance are included as Supplementary material.

Results and discussions
A comprehensive comparison between the experimental IDTs ( Table 1 ) and those predicted using NUIGMech1.2 is presented below.

Performance of NUIGMech1.2 and the correlations versus experimental data
The performance of NUIGMech1.  1 , is shown in Figs. 3 and 4 .The symbols refer to the experimental data, with the solid lines from NUIGMech1.2simulations and the dashed lines representing the correlation equation results.These correlations are derived over well-defined conditions, as discussed in Section 4.4 .
Figs. 3 and 4 show that NUIGMech1.2can reasonably reproduce all of the measured IDTs over the stochastically distributed conditions studied.It not only reproduces the effects of mixture composition and temperature on IDTs, but it can also reliably predict the effects of pressure, equivalence ratio, and the effect of dilution except at very high-pressures ( ≥ 90 bar) where the model underpredicts the IDTs for the ternary blends of the CH 4 /C 2 H 4 /C 2 H 6 mixtures.A detailed comparison between the performance of NUIG-Mech1.2 and other available mechanisms is provided in Section 9 At first glance, in Fig. 5 (a), the CH 4 /'air' mixture is the slowest to ignite over the entire temperature range studied in this work, while the C 2 H 6 /'air' is the next slowest in the low-temperature regime ( < 900 K), being ∼25% faster than CH 4 .However, C 2 H 6 /'air' becomes the second-fastest mixture to ignite, after ethylene, by inverting its trend at ∼ T C = 1100 K. Overall, C 2 H 6 /'air' mixtures are approximately an order of magnitude faster to ignite than CH 4 and are ∼50% faster compared to the C 3 H 8 /'air' mixtures, but are slower by a factor of ∼1.5 when compared to C 2 H 4 /'air' mixtures at temperatures higher than 1150 K.The C 2 H 4 /'air' mixtures are the fastest to ignite at high-temperatures, and exhibit shorter IDTs compared to the C 3 H 8 /'air' mixtures at temperatures below ∼1050 K, with propane being the fastest fuel to ignite in the lowtemperature regime.
To determine the chemistry controlling the reactivities of various fuel/'air' mixtures, reaction pathways based on multiple rate of production (ROP) analyses are illustrated in Fig. 6 [33] .This reaction channel directly competes with the chain-terminating reaction ĊH 3 + H Ȯ 2 ↔ CH 4 + O 2 for hydroperoxyl (H Ȯ 2 ) radicals, as shown in Fig. 7 (b), and the current mechanism utilizes the rate constant from Zhu et al. [34] for this inhibiting pathway.At the high temperature of 1450 K, ∼7% of methyl ( ĊH 3 ) radicals react with O 2 to form methyl peroxy (CH 3 Ȯ 2 ) radicals via ĊH 3 + O 2 ↔ CH 3 Ȯ 2 .These radicals further dissociate to CH 2 O and ȮH radicals, promoting reactivity as illustrated in Fig 7 (b).This direct dissociation of CH 3 Ȯ 2 radicals was not incorporated in the previous mechanism, and in NUIGMech1.2, the rate constant for this reaction is adopted based on an extensive study of rate rules by Villano et al. [35] derived using electronic structure calculations.According to the ROP and the sensitivity analysis of CH 4 /'air' mixtures, at high temperatures (1450 K) a significant quantity ( ∼55%) of the methyl radicals undergo self-recombination to form C 2 H 6 making methane the slowest fuel to ignite compared to the other single fuels.The self-recombination reaction of methyl radicals has been widely studied in the literature.Comparisons of pressure-dependent rate constants and various experimental measurements for ĊH 3 + ĊH 3 ( + M ) ↔ C 2 H 6 ( + M ) from the literature [37][38][39][40][41][42] are shown in Fig. 8 (b).NUIGMech1.1 employed a rate constant from the experimental and theoretical study of Wang et al. [41] for this reaction.Here NUIGMech1.2uses the fit derived from    a comprehensive experimental review by Blitz et al. [40] for the high-pressure limit rate constant.We have chosen this rate constant as the revised high-pressure limit is in good agreement with the recent experimental study by Sangwan et al. [38] in the temperature range 292 -714 K as well as with the high level theoretical calculation by Klippenstein et al. [42] .Fig. 8 (b) shows that the high-pressure limit reported by Wang et al. [41] is ∼17% slower than that measured by Sangwan et al. [38] at 714 K.In NUIG-Mech1.2 the low-pressure limit and the fall-off parameters are taken from Wang et al. [41] , as these satisfactory match the experimental measurements from Slagle et al. [37] and Glanzer et al. [39] in the pressure-dependent fall-off regime, Fig. 8 (b).
Unlike the case for CH 4 /'air' mixtures, the most important reaction promoting the reactivity at high temperatures for the binary, ternary and quaternary blends is the chain branching reaction Ḣ + O 2 ↔ Ӧ + ȮH.This is because C 2 H 4 , C 2 H 6, and C 3 H 8 mixtures generate larger concentrations of Ḣ atoms compared to CH 4 /'air' mixtures and consequently increases the fuel reactivities of the blended mixtures at the high-temperature conditions, as shown in Fig. 5 .In a previous study [4] , the important reactions, as well as the choice of their rate constants governing the oxidation behavior of C 2 H 4 /'air', C 2 H 6 /'air', C 3 H 8 /'air' mixtures and their binary blends, were discussed in detail.Thus, in this work, particular emphasis is placed on understanding the synergistic and antagonistic effects of ethane/ethylene/propane fuels on the ignition of methane/fuel mixtures.
The  On the contrary, the reactions CH 4 + Ḣ ↔ ĊH 3 + H 2 and CH 4 + ȮH ↔ ĊH 3 + H 2 O compete for the available Ḣ atoms and ȮH radicals which exhibit higher positive sensitive coefficients for the C 2 H 6. blended mixtures in comparison to the pure CH 4 case.
At low temperatures, the pure C 2 H 6 /air mixture is still faster to ignite compared to the pure CH 4 /'air' mixture, but the difference between the two decreases with a decrease in temperature, Fig. 5 (a).Furthermore, for the 50% CH 4 /50% C 2 H 6 binary blend, the IDT predictions overlap with the pure CH 4 ones at lowtemperatures.The sensitivity analysis shows that the sensitivities of strengthened by the addition of 50% C 2 H 6 to CH 4 /air mixtures.This is because, compared to the pure CH 4 case, H-atom abstraction from C 2 H 6 by H Ȯ 2 radicals exhibits a relatively larger negative sensitivity coefficient promoting reactivity, which is attributed to the higher rate constant of C 2 H 6 + H Ȯ 2 ↔ Ċ 2 H 5 + H 2 O 2 , which is approximately an order of magnitude faster than that for CH 4 + H Ȯ 2 ↔ ĊH 3 + H 2 O 2 .However, most of the Ċ 2 H 5 radicals formed add to O 2 producing ethylperoxy (C 2 H 5 Ȯ 2 ) radicals, followed by the concerted elimination reaction C 2 H 5 Ȯ 2 ↔ C 2 H 4 + H Ȯ 2, which plays an important role in inhibiting the ignition of the C 2 H 6 blended mixtures at low temperatures, Fig. 7 (a).
C 2 H 4 is an essential intermediate of C 2 H 6 oxidation and is the fastest fuel to ignite compared to the other single fuels at high temperatures.The effect on IDT predictions of the addition of C 2 H 4 to a CH 4 /C 2 H 6 /'air' mixture is presented in Fig 5 (a).The reactivity of the 50% CH 4 /25% C 2 H 6 /25% C 3 H 8 ternary blend is ∼15% faster than the CH 4 /C 2 H 6 binary blend at high temperatures ( > 10 0 0 K).Moreover, at low temperatures, the ternary blend is ∼80% faster than the binary blend.The main reason for the increased reactivity of C 2 H 4 blended mixtures at high temperatures is attributed to the substantial formation of Ḣ atoms due to the reaction sequence, reactions for the ternary blend compared to the binary blend mixtures, Fig. 7 (b).On the other hand, at low temperatures ( < 10 0 0 K), C 2 H 4 is primarily consumed by the addition of ȮH radicals producing hydroxyl-ethyl ( Ċ 2 H 4 OH) radicals which add to molecular oxygen to form hydroxyethylene-peroxy ( Ȯ 2 C 2 H 4 OH) radicals.The dissociation of Ȯ 2 C 2 H 4 OH radicals ultimately increases the reactivity of the ethylene blended mixtures at low temperatures by generating two molecules of formaldehyde and ȮH radicals.For the ternary mixtures, ȮH radicals are also formed through the reaction C 2 H 4 + H Ȯ 2 ↔ C 2 H 4 O1-2 (oxirane) + ȮH, further promoting reactivity.
Propane/'air' oxidation at high temperatures ( > 900 K) is mainly dominated by H-atom abstraction by ȮH radicals and Ḣ atoms, leading to the formation of n -propyl (n Ċ 3 H 7 ) and isopropyl  5 (a).The sensitivity analysis, Fig. 7 (b), shows that at 1450 K, the main chain branching reaction Ḣ + O 2 ↔ Ӧ + ȮH is relatively less sensitive to the IDT of the quaternary blends compared to the ternary blends leading to a lower reactivity of the 50% CH 4 /16.66%C 2 H 6 /16.66%C 2 H 4 /16.66%C 3 H 8 /air mixtures.Fig. 6 (b) shows that the addition of pure C 3 H 8 to CH 4 /C 2 H 6 /C 2 H 4 ternary blends becomes an important source of methyl radicals produced from the β-scission of n Ċ 3 H 7 radicals, is responsible for the slower reactivity of the quaternary blends compared to the ternary blends at higher temperatures.Fig. 7 (a) shows that at lower temperature the reactivity of the quaternary blends is governed by C 3 H 8 + ȮH ↔ n Ċ 3 H 7 + H 2 O being the most reactivity promoting channel, while the most sensitive reactions inhibiting the reactivity are C 3 H 8 + ȮH ↔ i Ċ 3 H 7 + H 2 O and n Ċ 3 H 7 O 2 ↔ C 3 H 6 + H Ȯ 2 .At T C < 900 K, n Ċ 3 H 7 radicals add to molecular oxygen producing n-propyl-peroxy (nC 3 H 7 Ȯ 2 ) radicals, which then isomerize to hydroperoxyl-propyl ( Ċ 3 H 6 OOH1-3) radicals that can add to molecular oxygen generating hydroperoxyl-propyl-peroxy (C 3 H 6 OOH1-3 Ȯ 2 ) radicals.The C 3 H 6 OOH1-3 Ȯ 2 radicals can further isomerize and generate a carbonyl hydroperoxide and an ȮH radical.Finally, the carbonyl hydroperoxide undergoes RO-OH bond cleavage, producing a second ȮH radical and a carbonyl-alkoxy (R Ȯ) radical in a chain branching process that increases the reactivity of propane at low temperatures, Fig. 6 (a).The addition of propane to the ternary blends significantly increases the reactivity of the 50% CH 4 /16.66%C 2 H 6 /16.66%C 2 H 4 /16.66%C 3 H 8 quaternary blends, this being just less reactive than the pure C 3 H 8 /'air' mixture at lower temperatures ( < 900 K), Fig. 5 (a).

Effect of pressure on ignition
Fig. 9 illustrates the effect of the pressure on IDTs for 50% CH 4 /25% C 2 H 4 /25% C 2 H 6 blend, 50% CH 4 /16.66%C 2 H 4 /16.66%C 2 H 6 /16.66%C 3 H 8 blend along with pure methane at ϕ = 1.0.Fig. 9 indicates that the reactivity of the mixtures increases at highpressure conditions due to the corresponding increase in concentration with pressure.Furthermore, it is observed that at lower temperatures ( < 830 K), the addition of C  10(b) show that for the ternary and quaternary blends, the total fluxes going through the ignition promoting pathways ĊH 3 + O 2 ↔ CH 3 Ȯ 2 and ĊH 3 + H Ȯ 2 ↔ CH 3 Ȯ + ȮH are around two times higher at 40 and 80 atm compared to 1 atm, while ∼32% less flux goes through the methyl radical recombination reaction ĊH 3 + ĊH 3 ( + M ) ↔ C 2 H 6 ( + M ) at 40 and 80 atm compared to 1 atm, thus increasing the overall reactivity of the blends at high pressure and low-temperature conditions.Moreover, for the ternary blend, as pressure rises, the carbon fluxes going through the channels generating hydroxyl-ethyl ( Ċ 2 H 4 OH) radicals from C 2 H 4 + ȮH followed by the O 2 addition and the subsequent dissociation of O 2 Ċ 2 H 4 OH leading to two formaldehyde and a hydroxyl radical increases, as seen in Fig. 10 (a

Correlation analysis
Reliable global correlations that can accurately reproduce experimental measurements are desirable tools for analytical, semiempirical, or computational fluid dynamics (CFD) calculations of reactive flows.Such correlations are versatile tools that can predict a mixture's sensitivity to any change in the correlated parameters.As discussed previously [ 2 , 3 ], these correlations significantly reduce the required simulation time in response to any change in the chemical system, a critical parameter in real-time combustion controlling and monitoring systems.A simple form of the correlations applied here follows those previously published [ 2 , 3 ] and is expressed in Eqn.(3) .As previously shown, this type of correlation can reasonably explain the IDT characteristics of single and binary blended C 1 -C 3 hydrocarbons over a wide range of conditions [ 2 , 3 ].Here, it should be noted that Eqn.(3) explains the general form of the correlations used in the study so that the "F" variable is zero for the tertiary blended C 1 -C 2 fuels where no propane is present.
where A represents the pre-exponential factor coefficient, B represents the activation energy divided by universal gas constant, and C -H represent coefficients for methane, ethylene, ethane, propane, oxidizer, and dilution, respectively.All of the correlations were derived using NUIGMech1.2based on thousands of adiabatic constant volume simulations, and their corresponding correlation coefficients are presented below and in Tables S12 -S15 of the Supplementary material.The correlation coefficients were calculated using a non-linear curve fitting routine available in OriginPro 8.5 [45] which provides linear regressions together with residuals, R 2 , χ 2 and standard errors for every coefficient value.Some of the examples of these conditions are presented in Fig. 5 (b) and Fig. 9 .However, according to the simple form of Eqn.(3) , the correlations are derived over ranges where the dependency of the IDT on the various parameters does not become highly non-linear.Moreover, for relatively long IDTs ( > 10 ms) measured using the RCM, the correlation results derived using the adiabatic constant volume calculations differs significantly from the simulations including heat loss effects for the facility.We have already discussed this effect in detail [ 2 , 4 ].
For 1100 ≤ T C ≤ 2000 K: For 800 ≤ T C ≤ 1100 K: At high temperatures (1100 -20 0 0 K), the coefficient associated with CH Furthermore, at high temperatures, the coefficient corresponding to the oxidizer is significantly higher than other coefficients showing the strong sensitivity of oxygen concentrations under the respective conditions.For the low-temperature range, 800 -1100 K, the coefficients for C 2 H 4 and C 3 H 8 become strongly negative.This is because, in this temperature range, C 2 H 4 and C 3 H 8 greatly enhance reactivity by producing higher concentrations of highly active ȮH radicals compared to the CH 4 and C 2 H 6 fuels.Moreover, the coefficient corresponding to the oxidizer in Eqns.(6) and 7 is comparable to other coefficients showing less importance on oxygen concentrations under these specific conditions.A more detailed correlation table is included for all conditions studied in this work as Supplemental material, which includes the coefficients, standard errors related to coefficients R 2 , and χ 2 .For both, low and high temperatures, the corresponding R 2 and χ 2 parameters are ranging from 0.985 -0.999 and 1.53 × 10 -4 -1.02 × 10 -11 , respectively.
Figures S51 and S52 in the Supplementary material compare the predictions of the quaternary correlations derived in this study with our prior published experimental IDTs of binary blends for C 2 H 4 /C 3 H 8 and C 2 H 6 /C 3 H 8 mixtures at high temperature and highpressure conditions [4] .The present correlations are also compared with the IDT predictions calculated using the binary correlations derived in our previous work [4] .It can be seen that the global correlation of the quaternary blend is unable to accurately predict the IDTs of the C 2 H 4 /C 3 H 8 and C 2 H 6 /C 3 H 8 binary blends.This is attributed to the fact that the present correlations were derived from simulations of quaternary mixtures with CH 4 as the major fuel component.Eq. (5) shows that, for quaternary mixtures, the coefficient associated with CH 4 is strongly positive compared to the other fuel components in the blend and thus CH 4 dominates the predicted reactivity of the fuel mixture.Since our previously published [4] binary blends did not include CH 4 as an additive component, the predictions of the present correlations differ significantly from those calculated using the binary mixture correlations [4] .

Conclusions
In the current study, a detailed experimental and kinetic modeling study of the IDT characteristics of C 1 -C 3 novel ternary and quaternary blends of CH 4 /C 2 H 4 /C 2 H 6 and CH 4 /C 2 H 4 /C 2 H 6 /C 3 H 8 mixtures was performed over a wide range of experimental conditions, temperature ( ∼750 -20 0 0 K), pressure (1 -135 bar), equivalence ratio (0.5 ≤ ϕ ≤ 2.0), and dilution ( ∼75 -90%).24 new IDT datasets, including approximately 360 data points were measured, which were not already available in the literature.Lowand high-temperature IDT characteristics of CH 4 /C 2 H 4 /C 2 H 6 and CH 4 /C 2 H 4 /C 2 H 6 /C 3 H 8 combinations were investigated using the ST and RCM facilities at NUIG and PCFC RWTH Aachen University.The results showed that NUIGMech1.2could predict the IDT characteristics of the blends studied with high fidelity over the wide range of conditions studied here.Therefore, NUIGMech1.2was used to perform studies on the blending and pressure effect on ignition.It was observed that for all blends used in this work, as the temperature and the pressure increase, the IDTs decrease.For high temperatures ( T > 1100 K), CH 4 exhibit the slowest reactivity because its chemistry is mainly driven by methyl chemistry, ĊH 3 + H Ȯ 2 ↔ CH 4 + O 2, and ĊH 3 + ĊH 3 ( + M ) ↔ C 2 H 6 ( + M ); these reaction pathways are responsible for the very slow ignition on pure CH 4 and the ternary and quaternary blends at this temperature.C 2 H 6 is the secondfastest mixture to ignite, and this is because the pressure dependant reaction from ethyl radical enhances the reactivity Ċ 2 H 5 ( + M ) ↔ C 2 H 4 + Ḣ ( + M ) at the same time, competition with the concerted elimination reaction Ċ 2 H 5 + O 2 ↔ C 2 H 4 + H Ȯ 2 slow down the ignition making C 2 H 6 slower than C 2 H 4 .C 2 H 4 mixtures show the fastest reactivity due to the vinyl ( Ċ 2 H 3 ) chemistry that makes C 2 H 4 the fastest fuel to ignite.Furthermore, C 3 H 8 is much slower than C 2 H 4 due to the high amount of methyl ( ĊH 3 ) radicals produced from n Ċ 3 H 7 channels.
It was observed that for low temperatures ( T < 1100 K), methyl chemistry is still responsible for the slow ignition exhibited by CH 4 .This is because of the large amount of ĊH 3 radicals reacting with O 2 to form methyl peroxy radicals ĊH 3 + O 2 ↔ CH 3 Ȯ 2, which further react through CH 3 Ȯ 2 + CH 3 Ȯ 2 ↔ CH 3 Ȯ + CH 3 Ȯ + O 2 to inhibit the reactivity.However, here C 2 H 6 is the second slowest due to the concerted elimination reaction, C 2 H 5 Ȯ 2 ↔ C 2 H 4 + H Ȯ 2 , which becomes very important at this temperature.Additionally, C 2 H 4 is the second-fastest in igniting due to hydroxy-ethyl-peroxy ( Ȯ 2 C 2 H 4 OH) and oxirane (C 2 H 4 O1-2) channels that enhance the reactivity.Moreover, C 3 H 8 exhibits the fastest ignition at this temperature due to the considerable amount of ȮH radicals generated

Fig. 1 .
Fig. 1.Experimental tests performed in the current study for% vol.composition of (a) ternary (CH 4 /C 2 H 4 /C 2 H 6 ) blends represented in solid symbols and quaternary (CH 4 /C 2 H 4 /C 2 H 6 /C 3 H 8 ) blends represented in half solid symbols, and (b) the input conditions for C 1 -C 3 blends studied in the current work at various equivalence ratios (x-axis), pressures (y-axis) and dilution levels (z-axis).

Fig. 2 .
Fig. 2. Definition for measuring IDT in the NUIG-RCM using Kistler pressure trace and PMT-CH * trace mounted on the side wall of the reaction chamber.

Fig. 5 .
Fig. 5. (a) Comparisons of IDT predictions for various single, binary, ternary, and quaternary fuels in air at p C = 40 atm and ϕ = 1.0, and (b) the predictions of their corresponding correlations (dotted lines).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

C 3 H
8 (blue) mixtures at p C = 40 atm and ϕ = 1.0, at the time of 20% fuel consumed for (a) T C = 800 K and (b) T C = 1450 K. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
4 in both blends (CH 4 /C 2 H 4 /C 2 H 6 /'air' and CH 4 /C 2 H 4 /C 2 H 6 /C 3 H 8 /'air') is strongly positive, while those for C 2 H 4 , C 2 H 6, and C 3 H 8 are negative.CH 4 produces high concentrations of ĊH 3 radicals, which are relatively less reactive compared to Ḣ atoms produced in the oxidation of C 2 H 4 , C 2 H 6 , and C 3 H 8 .Thus, increasing the CH 4 concentration in the ternary and quaternary blends will increase IDT, and increasing C 2 H 4 , C 2 H 6, and C 3 H 8 concentrations will decrease the mixture IDT.Moreover, the coefficient associated with C 3 H 8 in the quaternary (CH 4 /C 2 H 4 /C 2 H 6 /C 3 H 8 /'air') blend is smaller than the coefficients for C 2 H 4 and C 2 H 6 .This is because, at high temperatures, C 3 H 8 produces ĊH 3 radicals and C 2 H 4 molecules from n Ċ 3 H 7 radicals that decompose via β-scission reaction by n Ċ 3 H 7 ↔ C 2 H 4 + ĊH 3 , whereas C 2 H 4 and C 2 H 6 oxidation produces higher amounts of Ḣ atoms, thus more effectively enhancing the reactivity of the C 2 H 4 or C 2 H 6 blended mixtures than those of the C 3 H 8 blended ones.Therefore, although increasing the concentrations of C 2 H 4 and C 3 H 8 relative to CH 4 will increase mixture reactivity, the former will dominate the latter.