Dimethyl ether autoignition in a rapid compression machine: Experiments and chemical kinetic modeling

F U E L PR O CE SS I N G TE CH N O LOG Y 89 ( 20 0 8 ) 1 2 44 –1 2 5 4

a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m

w w w. e l s e v i e r. c o m / l o c a t e / f u p r o c

Dimethyl ether autoignition in a rapid compression machine: Experiments and chemical kinetic modeling Gaurav Mittal a , Marcos Chaos b,⁎, Chih-Jen Sung a , Frederick L. Dryer b a

Department of Mechanical and Aerospace Engineering, Case Western Reserve University, Cleveland, OH 44106, USA Department of Mechanical and Aerospace Engineering, Princeton University, Princeton, NJ 08544, USA

b

AR TIC LE I N FO

ABS TR ACT

Article history:

Dimethyl ether (DME) autoignition at elevated pressures and relatively low temperatures is

Received 15 February 2008

experimentally investigated using a rapid compression machine (RCM). DME/O2/N2

Received in revised form 27 May 2008

homogeneous mixtures are studied over an equivalence ratio range of 0.43–1.5 and at

Accepted 29 May 2008

compressed pressures ranging from 10 to 20 bar and compressed temperatures from 615 to 735 K. At these conditions RCM results show the well-known two-stage ignition

Keywords:

characteristics of DME and the negative temperature coefficient (NTC) region is noted to

Dimethyl ether

become more prominent at lower pressures and for oxygen lean mixtures. Furthermore, the

Ignition

first-stage ignition delay is found to be insensitive to changes in pressure and equivalence

Rapid compression

ratio. To help interpret the experimental results, chemical kinetic simulations of the ignition

Chemical kinetics

process are carried out using available detailed kinetic models and, in general, good

Induction chemistry

agreement is obtained when using the model of Zhao et al. [Int. J. Chem. Kinet. 40, 2008, 1– 18]. Sensitivity analyses are carried out to help identify important reactions. Lastly, while it is implicitly assumed in many rapid compression studies that chemical changes from the initial charge conditions that might occur during compression are negligible, it is herein shown with the help of Computational Singular Perturbation (CSP) analyses that chemical species formed during compression with little evolved exothermicity can considerably affect autoignition observations. Therefore, it is essential to simulate both compression and post-compression processes occurring in the RCM experiment, in order to properly interpret RCM ignition delay results. © 2008 Elsevier B.V. All rights reserved.

1.

Introduction

Environmental pollution, energy security, and future oil supplies are concerns that have driven the global community to seek nonpetroleum-based alternative fuels, along with more advanced energy technologies to increase the efficiency of energy use. Dimethyl ether (DME) appears to have a large potential as an energy source. DME has been proposed as a promising alternative to diesel fuels due to its high cetane number [1]. In recent years DME has emerged as a fuel additive to reduce particulate and NOx emissions due to its overall low

sooting and polluting potentials. DME can be mass-produced from extensive feedstocks, including natural gas, coal, and biomass; its physical properties are similar to those of Liquified Petroleum Gases (LPG), so DME can be stored and distributed using established LPG infrastructures [2,3]. In addition, DME can also be used as an ignition enhancer in propulsion systems and internal combustion engines [4,5]. Autoignition of engine-relevant fuels, such as DME, at practical conditions (i.e. high pressures and intermediate to low temperatures) is of fundamental importance to emerging technologies such as Homogeneous Charge Compression

⁎ Corresponding author. Department of Mechanical and Aerospace Engineering, Princeton University, Princeton, New Jersey 08544-5263, USA. Tel.: +1 609 258 2947; fax: +1 609 258 6109. E-mail address: [email protected] (M. Chaos). 0378-3820/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.fuproc.2008.05.021

1245

FUE L PR O CE SS I N G TE CH N O LO G Y 89 ( 20 0 8 ) 1 2 4 4–1 2 5 4

Ignition (HCCI) [6]. Furthermore, interest and advancements in fuel-flexible gas turbine power generation systems that use DME [7] require the knowledge of its ignition and burning characteristics. DME shock tube ignition studies are available in the literature [8–11]; however, with the exception of the studies of Pfhal et al. [8] and Zinner and Petersen [11], few DME ignition results are available at pressure and temperature conditions of practical interest. In kinetic research, several detailed chemical models for low and high temperature DME oxidation [12–17] have been developed and validated against multiple experimental observations. Of particular importance to the present study is the ability of available chemical kinetic models to accurately reproduce DME autoignition properties at engine-like conditions. DME displays the classical two-stage, negative temperature coefficient (NTC) ignition behavior similar to that observed with linear alkanes [6,8,18,19]. This behavior stems from low temperature reactions involving hydrocarbon radicals and molecular oxygen [19]. Therefore, a comprehensive detailed DME kinetic model for gas turbine and engine applications should correctly predict these low-temperature autoignition features. Due to the lack of ignition studies at elevated pressures and low-to-intermediate temperatures noted above, the focus of this work is to further the understanding of DME autoignition behavior under such conditions. Autoignition experiments are conducted for DME/oxidizer mixtures in a rapid compression machine (RCM) over a range of compressed pressures, compressed temperatures, and equivalence ratios. This experimental dataset is then used as a basis for validation and refinement of recently developed kinetic models for DME oxidation, with special emphasis on prediction of autoignition characteristics.

2.

Experimental

The RCM system used in the present investigation has been described in detail previously [20,21] and only a brief overview will be given here. The RCM consists of a driver piston, a reactor piston, a hydraulic motion control chamber, and a driving air tank. The driver cylinder has a bore of 12.7 cm and the reactor cylinder bore is 5 cm. The machine is pneumatically driven and hydraulically stopped. The machine allows variations of stroke and clearance height. The reaction chamber is equipped with sensing devices for measuring pressure and temperature, gas inlet/outlet ports for preparing the reactant mixture, and quartz windows for optical access. Additionally, the machine incorporates an optimized creviced piston head design to promote a homogeneous and adiabatic zone at the core of the reaction chamber [20]. Homogeneous reactant mixtures are prepared manometrically inside a mixing tank equipped with a magnetic stirrer. DME/O2/N2 mixtures are studied over the temperature range of 615–735 K, pressure range of 10–20 bar, and equivalence ratio range of 0.43–1.5. Table 1 lists the compositions of gas mixtures tested herein. Note that in Table 1, equivalence ratio (ϕ) is changed by altering the mole fractions of O2 and N2 while keeping a constant mole fraction of DME. Here, equivalence ratio is calculated by ϕ = 3 XDME / XO2, where

Table 1 – Molar composition and stoichiometries of gas mixtures tested DME

O2

N2

ϕ

ϕΩ

1 1 1

7 4 2

27 30 32

0.43 0.75 1.50

0.47 0.78 1.40

XDME and XO2 are the mole fractions of DME and O2, respectively. In certain studies, especially those dealing with emissions [22], ϕ may not be a good measure of mixture stoichiometry in comparing properties of oxygenated versus non-oxygenated fuels. In these cases, the definition of an “oxygen equivalence ratio,” ϕΩ [22,23], becomes relevant. For the purposes of the present study, however, the use of ϕ is sufficient as no comparisons to other fuels are performed and the main interest is in examining the temperature and pressure dependence of DME autoignition. For oxygenated hydrocarbon fuels (i.e. those containing H, C, and O) ϕΩ is defined as the amount of oxygen atoms required to convert all C and H atoms in the fuel/oxidizer mixture to CO2 and H2O divided by the amount of oxygen atoms present in the fuel/ oxidizer mixture. For the present DME mixtures, ϕΩ = 7XDME / (XDME + 2XO2) or ϕΩ = 7ϕ / (6 + ϕ). The oxygen equivalence ratio is also listed in Table 1. Note that the two definitions vary by no more than 9% over the conditions studied. In this RCM investigation, DME (supplied by Fisher Scientific) is 99.5% pure; O2 and N2 gases (supplied by Praxair) are of ultra high purity (99.993% and 99.999%, respectively). For a given mixture composition with known initial temperature, the compressed gas temperature at the end of the compression stroke (top dead center, TDC), Tc, is varied by altering the compression ratio; whereas the desired pressure at TDC is obtained by varying the initial pressure of the reacting mixture for a given compression ratio. The temperature at TDC is determined from measured pressures by the adiabatic core hypothesis according to the relation   Z Tc g dT Pc ¼ Ln P0 T0 g
1 T where P0 is the initial pressure, T0 is the initial temperature, γ is the temperature-dependent specific heat ratio, and Pc is the measured pressure at TDC.

3.

Numerical modeling

Due to heat loss to the combustion chamber walls, compression in an RCM is not truly adiabatic, and the pressure also decreases during the post-compression period. Consequently, a numerical model that accounts for the effect of heat loss is required to correctly simulate the RCM experimental data. For a properly designed creviced piston, such as the one used in the present experiments, isentropic core compression can be assumed and the effect of heat losses can be represented numerically by comparing the computed and measured pressure traces [21]. Compression may also be non-ideal due to small amounts of piston blow-by, but the method described below also accounts for this effect as well. In the present study, non-adiabatic effects are expressed through an adiabatic expansion by prescribing an

1246

F U E L PR O CE SS I N G TE CH N O LOG Y 89 ( 20 0 8 ) 1 2 44 –1 2 5 4

effective volume as a function of time, during and after compression. The volume expansion parameters are determined empirically from experiments using nonreactive mixtures with the same heat capacity, initial temperature, and initial pressure as the reactive cases. Details of this procedure can be found in [21] and [24]. The empirical effective volume parameters used for the specific DME experiments discussed herein are available in [25]. Numerical modeling of the experiments is performed using the SENKIN code [26] coupled with CHEMKIN II [27] libraries. The RCM is modeled as an adiabatic system with volume specified as a function of time. The modeling begins from the start of the compression stroke, and the importance of modeling both compression and post-compression processes will be further elaborated upon below. A recently developed kinetic model for DME oxidation and pyrolysis [17] is used for simulating the present experimental measurements. This model consists of 55 species undergoing 290 reversible reactions and has been validated for a wide range of physical conditions, including flow reactor species-time history profiles at pressures up to 18 atm and temperatures near 1000 K [14,15] as well as flow reactor reactivity profiles at 12.5 atm and initial reaction temperatures from 550 K–900 K [15], jet-stirred reactor species profiles up to 10 atm over the temperature range of 550 K–1100 K [10,12], pyrolysis in a flow reactor up to 10 atm [14,17], shock tube ignition delay measurements at low and high pressures [8–10], species profiles in low pressure and atmospheric burner stabilized flames [28–30], and laminar flame speeds up to 10 atm [31–33]. Updates to previous modeling work [14,15] have been continued by Curran and co-workers [16], and the most recent version of this model [34] is also used to produce comparisons.

4.

Experimental and numerical results

4.1.

Ignition delay time definitions and modeling approach

Fig. 1 shows a typical pressure trace for the ignition of a DME/ O2/N2 mixture (1/4/30 by mole) at initial conditions of 297 K and 591 Torr, along with the definitions of the first-stage

Fig. 2 – Experimental and simulated pressure traces (using the model of [17]) for the conditions shown in Fig. 1. Both reactive and nonreactive conditions are shown.

ignition delay (τ1) and the overall ignition delay (τ). Specifically, the ignition delays (τ1, τ) are defined as the time elapsed from the end of the compression stroke (t = 0) to the instant of maximum pressure gradient deduced from the pressure history. For the case shown in Fig. 1, the measured pressure and the deduced temperature at the end of compression (t = 0) are Pc = 15.1 bar and Tc = 651 K, respectively. Fig. 2 illustrates an example of the present RCM modeling. Experimental and simulated pressure traces for both the reactive mixture and the corresponding nonreactive mixture with the same specific heat ratio are shown and compared in Fig. 2. As explained above, based on the pressure trace of the nonreactive experiment, the parameters required for the heat transfer model (i.e. the effective volume parameters) are deduced [21]. The experimental and computed pressure traces for the nonreactive mixture are seen to match very well for both compression and post-compression events, indicating the adequacy of the present heat transfer model. The empirically-determined parameters are then used for simulating the corresponding reactive case. Furthermore, the present heat loss modeling approach ensures that along with the pressure history, the temperature history is also correctly simulated [24]. Compared to the experimental data, the simulated pressure trace obtained using the model of [17] for the condition of Fig. 1 over-predicts the first-stage and overall ignition delays.

4.2.

Fig. 1 – Pressure traces illustrating the ignition characteristics of DME and the definitions of the first-stage and overall ignition delays. Molar composition: DME/O2/N2 = 1/4/30.

CSP analyses and compression stroke effects

It is noted from Fig. 1 that while the compression time is ~30 ms, approximately 45% of the total pressure rise occurs in the last 2 ms of the compression stroke. This compression history is desirable in order to minimize the extent of chemical reaction during compression, as it may affect subsequent observations after compression ceases in two ways. First, if chemical induction processes that control the initial production of radicals begin during compression, radicals so generated may affect the observations after compression ceases. Secondly, if significant chemical enthalpy changes occur during compression, the temperature, Tc, immediately after compression will differ from that determined from the measured

FUE L PR O CE SS I N G TE CH N O LO G Y 89 ( 20 0 8 ) 1 2 4 4–1 2 5 4

Pc and the assumptions of isentropic, non-reactive compression with heat loss. If only small amounts of induction chemistry occur, negligible mixture chemical enthalpy changes will result, but overall ignition may still be affected by the small radical concentrations produced during compression. It is also well known from motored engine studies [35] that reactions may occur during the compression stroke producing intermediates species and heat release both of which may affect the observed chemical evolution and autoignition behavior over the engine cycle. In interpreting RCM observations, it has frequently been the case that the conditions used as reference for simulations have been the experimentally observed pressure and temperature at the end of compression (e.g. [36]). Recently, some investigators [37,38] have referenced autoignition observations to effective pressure and temperatures empirically derived from measured data. It is implicitly assumed in such approaches that chemical changes from the initial charge conditions that might occur during compression do not affect the calculations of autoignition delay parameters. Some RCM studies have also reported considerable fuel consumption during the compression stroke [39], inferring substantial chemical enthalpy changes. However, no analyses are available in the literature that investigate the significance of the above two sources of perturbations on observed ignition delays. Such analyses have been performed here and are summarized below. In the present study, a Computational Singular Perturbation (CSP) [40] methodology recently demonstrated by Kazakov et al. [41] is used. As developed [41], the methodology is applicable to the analysis of systems that can be modeled as constant volume processes (e.g. which has until recently [42] been a common assumption in modeling shock tube ignition delay). To analyze RCM data, the technique was further modified to accommodate systems with time-changing volume, as described by Li et al. [43]. Similar modifications can be made to accommodate modeling of shock tube phenomena under known pressure history [42]. The implementation of the CSP methodology used

1247

Fig. 4 – Modes with largest amplitudes obtained from CSP analysis for the conditions shown in Fig. 2 at 0.5 ms before the end of the compression stroke. Open bars correspond to stable modes, solid bars to explosive modes. The value of the real component of the associated eigenvalues is also shown.

in the present study is briefly outlined below. The chemical kinetic reaction system may be represented by the following set of ordinary differential equations dz ¼ gðzÞ; dt ~ ~ where z = [T y1 y2 … yn – 1 yn]T is the state variable vector, T the normalized temperature, yi the species mass fractions (n total), and g the overall reaction rate vector. Unlike prior implementations of CSP, the inclusion of temperature as one of the CSP state variables is essential for the direct analysis of thermokinetic feedback so that factors controlling ignition and heat release can be unambiguously determined. At any given time t the reaction rate vector can be differentiated (i.e. “perturbed”) with respect to time, dg/dt = J•g so that a local Jacobian matrix is defined, J = dg/ dz. One can, thus, perform the following decomposition on J: J ¼ VLV
1 ; where V = (v1 v2 …vn vn + 1 is the matrix of eigenvectors and Λ the diagonal matrix containing eigenvalues. The differentiation essentially yields a system of linear ordinary differential equations for g. Using the decomposition above, the reaction rate vector can be represented as a sum of individual modes: gðt þ DtÞc

nþ1 X

fi vi expðki DtÞ;

i¼1

Fig. 3 – Temperature (dashed lines) and CSP eigenvalue spectrum (solid lines) time evolution for ignition of a DME/O2/ N2 mixture (1/4/30 molar composition) initially at 523 Torr and 297 K (Pc = 20.1 bar, Tc = 720 K). Real eigenvalues are plotted corresponding to the leading modes (i.e. the modes with highest amplitudes). The figure insert shows results during the compression stroke.

where fi is the mode amplitude (indicating the mode importance) and λi the corresponding eigenvalue (indicating the mode time scale and physical behavior). It is evident from the equation above that the sign of the real component of the eigenvalues, Re(λi), provides information about the dynamics of the system. The modes with negative Re(λi) are referred to as stable (decaying) modes, while the modes with positive Re(λi) are unstable (explosive) modes. The explosive modes control the ignition behavior of the kinetic system [41]. Results from the CSP analyses applied to one of the most reactive cases (i.e. high Pc and Tc) considered in the present DME study are shown in Fig. 3. As expected, two-stage ignition behavior is clearly evident in the predicted temperature

1248

F U E L PR O CE SS I N G TE CH N O LOG Y 89 ( 20 0 8 ) 1 2 44 –1 2 5 4

profile. Also shown in Fig. 3 are the real parts of the eigenvalues, Re(λ), corresponding to the leading modes (i.e., the modes with highest amplitudes). Of special importance are the trends shown during the compression stroke (the inset of Fig. 3). CSP analyses reveal that the dominant mode during compression is a stable mode (i.e. Re(λ) b 0), which quickly loses its stable nature as the compression ends. Shown in Fig. 4 are the amplitudes of the leading modes at a time corresponding to 0.5 ms before the end of the compression stroke for the conditions shown in Fig. 3. Note that while an explosive mode with sufficiently small time scale (i.e. Re(λ) = 5.55 ms− 1) is present, its amplitude is nearly five orders of magnitude smaller than the leading stable mode and, thus, can be considered negligible, even at this late stage during compression. The explosive modes shown in Fig. 4 are mainly due to initiation reactions involving fuel and molecular oxygen. CSP results shown in Fig. 3 after the compression stroke reveal an interesting pattern and warrant further discussion. DME ignition exhibits similar patterns to those observed by Kazakov et al. [41] in studying the two-stage ignition of nheptane. In the beginning of the first stage, the ignition is characterized by a single explosive mode. As the system approaches the end of the first stage, in addition to the existing dominant explosive mode, another, slower explosive mode with a lower amplitude appears. The two explosive modes collapse, with their Re(λ) exhibiting a rapid decrease passing through zero. Hence, the two explosive modes lose Fig. 6 – Effect of the compression stroke on modeled (using the model of [17]) pressure traces for a DME/O2/N2 mixture (1/4/30 molar composition). Open symbols represent calculations performed considering the RCM compression stroke; lines are results obtained by initializing the calculations at TDC for the compressed pressure and temperature conditions listed and using the initial mixture composition. Heat loss effect is included in both calculations.

Fig. 5 – Predicted (using the model of [17]) species profiles and heat release during compression and ignition of DME/O2/N2 mixture (1/4/30 molar composition) at compressed conditions of 20.1 bar and 720 K.

their explosive nature, indicating the end of the first stage. After the end of the first stage, the system again develops a single dominant explosive mode that controls the ignition runaway during the second stage. The reactions governing the explosive modes that define the first stage are all low temperature branching processes involving the internal isomerization and decomposition of peroxy radicals, whereas the second ignition stage is nearly exclusively driven by the decomposition of hydrogen peroxide [41]. To further support the discussion above, Fig. 5 shows fuel and oxidizer profiles as well as the heat release rate during compression and ignition under the conditions studied above and shown in Fig. 3 (i.e. Pc = 20.1 bar, Tc = 720 K). It is evident that, during the compression stroke, there is very little chemical reaction and, consequently, no measurable exothermic/endothermic effects on system sensible enthalpy. However, during the compression stroke, radical initiation processes do begin to occur that subsequently play a role in the further development of the radical pool after compression ceases. Thus, even though there is little overall chemical reaction during compression, the ignition processes observed thereafter may be responsive to chemistry occurring during

FUE L PR O CE SS I N G TE CH N O LO G Y 89 ( 20 0 8 ) 1 2 4 4–1 2 5 4

the compression stroke. Such chemistry effects are included in the present modeling approach, as predictions are calculated based upon the initial conditions prior to compression and the pressure history, corrected for heat loss, over the entire experimental procedure. To further elaborate on the effects of small radical production during compression, model results are presented in Fig. 6 using two different modeling constraints: 1) the entire compression and post-compression processes are modeled using the experimental initial mixture conditions; 2) only post-compression processes are modeled using the thermodynamic state at TDC (i.e. Pc and Tc) as the initial pressure and temperature, and the initial experimental reaction mixture composition. While the effect of heat loss is accounted for in both calculations, the two modeling predictions can differ significantly in computed ignition delay following compression, especially for experimental conditions that result in short ignition delays. For example, from Fig. 6, at compressed conditions of 20.1 bar and 639 K, the two modeling approaches yield ignition delays of 31.87 ms and 34.31 ms, respectively. At 720 K, however, these values are 2.34 ms and 3.69 ms, respectively, almost a 60% difference. Fig. 7 shows the time histories of important radicals in the DME system (i.e. H, O, OH, HO2, CH3) during and after the compression stroke for the conditions shown in Figs. 3 and 5. It is noted that these radicals are all present in sub-ppm (i.e. HO2, CH3) and sub-ppb (i.e. H, O, OH) levels at the end of compression. The radical pool developed during compression, however small, can have a marked effect on characteristic ignition delay observations for the pressures and temperatures studied here, simply because at these conditions chemical induction processes have characteristic times that are comparable to the observed ignition delays. While the radical pool formed in the compression stroke has an important effect on the first-stage ignition delay, its effect on the second-stage chemical activities is minimal. Specifically, by defining the second-stage ignition delay, τ2, as the time elapsed from the end of the first-stage ignition to the instant of hot ignition (i.e. τ2 = τ − τ1), comparison of the computed pressure histories shown in Fig. 6 using the two

Fig. 7 – Evolution of main radicals (computed using the model of [17]) during compression and ignition of a DME/O2/N2 mixture (1/4/30 molar composition) at compressed conditions of 20.1 bar and 720 K.

1249

above-mentioned modeling approaches shows that τ2 is nearly independent of the chemical induction processes that occur during compression. The discussion above indicates that chemical induction processes are influenced by even small perturbations to the initial radical pool formed during compression. This result supports the views of Dryer and Chaos [44,45] put forth in analyzing the recently reported discrepancies between chemical kinetic predictions and experiments of syngas/air ignition delay [46]. As mentioned above, some rapid compression studies have modeled their experiments as constant volume processes using conditions at TDC [36] or experimentally determined effective pressures and temperatures [37,38] as initial reference parameters. These investigations [36,37] also report small differences (within experimental uncertainty) in computed ignition delays between these approaches and ones that include compression and post-compression (similar to the one used here.) However, the validity of modeling rapid compression experiments in this manner should always be verified since, as shown above, this can lead to considerable differences, especially for mixtures with short ignition delays, such as those studied by Mittal et al. [47]. Furthermore, recent shock tube studies [42,48] under similar pressure and temperature conditions as those studied here have drawn attention to the considerable preignition pressure rise that is observed in some shock tube experiments. This pressure rise can be considered as an adiabatic compression and treated in a very similar manner as the rapid compression processes described above, leading to substantial differences in the interpretation of the measured ignition delays in comparison to the use of constant volume (V) and internal energy (U) modeling assumptions, an approach commonly adopted by most shock tube ignition delay modelers over the last few decades. This approach is reasonable in terms of ignition delay interpretations at short residence times, but cannot be applied in interpreting long ignition delay data or those found for high energy density mixtures. Shock tube experimentalists have contributed to fostering this latter misapplication by continuing to apply the constant U, V assumptions to such data, and in reporting much of the shock tube ignition delay data without pressure history data where pre-ignition pressure changes may be discerned. A physical explanation of this behavior, however, has not been given in these recent works [42,48]. Michael and Sutherland [49] have discussed at length how pre-ignition pressure changes can arise from non-idealities due to boundary layer interactions and residual velocities behind the reflected shock at the end wall. It appears that at temperatures below around 1000 K, where chemical induction processes are of significance to measured ignition delays, these types of perturbations result in significant departure of observations from interpretation of the data as constant volume and internal energy processes. Moreover, many observations of ignition processes in this regime show them to be nonhomogenous in nature [45], leading to additional sources of pressure rise. Further research is needed to fully document the processes that lead to non-homogeneous ignition and subsequent pre-ignition pressure rise in diatomic bath gases and at high fuel/oxidizer concentrations.

1250

F U E L PR O CE SS I N G TE CH N O LOG Y 89 ( 20 0 8 ) 1 2 44 –1 2 5 4

Fig. 8 – Effect of equivalence ratio on experimental first-stage (open symbols) and total (filled symbols) ignition delays of DME at a compressed pressure of 10 bar; circles — ϕ = 0.43; triangles — ϕ = 0.75; squares — ϕ = 1.5.

Fig. 10 – Effect of compressed pressure on experimental firststage (open symbols) and total (filled symbols) ignition delays of DME for an equivalence ratio of 0.75; circles — Pc = 10 bar; triangles — Pc = 15 bar; squares — Pc = 20 bar.

Therefore, validation of chemical kinetic models using ignition delay data gathered in shock tubes and rapid compression machines for undiluted mixtures under elevated pressures and low-to-intermediate temperatures (similar to those studied here) strongly depend on the proper interpretation of experimental data [44]. The above discussion also supports that ignition delays measured under these conditions cannot be directly compared between different RCM and/ or shock tube studies without accounting for these “facility dependent” effects [48]. In summary, the present work shows that it is important to simulate the entire processes occurring in the experiment, including the compression stroke.

while keeping the mole fraction of DME constant. It is seen from Figs. 8 and 9 that τ increases for increasing equivalence ratio at higher compressed temperatures, whereas at lower compressed temperatures τ changes little with variations in equivalence ratio. It is also noted that the NTC behavior becomes more pronounced under fuel rich conditions. This behavior is different from ignition studies of fuel/air mixtures

4.3.

Experimental and modeling results

Figs. 8 and 9 show the effect of equivalence ratio on measured first-stage and overall ignition delays as a function of compressed temperature for compressed pressures of 10 bar and 15 bar, respectively. As mentioned earlier, equivalence ratio is varied through the change in the mole fraction of O2,

Fig. 9 – Effect of equivalence ratio on experimental first-stage (open symbols) and total (filled symbols) ignition delays of DME at a compressed pressure of 15 bar; circles — ϕ = 0.43; triangles — ϕ = 0.75; squares —ϕ = 1.5.

Fig. 11 – Measured (symbols) and computed (lines) ignition delays for a compressed pressure of 10 bar as a function of equivalence ratio. Simulations use the model of [17]. Dashed lines correspond to open symbols. For comparison, results obtained using the model of [34] for ϕ = 0.75 are shown as gray lines.

FUE L PR O CE SS I N G TE CH N O LO G Y 89 ( 20 0 8 ) 1 2 4 4–1 2 5 4

1251

(e.g. [8]) where τ in the NTC region decreases as the mixture stoichiometry changes from lean to rich. This is due to the fact that in the present study the fuel mole fraction does not vary and NTC reactivity is determined by the amount of O2 available in the system. It is well known [19] that the fuel peroxy species formed in the NTC region dictate the amount of radical branching. For the cases considered herein, rich mixtures have longer ignition times (i.e. slower branching) in the NTC region because peroxy radical formation is reduced due to the lower amount of oxygen present in the system compared to lean cases. Ignition delay variations as a function of compressed pressure are shown in Fig. 10 for an equivalence ratio of 0.75. It is noted that the NTC behavior becomes more pronounced at lower compressed pressures. Moreover, for the pressure conditions investigated, similar to the trends shown in Figs. 8 and 9, the first-stage ignition delay appears to be independent of pressure. Further analysis shows that τ1 is also a weak function of the mole fraction of oxidizer, leaving temperature as the primary variable. Further modeling results are compared with experimental data in Figs. 11–13. The predictions are observed to be in good agreement with the data and the general trends are properly reproduced. That is, compared to the experimental data, the model exhibits similar variations in first-stage and overall ignition delays with varying compressed pressure and equivaFig. 13 – Measured (symbols) and computed (lines) ignition delays for an equivalence ratio of 0.75 as a function of compressed pressure. Simulations use the model of [17]. Dashed lines correspond to open symbols. For comparison, results obtained using the model of [34] are shown for Pc = 15 bar as gray lines.

Fig. 12 – Measured (symbols) and computed (lines) ignition delays for a compressed pressure of 15 bar as a function of equivalence ratio. Simulations use the model of [17]. Dashed lines correspond to open symbols. For comparison, results obtained using the model of [34] for ϕ = 0.75 are shown as gray lines.

lence ratio. For comparison with the above predictions, all yielded using the model of Zhao et al. [17], representative results using an updated model [34] based on previous development [5,14–16] are also shown in Figs. 11 and 12 for ϕ = 0.75 and in Fig. 13 for Pc = 15 bar. Larger discrepancies are seen between computed and measured ignition delays when using this model, especially at the lower compressed temperatures. One of the trends observed in Figs. 11–13 is that the model of Zhao et al. [17], overall, reproduces the overall ignition delays better, whereas the first-stage ignition delays are consistently over-predicted. To further investigate this issue, a brute-force sensitivity analysis is performed to determine what reactions are controlling the ignition delay at compressed conditions of 15 bar and 650 K for an equivalence ratio of 0.75. The ignition delay sensitivity to each reaction in the model is computed by multiplying and dividing each reaction rate, k, by a factor of two, and computing the ignition delay (i.e. τign_up or τign_dwn for results obtained by adjusting rates upwards or downwards, respectively). The sensitivity coefficient is calculated on a logarithmic basis, as it is usually reported in the literature, namely S = ∂Ln(τ)/∂Ln(k) or, in this case, S = Ln(τign_up/τign_dwn)/Ln(4). The results of such an analysis are summarized in Fig. 14. Fig. 14 shows that at the pressure and temperature conditions studied herein, both first-stage and overall ignition

1252

F U E L PR O CE SS I N G TE CH N O LOG Y 89 ( 20 0 8 ) 1 2 44 –1 2 5 4

improved by changing this reaction rate without affecting the performance of model against other targets; however, further theoretical and experimental studies of the hydroperoxymethyl formate decomposition reaction are needed to substantiate this change.

5.

Fig. 14 – First-stage (s1) and overall (s) ignition delay time sensitivity to reaction rates for a DME/O2/N2 = 1/4/30 (ϕ = 0.75) mixture under the compressed conditions of Pc = 15 bar and Tc = 650 K.

delays are most sensitive to the isomerization of the methoxymethyl-peroxy radical, CH3OCH2O2, forming a hydroperoxy-methoxymethyl radical CH2OCH2O2H. CH2OCH2O2H may undergo two routes: 1) β-scission releasing two formaldehyde molecules and a hydroxyl radical (i.e. CH2OCH2O2H = 2CH2O + OH); 2) reaction with molecular oxygen to form the O2CH2OCH2O2H radical, which consequently isomerizes and decomposes releasing two OH radicals through the reactions O 2 CH 2 OCH 2 O 2 H = HO 2 CH 2 OCHO + OH and HO 2 CH 2 OCHO = OCH2OCHO + OH. This latter sequence provides chain branching at low temperatures. As seen in Fig. 14, the β-scission of CH2OCH2O2H is the main reaction “opposing” ignition; as temperature increases this step dominates leading to the negative temperature coefficient (NTC) region as the reactivity of the system decreases since only one reactive hydroxyl molecule is released. Rate expressions for the kinetic steps described above were optimized in the study of Zhao et al. [17] and any significant changes are likely to degrade agreement with other systems against which the model was validated. One possible change, however, is to the decomposition of hydroperoxymethyl formate (HO2CH2OCHO). Ignition delay times are most sensitive to this reaction at low temperatures (T b 650 K, approximately), with essentially no sensitivity at higher temperatures. This is also the case in other systems (e.g. flow reactors, flames, etc.) [17]. In the study of Zhao et al. [17] the rate of hydroperoxymethyl formate decomposition was increased; however, based on present results this rate may need to be increased further. An increase of a factor of two would reduce computed first-stage ignition delays by approximately 40% while minimally affecting total ignition delays. This change, however, would make the rate of hydroperoxymethyl formate decomposition considerably different from the values recommended by Sahetchian et al. [50] for the decomposition of organic hydroperoxides. Therefore, it is noted that model predictions of present RCM data can be

Conclusion

DME autoignition was studied in a rapid compression machine under a wide range of stoichiometries as well as compressed pressures and temperatures. Results show two-stage ignition behavior and the presence of a NTC region that is accentuated at lower pressures and for oxygen lean mixtures. First-stage ignition delays were found to be insensitive to variations in pressure and equivalence ratio. Available chemical kinetic models were used to simulate experimental results. The model of Zhao et al. [17], developed and validated a priori, was found to provide good agreement overall; thus, extending its range of validation. First-stage ignition delays were consistently over-predicted, however. Analysis of present RCM data coupled with results from model sensitivity analyses have helped identify the decomposition of hydroperoxymethyl formate as a reaction that may need updating to better reproduce ignition delays under the conditions of the present study; further experimental and theoretical studies of this reaction are needed. Modeling the compression process accurately is important to properly interpreting RCM ignition delay results. The use of effective thermodynamic parameters when modeling RCM ignition delay, as it is commonly done in the literature, cannot account for the establishment of the initial radical pool during compression and its effect on induction chemistry. This study has demonstrated the absence of essential thermochemical activity during compression; nonetheless, and for the first time, it is shown that induction chemistry occurring during compression, even resulting in negligible changes in the chemical energy release of the mixture, can result in substantive changes in the ignition delays observed after compression. Hence the specific design of the compression process in various RCMs as well as their heat loss characteristics must be taken into account in comparing observations among different experimental venues. This perspective also applies to the interpretation of shock tube ignition delay measurements at similar pressures and temperatures, i.e. conditions where observations are sensitive to chemical induction processes, which in general, are strongly sensitive to experimental perturbations.

Acknowledgements The authors would like to thank Prof. Henry Curran for providing an electronic copy of the model used in this study (Ref. 34) This work has been supported by the Air Force Office of Scientific Research under Grant No. FA9550-07-1-0515 and by the Chemical Sciences, Geosciences and Biosciences Division, Office of Basic Energy Sciences, Office of Science, U. S. Department of Energy under Grant No. DE-FG02-86ER13503.

FUE L PR O CE SS I N G TE CH N O LO G Y 89 ( 20 0 8 ) 1 2 4 4–1 2 5 4

REFERENCES [1] R. Verbeek, J.V. Weide, Global assessment of dimethyl-ether: comparison with other fuels, SAE Paper 971607, 1997. [2] T.H. Fleisch, A. Basu, M.J. Gradassi, J.G. Masin, Dimethyl ether: a fuel for the 21st century, Stud. Surf. Sci. Catal. 107 (1997) 117–125. [3] T.A. Semelsberger, R.L. Borup, H.L. Greene, Dimethyl ether (DME) as an alternative fuel, J. Power Sources 156 (2006) 497–511. [4] T. Amano, F.L. Dryer, Effect of dimethyl ether, NOxNOx, and ethane on CH4 oxidation: High pressure, intermediate-temperature experiments and modeling, Proc. Combust. Inst. 27 (1998) 397–404. [5] Z. Chen, X. Qin, Y. Ju, Z. Zhao, M. Chaos, F.L. Dryer, High temperature ignition and combustion enhancement by dimethyl ether addition to methane-air flames, Proc. Combust. Inst. 31 (2007) 1215–1222. [6] N. Iida, T. Igarashi, Auto-ignition and combustion of n-butane and DME/air mixtures in a homogeneous charge compression ignition engine. SAE Paper 2000-01-1832. [7] D. Cocco, V. Tola, G. Cau, Performance evaluation of chemically recuperated gas turbine (CRGT) power plants fuelled by di-methyl-ether (DME), Energy 31 (2006) 1446–1458. [8] U. Pfahl, K. Fieweger, G. Adomeit, Self-ignition of diesel-relevant hydrocarbon-air mixtures under engine conditions, Proc. Combust. Inst. 26 (1996) 781–789. [9] P. Dagaut, C. Daly, J.M. Simmie, M. Cathonnet, The oxidation and ignition of dimethylether from low to high temperature (500–1600 K): experiments and kinetic modeling, Proc. Combust Inst. 27 (1998) 361–369. [10] C.T. Bowman, D.M. Golden, R.K. Hanson, H. Pitsch, D.F. Davidson, A. Bardos, R. Cook, Z. Hong, P. Iyengar, S. Shashank, S. Vasu, K. Walters, R. Malhotra, GCEP Technical Report, 2006 http://gcep.stanford.edu/pdfs/QeJ5maLQQrugiSYMF3ATDA/ 2.6.5.bowman_06.pdf. [11] C.M. Zinner, E.L. Petersen, Methane and dimethyl ether oxidation at elevated temperatures and pressures. 2007 Fall Technical Meeting of the Eastern States Section of the Combustion Institute, University of Virginia, October 21–24 (2007) Paper A-09. [12] P. Dagaut, J.C. Boettner, M. Cathonnet, Chemical kinetic study of dimethylether oxidation in a jet stirred reactor from 1 to 10 atm: experiments and kinetic modeling, Proc. Combust. Inst. 26 (1996) 627–632. [13] H.J. Curran, W.J. Pitz, C.K. Westbrook, P. Dagaut, J.-C. Boettner, M. Cathonnet, A wide range modeling study of dimethyl ether oxidation, Int. J. Chem. Kinet. 30 (1998) 229–241. [14] S.L. Fischer, F.L. Dryer, H.J. Curran, The reaction kinetics of dimethyl ether. I: High-temperature pyrolysis and oxidation in flow reactors, Int. J. Chem. Kinet. 32 (2000) 713–740. [15] H.J. Curran, S.L. Fischer, F.L. Dryer, The reaction kinetics of dimethyl ether. II: Low-temperature oxidation in flow reactors, Int. J. Chem. Kinet. 32 (2000) 741–759. [16] X.L. Zheng, T.F. Lu, C.K. Law, C.K. Westbrook, H.J. Curran, Experimental and computational study of nonpremixed ignition of dimethyl ether in counterflow, Proc. Combust. Inst. 30 (2005) 1101–1109. [17] Z. Zhao, M. Chaos, A. Kazakov, F.L. Dryer, Thermal decomposition reaction and a comprehensive kinetic model of dimethyl ether, Int. J. Chem. Kinet. 40 (2008) 1–18. [18] K. Fieweger, R. Blumenthal, G. Adomeit, Self-ignition of S.I. engine model fuels: a shock tube investigation at high pressure, Combust. Flame 109 (1997) 599–619. [19] H.J. Curran, P. Gaffuri, W.J. Pitz, C.K. Westbrook, A comprehensive modeling study of n-heptane oxidation, Combust. Flame 114 (1998) 149–177. [20] G. Mittal, C.-J. Sung, Aerodynamics inside a rapid compression machine, Combust. Flame 145 (2006) 160–180.

1253

[21] G. Mittal, C.-J. Sung, A rapid compression machine for chemical kinetics studies at elevated pressures and temperatures, Combust. Sci. Tech. 179 (2007) 497–530. [22] C.J. Mueller, W.J. Pitz, L.M. Pickett, G.C. Martin, D.L. Siebers, C.K. Westbrook, Effects of oxygenates on soot Processes in DI diesel engines: experiments and numerical simulations, SAE Paper 2003-01-1791, 2003. [23] C.J. Mueller, The quantification of mixture stoichiometry when fuel molecules contain oxidizer elements or oxidizer molecules contain fuel elements, SAE Paper 2005-01-3705, 2005. [24] G. Mittal, C.-J. Sung, Autoignition of toluene and benzene at elevated pressures in a rapid compression machine Combust. Flame 150 (2007) 355–368. [25] http://www.mae.case.edu/facilities/cdl/projects/rapidcomp/ rapiddatabase. [26] A.E. Lutz, R.J. Kee, J.A. Miller, Senkin: a Fortran program for predicting homogeneous gas phase chemical kinetics with sensitivity analysis, Technical Report SAND87-8248, Sandia National Laboratories, 1998. [27] R.J. Kee, F.M. Rupley, J.A. Miller, Chemkin-II: a Fortran chemical kinetics package for the analysis of gas-phase chemical kinetics, Technical Report SAND89-8009, Sandia National Laboratories, 1989. [28] E.W. Kaiser, T.J. Wallington, M.D. Hurley, J. Platz, H.J. Curran, W.J. Pitz, C.K. Westbrook, Experimental and modeling study of premixed atmospheric-pressure dimethyl ether-air flames, J. Phys. Chem. A 104 (2000) 8194–8206. [29] A. McIlroy, T.D. Hain, H.A. Michelsen, T.A. Cool, A laser and molecular beam mass spectrometer study of lowpressure dimethyl ether flames, Proc. Combust. Inst. 28 (2000) 1647–1653. [30] T.A. Cool, J. Wang, N. Hansen, P.R. Westmoreland, F.L. Dryer, Z. Zhao, A. Kazakov, T. Kasper, K. Kohse-Höinghaus, Photoionization mass spectrometry and modeling studies of the chemistry of fuel-rich dimethyl ether flames, Proc. Combust. Inst. 31 (2007) 285–293. [31] C.A. Daly, J.M. Simmie, J. Würmel, N. Djebaïli, C. Paillard, Burning velocities of dimethyl ether and air, Combust. Flame 125 (2001) 1329–1340. [32] Z. Zhao, A. Kazakov, F.L. Dryer, Measurements of dimethyl ether/air mixture burning velocities by using particle image velocimetry, Combust. Flame 139 (2004) 52–60. [33] X. Qin, Y. Ju, Measurements of burning velocities of dimethyl ether and air premixed flames at elevated pressures, Proc. Combust. Inst. 30 (2005) 233–240. [34] H.J. Curran, personal communication (2007). [35] W.R. Leppard, The autoignition chemistry of n-butane: an experimental study, SAE Paper 872150, 1987. [36] M. Ribaucour, R. Minetti, L.R. Sochet, H.J. Curran, W.J. Pitz, C.K. Westbrook, Ignition of isomers of pentane: an experimental and kinetic modeling study, Proc. Combust. Inst. 28 (2000) 1671–1678. [37] X. He, M.T. Donovan, B.T. Zigler, T.R. Palmer, S.M. Walton, M.S. Wooldridge, A. Atreya, An experimental and modeling study of iso-octane ignition delay times under homogeneous charge compression ignition conditions, Combust. Flame 142 (2005) 266–275. [38] S.M. Walton, X. He, B.T. Zigler, M.S. Wooldridge, An experimental investigation of the ignition properties of hydrogen and carbon monoxide mixtures for syngas turbine applications, Proc. Combust. Inst. 31 (2007) 3147–3154. [39] A. Cox, J.F. Griffiths, C. Mohamed, H.J. Curran, W.J. Pitz, C.K. Westbrook, Extents of alkane combustion during rapid compression leading to single- and two-stage ignition, Proc. Combust. Inst. 26 (1996) 2685–2692. [40] S.H. Lam, Using CSP to understand complex chemical kinetics, Combust. Sci. Tech. 89 (1993) 375–404.

1254

F U E L PR O CE SS I N G TE CH N O LOG Y 89 ( 20 0 8 ) 1 2 44 –1 2 5 4

[41] A. Kazakov, M. Chaos, Z. Zhao, F.L. Dryer, Computational singular perturbation analysis of two-stage ignition of large hydrocarbons, J. Phys. Chem. A 110 (2006) 7003–7009. [42] G.A. Pang, D.F. Davidson, R.K. Hanson, Shock tube ignition delay times for hydrogen-oxygen-argon mixtures at low temperatures and elevated pressures. Proceedings of the Fall Meeting of the Western States Section of the Combustion Institute, Livermore, CA, October 16, 17, 2007, paper 07F-12. [43] J. Li, Z. Zhao, A. Kazakov, M. Chaos, F.L. Dryer, J.J. Scire Jr., A comprehensive kinetic mechanism for CO, CH2O, CH3OH combustion, Int. J. Chem. Kinet. 39 (2007) 109–136. [44] F.L. Dryer, M. Chaos, Ignition of syngas/air and hydrogen/air mixtures at low temperatures and high pressures: experimental data interpretation and kinetic modeling implications, Combust. Flame 152 (2008) 293–299. [45] M. Chaos, F.L. Dryer, Syngas combustion kinetics and applications, Combust. Sci. Tech. 180 (2008) 1051–1094. [46] E.L. Petersen, D.M. Kalitan, A.B. Barrett, S.C. Reehal, J.D. Mertens, D.J. Beerer, R.L. Hack, V.G. McDonell, New syngas/air

[47]

[48]

[49]

[50]

ignition data at lower temperature and elevated pressure and comparison to current kinetics models, Combust. Flame 149 (2007) 244–247. G. Mittal, C.-J. Sung, RA. Yetter, Autoignition of H2/CO at elevated pressures in a rapid compression machine, Int. J. Chem. Kinet. 38 (2006) 516–539. E.L. Petersen, M. Lamnaouer, J. de Vries, H. Curran, J. Simmie, M. Fikri, C. Schulz, G. Bourque, Discrepancies between shock-tube and rapid compression machine ignition at low temperatures and high pressures. Proceedings of the 26th International Symposium on Shock Waves, Göttingen, Germany, July 15–20, 2007, paper 0911. J.V. Michael, J.W. Sutherland, The thermodynamic state of the hot gas behind reflected shock waves: implication to chemical kinetics, Int. J. Chem. Kinet. 18 (1986) 409–436. K.A. Sahetchian, R. Rigny, J.T. De Maleissye, L. Batt, M. Anwar-Khan, S. Matthews, Pyrolysis of organic hydroperoxides (ROOH), Proc. Combust. Inst. 24 (1992) 637–643.