Reduced Mechanism for Combustion of Ammonia and Natural Gas Mixtures

: A fuel mixture of ammonia and natural gas as a low-carbon alternative for future power generation and transportation is an attractive option. In this work, a 50-species reduced mechanism, NH 3 NG, suitable for computational ﬂuid dynamics simulations (CFD), is developed for ammonia– natural gas coﬁring while addressing important emission issues, such as the formation of nitrogen oxides (NOx), soot, carbon monoxide, and unburnt methane/ammonia. The adoption of reduced mechanisms is imperative not only for saving computer storage and running time but also for numerical convergence for practical applications. The NH 3 NG reduced mechanism can predict soot emission because it includes soot precursor species. Further, it can handle heavier components in natural gas, such as ethane and propane. The absolute error is 5% for predicting NOx and CO emissions compared to the full Modiﬁed Konnov mechanism. Validation with key performance parameters (ignition delay, laminar ﬂame speed, adiabatic temperature, and NOx and CO emissions) indicates that the predictions of the reduced mechanism NH 3 NG are in good agreement with published experimental data. The average prediction error of 13% for ignition delay is within typical experimental data uncertainties of 10–20%. The predicted adiabatic temperatures are within 1 ◦ C. For laminar ﬂame speed, the R 2 between prediction and data is 0.985. NH 3 NG over-predicts NOx and CO emissions, similar to all other literature methods, but the NOx predictions are closer to the experimental data.


Introduction
Since wind and solar power sources are intermittent in nature and are generally far from urban centers, there is a compelling need to store large quantities of renewable energy with a faster response. The popular power from hydrogen concept via a water splitter can be used to create ammonia by reacting hydrogen with nitrogen in an air separation unit (ASU). Ammonia is currently manufactured at a large scale as an industrial chemical and fertilizer. Owing to its various merits-ammonia (1) is easy to store as a liquid; similar to propane; (2) has a high energy density; and (3) handling experience and infrastructure are already available-nowadays, ammonia is widely considered as an important hydrogencarrier for future marine shipping or commercial aviation to drive turbines or engines, Figure 1. Ammonia can be easily cracked back to pure hydrogen for those applications that use pure hydrogen or a mix of hydrogen with other fuels, including ammonia itself [1][2][3][4][5][6].
However, through fuel NOx mechanisms, nitrogen-containing ammonia can lead to significant NOx emissions. Further, there are also concerns about unburnt ammonia and slow ammonia kinetics at low temperatures. On the other hand, using natural gas (NG) as a fuel has distinct advantages, such as (1) Higher hydrogen content relative to gasoline, diesel, and coal; (2) High adiabatic flame temperature and high laminar flame speed; (3) Well-established infrastructure; (4) Abundance of natural gas reserves in the US. (5) A worldwide increase in liquefied natural gas (LNG) plants and terminals.
(1) Higher hydrogen content relative to gasoline, diesel, and coal; (2) High adiabatic flame temperature and high laminar flame speed; (3) Well-established infrastructure; (4) Abundance of natural gas reserves in the US. (

5) A worldwide increase in liquefied natural gas (LNG) plants and terminal
Because of the above-mentioned advantages, cofiring ammonia with natu gained significant attention as a clean source of energy in gas turbines for pow tion as well as in internal combustion engines (ICE) for transportation. Even th ing ammonia with natural gas can ease the above-mentioned emission conce busting ammonia alone, additional emissions, such as soot, carbon monoxid hydrocarbons, and Volatile Organic Compounds (VOCs), still need to be ad 19]. There are full mechanisms in the literature that address combustion inv monia in the air. The detailed Konnov mechanism for ammonia consists of and 957 reactions [11]. A new version of the Konnov model targeting the amm and an improved hydrocarbon subset has also been reported [13,14]. The CEUanism for ammonia and methane/methanol/ethanol contains 91 species and 44 [15]. The UC San Diego mechanisms are a suite of mechanisms, including ni hydrocarbon-based chemistry [16]. The USCII full mechanism, developed b Hai Wang, is a detailed kinetic mechanism tailored for hydrogen and C1 to C4 c with 111 species in 784 reversible reactions [17][18][19].
The objective of this study is to develop a reaction mechanism that can computational fluid dynamics' (CFD) modeling to predict NOx/soot/unbur burnt hydrocarbons/volatile organic compound (VOC) emissions under gas t internal combustion engine (ICE) conditions. The Konnov and the USCII m [10,[17][18][19] were combined to become the full mechanism (Modified Konnov M The full Konnov mechanism has 129 species. Additional species (C4H10, pC4H9, and associated mechanisms from USCII were added to the Konnov mechanis the Modified Konnov Mechanism, which has 132 species and 1238 reactions.
Reduced mechanisms are required for practical turbine/ICE applications There are full mechanisms in the literature that address combustion involving ammonia in the air. The detailed Konnov mechanism for ammonia consists of 129 species and 957 reactions [11]. A new version of the Konnov model targeting the ammonia flame and an improved hydrocarbon subset has also been reported [13,14]. The CEU-NH 3 mechanism for ammonia and methane/methanol/ethanol contains 91 species and 444 reactions [15]. The UC San Diego mechanisms are a suite of mechanisms, including nitrogen and hydrocarbonbased chemistry [16]. The USCII full mechanism, developed by Professor Hai Wang, is a detailed kinetic mechanism tailored for hydrogen and C 1 to C 4 combustion with 111 species in 784 reversible reactions [17][18][19].
The objective of this study is to develop a reaction mechanism that can be used in computational fluid dynamics' (CFD) modeling to predict NOx/soot/unburnt NH 3 /unburnt hydrocarbons/volatile organic compound (VOC) emissions under gas turbine and internal combustion engine (ICE) conditions. The Konnov and the USCII mechanisms [10,[17][18][19] were combined to become the full mechanism (Modified Konnov Mechanism). The full Konnov mechanism has 129 species. Additional species (C 4 H 10 , pC 4 H 9 , and nC 3 H 7 ) and associated mechanisms from USCII were added to the Konnov mechanism to make the Modified Konnov Mechanism, which has 132 species and 1238 reactions.
Reduced mechanisms are required for practical turbine/ICE applications due to the complexity of 3D computational fluid dynamics' (CFD) simulations. The CFD simulations are widely used in turbine, engine, and furnace design for analyzing flame stability, autoignition zone, temperature/velocity profiles, fuel/oxidizer mixing patterns, and emission rates. However, coupling detailed reaction mechanisms with momentum, heat, mass transport equations, and stiff differential equations proved to be a serious burden on computational speed and numerical convergence. Thus, it is imperative to adopt reduced mechanisms not only for saving computer storage and running time but also for numerical convergence for practical applications. For instance, ANSYS Fluent sets the maximum number of species in the reactions to 50. One of the goals of this paper is to contribute an ammonia-methane cofiring reduced mechanism that can handle heavier hydrocarbons (C 2 and C 3 ) in natural gas streams, Table 1. Further, many existing reduced mechanisms for ammonia and methane mixtures are unsuitable for predicting soot emissions. Therefore, the developed reduced mechanism NH 3 NG also included soot precursor species C 2 H 2 and C 2 H 4 to facilitate soot emission estimation. NH 3 NG compares well with the full mechanism and existing reduced mechanisms when validated against experimental ignition delay, adiabatic temperature, laminar flame speed, and NOx/CO emissions [20][21][22][23][24][25][26][27][28][29][30][31][32][33].

Methodology
The mechanism reduction was performed using Chemkin 2019 R2 [33][34][35][36]. The following steps were taken to determine the 50 species to be included in the reduced mechanism NH 3 NG:

1.
The species in the reduced mechanism should be suitable for the targeted applications in turbine/engine design with CFD. In this study, the reduced mechanism should include fuel species, emission species (NO, NO 2 , NH 3 , HCN, N 2 O), and soot precursors (C 2 H 2 , C 2 H 4 ).

2.
The next step is to utilize the Reaction Path Analyzer (RPA) tool in Chemkin. RPA provides a visualization of the inner relationships of the chemistry model, as shown in Figure 2.

3.
The final step in species selection is to apply the Reaction Rate Analysis (RRA) tool. In this Chemkin analysis, the Perfectly Stirred Reactor (PSR) model is adopted. In these PSR runs, the species with a higher rate of production or destruction are expected to be dominant in the reaction mechanism and, thus, are ranked higher.
The soot precursor species C2H2 and C2H4 are important in that the soot emission can be predicted with precursor-based soot models, such as the Moss-Brookes' equation (with the built-in Fenimore-Jones soot oxidation model) listed in ANSYS Fluent [26][27][28][29][30][31][32]. Additional species (C, CH, CH2, C3H3, and C4H4) were also included as they were part of the reaction intermediates.

Reaction Path Analyzer (RPA)
The reaction path analyzer (RPA) displays reaction pathways connecting the species. Since the path width is proportional to its rate of production, RPA helps identify the main intermediates during reactions and eliminates any species with negligible contributions [28]. In this study, reaction paths between fuel components and the major intermediates/products (CO, CO2, NO, NO2, C2H4) were analyzed. The major contributing reactions and participating species were identified. Figure 2 shows the reaction pathways for CO production.

Reaction Rate Analysis (RRA)
The important intermediate species were selected based on the active participation of the species in the reaction. To determine the key intermediate species, the production and destruction rates were added and sorted to obtain the ranks of the species. The species with a higher rate of production or destruction were ranked higher. From the rankings list, it was observed that most of the species were already included in the reduced mechanism. However, some species, such as C, CH, CH2, C3H3, and C4H4, were still needed (e.g., for soot production). Therefore, they were subsequently added to the list, and the mechanism was reduced further. Figure 3 gives the reaction rate analysis for the key intermediate C2H4, a soot precursor species. Table 2 shows the final 50 selected species.

Preliminary Species Selection
Mechanism reduction was conducted at a temperature of 1500 K and pressure of 300 bar, using Chemkin 2019 R2 [36]. For the reduced mechanism, the initial species list included Fuel Species (NH 3 , CH 4 , C 2 H 6 , C 3 H 8 ), NO X and NO y (NO, NO 2 , HONO), Soot Precursors (C 2 H 2 , C 2 H 4 ), and Air (O 2 , N 2 , Ar).
The soot precursor species C 2 H 2 and C 2 H 4 are important in that the soot emission can be predicted with precursor-based soot models, such as the Moss-Brookes' equation (with the built-in Fenimore-Jones soot oxidation model) listed in ANSYS Fluent [26][27][28][29][30][31][32]. Additional species (C, CH, CH 2 , C 3 H 3 , and C 4 H 4 ) were also included as they were part of the reaction intermediates.

Reaction Path Analyzer (RPA)
The reaction path analyzer (RPA) displays reaction pathways connecting the species. Since the path width is proportional to its rate of production, RPA helps identify the main intermediates during reactions and eliminates any species with negligible contributions [28]. In this study, reaction paths between fuel components and the major intermediates/products (CO, CO 2 , NO, NO 2 , C 2 H 4 ) were analyzed. The major contributing reactions and participating species were identified. Figure 2 shows the reaction pathways for CO production.

Reaction Rate Analysis (RRA)
The important intermediate species were selected based on the active participation of the species in the reaction. To determine the key intermediate species, the production and destruction rates were added and sorted to obtain the ranks of the species. The species with a higher rate of production or destruction were ranked higher. From the rankings list, it was observed that most of the species were already included in the reduced mechanism. However, some species, such as C, CH, CH 2 , C 3 H 3 , and C 4 H 4 , were still needed (e.g., for soot production). Therefore, they were subsequently added to the list, and the mechanism was reduced further. Figure 3 gives the reaction rate analysis for the key intermediate C 2 H 4 , a soot precursor species. Table 2 shows the final 50 selected species.

Comparison with the Full Mechanism
The new reduced mechanism was compared against the full modified Konnov mechanism at a residence time of 1 s and ignition temperature of 1500 K, and pressure of 300 bar. The predicted mole fractions of the important combustion products obtained from the Chemkin simulation of the reduced mechanism at various residence times were recorded. Table 3 summarizes the prediction errors at a 1 s residence time for an ammoniamethane mixture. The predicted mole fraction differed from the full Modified Konnov Mechanism mole fractions by an average absolute error of 1.90 × 10 −6 for the major species. Table 3. Comparison of prediction errors of reduced mechanisms for the mole fraction of the major species at a residence time of 1 s for CH4-NH3 fuel.

Comparison with the Full Mechanism
The new reduced mechanism was compared against the full modified Konnov mechanism at a residence time of 1 s and ignition temperature of 1500 K, and pressure of 300 bar. The predicted mole fractions of the important combustion products obtained from the Chemkin simulation of the reduced mechanism at various residence times were recorded. Table 3 summarizes the prediction errors at a 1 s residence time for an ammonia-methane mixture. The predicted mole fraction differed from the full Modified Konnov Mechanism mole fractions by an average absolute error of 1.90 × 10 −6 for the major species. The average absolute % error of 5.44% between the NH 3 NG reduced mechanism and modified Konnov full mechanism is reasonable, considering the goals of NH 3 NG are to accommodate soot prediction and to handle C 2 -C 3 hydrocarbon species. All in all, NH 3 NG offers a simple mechanism (with 50 species) suitable for CFD applications. The larger absolute % error of 11.79% for NH 3 represents the compromise between the original Konnov mechanism designed for NH 3 combustion and the USC II mechanism designed for hydrocarbon combustion.

Validation with Experimental Data
The 50 species reduced mechanism for ammonia-natural gas cofiring NH 3 NG was validated against key experimental performance indicators such as laminar flame speed, ignition delay, adiabatic temperature, and NOx and CO emissions [16,20,21]. The reduced mechanism containing 50 species and reactions, along with the thermodynamics file, was preprocessed using the Chemkin-Pro preprocessor. The validated reduced mechanism, NH 3 NG, can then be used in other ANSYS FLUENT CFD simulations .

Laminar Flame Speed
Flame speed is a fundamental property of the fuel-air mixture that influences the design of combustion equipment. Laminar flame speed is the speed at which a laminar flame propagates through a pre-mixture of fuel and air. The flame speed depends on the properties of the fuel mixture and the thermodynamic conditions, such as temperature and pressure, when ignited.
The flame speed calculation model in Chemkin was used to evaluate the performance of the reduced mechanism against the experimental values of the NH 3 /CH 4 /air system found in Rocha et al. [20]. The experimental results were used to validate the mechanism for different mixtures of ammonia and methane in the air. During validation, the temperature and pressure inside the reactor were set at 423 K and 1-3 atm., respectively. The equivalence ratio varied between 0.8 and 1.2 for all fuel mixtures. The ammonia blending fractions in the binary fuel mixtures were also varied, as shown in Table 4. The equivalence ratio is defined as the ratio of the actual fuel/oxidizer ratio to the fuel/oxidizer ratio in the stoichiometric equation. Many properties of combustion processes strongly depend on the stoichiometry of the combustion mixture. Table 4 shows an average absolute error of 7.7%, which is well within the reported experimental uncertainty of 10% [28]. As shown in Figure 4, the predicted results from the NH 3 NG mechanism agreed with the experimental data very well, with an R 2 = 0.985.

Ignition Delay
The period between the creation of a combustible mixture when the fuel is injected into an oxidizing environment and the sustained onset of the rapid reaction phase leading to a rise in temperature and pressure is defined as the ignition delay time. For ignition delay time calculations, a closed homogeneous reactor model was used. The reduced mechanism was compared against the experimental data from J. Huang et Al. [30]. For validation in Chemkin, the pressure was taken as 16 atm. and the temperature varied between 1227 and 1307 K. The fuel composition was CH4/C2H6/O2/N2 (8.93-0.34-19.05-71.68%). Table 5 shows the numerical comparison between the predictions and lab data with an average absolute error of 13%, which is well within the combined temperature/concentration uncertainties (8%) [30] and typical experimental data uncertainties (10-20%) [33]. Figure 5 shows the graphical representation of the experimental and simulation results between ignition delay time and for the said mixture with the new mechanism at the conditions mentioned above.

Ignition Delay
The period between the creation of a combustible mixture when the fuel is injected into an oxidizing environment and the sustained onset of the rapid reaction phase leading to a rise in temperature and pressure is defined as the ignition delay time. For ignition delay time calculations, a closed homogeneous reactor model was used. The reduced mechanism was compared against the experimental data from J. Huang et Al. [30]. For validation in Chemkin, the pressure was taken as 16 atm. and the temperature varied between 1227 and 1307 K. The fuel composition was CH 4 /C 2 H 6 /O 2 /N 2 (8.93-0.34-19.05-71.68%). Table 5 shows the numerical comparison between the predictions and lab data with an average absolute error of 13%, which is well within the combined temperature/concentration uncertainties (8%) [30] and typical experimental data uncertainties (10-20%) [33]. Figure 5 shows the graphical representation of the experimental and simulation results between ignition delay time and for the said mixture with the new mechanism at the conditions mentioned above.

Adiabatic Temperature
The adiabatic flame temperature is the maximum temperature when a particular gas mixture reaches equilibrium under an adiabatic combustion condition. In a real system, due to heat losses and chemical kinetic/mass transport limitations, the flame temperature is likely to be lower. In this validation, the adiabatic temperature was estimated with the Chemkin equilibrium model. The inlet composition of fuel and air mixtures was set at 0.8 equivalence ratios. The initial temperature and pressure were set at 293 K and 1 atm., respectively, for both fuel and air. An adiabatic flame temperature comparison of the Chemkin simulation and experimental data for ammonia and methane flames at different mole fractions of NH3 is shown in Figure 6 and Table 6 [9]. As seen in Table 6, the predictions were right on target, which indicates the quality of the thermodynamic properties used in the NH3NG mechanism.

Adiabatic Temperature
The adiabatic flame temperature is the maximum temperature when a particular gas mixture reaches equilibrium under an adiabatic combustion condition. In a real system, due to heat losses and chemical kinetic/mass transport limitations, the flame temperature is likely to be lower. In this validation, the adiabatic temperature was estimated with the Chemkin equilibrium model. The inlet composition of fuel and air mixtures was set at 0.8 equivalence ratios. The initial temperature and pressure were set at 293 K and 1 atm., respectively, for both fuel and air. An adiabatic flame temperature comparison of the Chemkin simulation and experimental data for ammonia and methane flames at different mole fractions of NH 3 is shown in Figure 6 and Table 6 [9]. As seen in Table 6, the predictions were right on target, which indicates the quality of the thermodynamic properties used in the NH 3 NG mechanism. Chemkin equilibrium model. The inlet composition of fuel and air mixtures was set at 0.8 equivalence ratios. The initial temperature and pressure were set at 293 K and 1 atm., respectively, for both fuel and air. An adiabatic flame temperature comparison of the Chemkin simulation and experimental data for ammonia and methane flames at different mole fractions of NH3 is shown in Figure 6 and Table 6 [9]. As seen in Table 6, the predictions were right on target, which indicates the quality of the thermodynamic properties used in the NH3NG mechanism. Figure 6. Adiabatic temperature (experimental vs. NH3NG reduced mechanism) [9].

NOx and CO Emissions
In this section, we compared the predicted NOx and CO emissions from NH 3 NG reduced mechanism with the experimental emission data, as well as those from different mechanisms in the literature [9,16,20,21]. These reactions were carried out in an adiabatic reactor with different CH 4 /NH 3 ratios, as shown in Table 6. The NOx and CO measurements and predictions for a thermal input of 1200 W and an equivalence ratio of 0.8 are shown in Figures 7 and 8.

NOx and CO Emissions
In this section, we compared the predicted NOx and CO emissions from NH3NG duced mechanism with the experimental emission data, as well as those from diffe mechanisms in the literature [9,16,20,21]. These reactions were carried out in an adiab reactor with different CH4/NH3 ratios, as shown in Table 6. The NOx and CO meas ments and predictions for a thermal input of 1200 W and an equivalence ratio of 0.8 shown in Figures 7 and 8.   Figure 7. Comparison of NOx emission predictions versus NOx measurements for a thermal i of 1200 W and equivalence ratio of 0.8 [9,16,20,21].
As shown in Figures 7 and 8, all mechanisms, including the present work, overdict NOx and CO emissions in relation to the published data available [9,16,20,21]. O all, the predictions from the NH3NG mechanism were in line with other literature m anisms. Compared to other mechanisms in the literature, the predicted CO emissions f NH3NG were higher (further from the experimental data). In comparison, the predi NOx emissions were lower (closer to the experimental data).

Discussion
Even though the mechanism reduction with Chemkin was conducted at a temperature of 1500 K and pressure of 300 bar for the purpose of developing a reduced mechanism that is useful for high-efficiency turbine and engine applications, the reduced mechanism NH3NG did remarkably well in validation at a wide range of experimental conditions, including those conducted at lower temperatures and pressures (e.g., laminar flame speed data at 423 K and 1-3 atm. [28] and ignition delay data at 16 atm. and 1227-1307 K [36]).
The Moss-Brookes model for soot prediction is applicable to higher hydrocarbon species (ethane and propane) by including appropriate soot precursors and participating surface growth species. In our earlier work of the LU 3.0.1 reduced mechanism for combustion of C1 to C4 light hydrocarbons, three precursor and surface growth species C2H2, C2H4, and C6H6, were involved. In this work, due to the limitation of the number of species and the addition of ammonia fuel, only two soot precursor species, C2H2 and C2H4, were included in NH3NG. However, these two precursor/surface growth species were deemed sufficient to predict soot emission because even if neither is present, curve fitting can be used in Fluent to determine the precursor and surface growth species' mass fractions [37][38][39][40].
The reactor used in Rocha et al.'s 2019 experiments [9] was a porous media flat burner type, while in the Chemkin simulation, the available types were the batch, constant volume, constant pressure reactor, perfectly/partially stirred, and plugged flow reactors. In our predictions, we assumed an equilibrium (adiabatic) perfectly stirred reactor for simplicity. However, NH3NG predictions still largely agreed with the literature values. Another feature that differentiates our work from other literature mechanisms (i.e., those shown in Figures 7 and 8) is that the full mechanism (the Modified Konnov Mechanism, 132 species and 1238 reactions) used in this work was a combination of Konnov's mechanism and the USCII mechanism, while others are in the family of the Konnov mechanism for ammonia/methane combustion. NH3NG allowed the simulation of heavier hydrocarbons, such as ethane and propane, and also had a better capability to predict soot emission with soot precursors C2H2 and C2H4 using the Moss-Brookes' equation in ANSYS Fluent [38,39].
In future work, NH3NG can be used in ANSYS Fluent CFD simulations to further study the combustion process. For example, 3-D CFD simulations for various types of low NOx and 2-stage burners that are suitable for modern gas/supercritical-CO2 turbines and ICE applications can be performed [43,44]. NOx/soot/CO/unburnt fuel/VOC emissions under various equivalence ratios, NH3 to NG, NG composition, fuel flow rate, and air/fuel As shown in Figures 7 and 8, all mechanisms, including the present work, over-predict NOx and CO emissions in relation to the published data available [9,16,20,21]. Overall, the predictions from the NH 3 NG mechanism were in line with other literature mechanisms. Compared to other mechanisms in the literature, the predicted CO emissions from NH 3 NG were higher (further from the experimental data). In comparison, the predicted NOx emissions were lower (closer to the experimental data).

Discussion
Even though the mechanism reduction with Chemkin was conducted at a temperature of 1500 K and pressure of 300 bar for the purpose of developing a reduced mechanism that is useful for high-efficiency turbine and engine applications, the reduced mechanism NH 3 NG did remarkably well in validation at a wide range of experimental conditions, including those conducted at lower temperatures and pressures (e.g., laminar flame speed data at 423 K and 1-3 atm. [28] and ignition delay data at 16 atm. and 1227-1307 K [36]).
The Moss-Brookes model for soot prediction is applicable to higher hydrocarbon species (ethane and propane) by including appropriate soot precursors and participating surface growth species. In our earlier work of the LU 3.0.1 reduced mechanism for combustion of C 1 to C 4 light hydrocarbons, three precursor and surface growth species C 2 H 2 , C 2 H 4 , and C 6 H 6 , were involved. In this work, due to the limitation of the number of species and the addition of ammonia fuel, only two soot precursor species, C 2 H 2 and C 2 H 4 , were included in NH 3 NG. However, these two precursor/surface growth species were deemed sufficient to predict soot emission because even if neither is present, curve fitting can be used in Fluent to determine the precursor and surface growth species' mass fractions [37][38][39][40].
The reactor used in Rocha et al.'s 2019 experiments [9] was a porous media flat burner type, while in the Chemkin simulation, the available types were the batch, constant volume, constant pressure reactor, perfectly/partially stirred, and plugged flow reactors. In our predictions, we assumed an equilibrium (adiabatic) perfectly stirred reactor for simplicity. However, NH 3 NG predictions still largely agreed with the literature values. Another feature that differentiates our work from other literature mechanisms (i.e., those shown in Figures 7 and 8) is that the full mechanism (the Modified Konnov Mechanism, 132 species and 1238 reactions) used in this work was a combination of Konnov's mechanism and the USCII mechanism, while others are in the family of the Konnov mechanism for ammonia/methane combustion. NH 3 NG allowed the simulation of heavier hydrocarbons, such as ethane and propane, and also had a better capability to predict soot emission with soot precursors C 2 H 2 and C 2 H 4 using the Moss-Brookes' equation in ANSYS Fluent [38,39].
In future work, NH 3 NG can be used in ANSYS Fluent CFD simulations to further study the combustion process. For example, 3-D CFD simulations for various types of low NOx and 2-stage burners that are suitable for modern gas/supercritical-CO 2 turbines and ICE applications can be performed [43,44]. NOx/soot/CO/unburnt fuel/VOC emissions under various equivalence ratios, NH 3 to NG, NG composition, fuel flow rate, and air/fuel mixing patterns, can be estimated using ANSYS Fluent. CFD features, such as lean blow-off (LBO), can be used to investigate injection patterns and flame stability [46].
The developed NH 3 NG Chemkin files are available from the corresponding author upon request.

Conclusions
A fuel mixture of ammonia and natural gas provides clean, low-carbon energy to run gas turbines and internal combustion engines for power generation, railroad/truck transportation, marine shipping, and commercial aviation. Reduced mechanisms are required in certain CFD software and can save computer time and storage in others for 3D CFD combustion applications. The newly proposed NH 3 NG reduced mechanism can be employed in CFD works for the prediction of NOx, CO, and soot emissions involving ammonia and natural gas combustion. Soot emission can be predicted by applying the Moss-Brookes-Hall model using soot precursor species.
The prediction errors for NOx and CO emissions associated with the use of the NH 3 NG mechanism for ammonia-natural gas mixtures were also minimal (abs. error 5%) in comparison to the full Modified Konnov mechanism (a combination of the Konnov mechanism and the USCII mechanism). The NH 3 NG reduced mechanism was validated with experimental laminar flame speed, ignition delay, adiabatic temperature, and NOx and CO emissions' data.
The effect of ammonia in such fuel mixtures was also analyzed by comparing the Laminar flame speeds of the reduced mechanism to values found in the literature. An increase in the ammonia concentration in such a fuel mixture decreased the laminar flame speed of the combustion flame. An average absolute error of 7.7% was obtained for the laminar flame speed prediction, which is well within the reported experimental uncertainty of 10% [28]. The R 2 between prediction and data was 0.985. The ignition delay time of the NH 3 NG mechanism also agrees with the experimental data found in literature. An increase in the ammonia concentration in such a fuel mixture increased the ignition delay of the combustion flame. An average ignition delay prediction error of 13% was also within typical experimental data uncertainties (10-20%). The predictions of adiabatic temperatures were within 1 • C. Similar to all mechanisms in the literature, the present work overpredicted NOx and CO emissions in relation to the published data [9,16,20,21]. Overall, the predictions from the NH 3 NG mechanism are in line with other literature mechanisms. Compared to other mechanisms in the literature, the predicted CO emissions from NH 3 NG were higher (further from the experimental data), while the predicted NOx emissions were lower (closer to the experimental data). Funding: This work was supported by the Texas Air Research Center (TARC Grant #110LUB0179A). TARC was not involved in study design, in the collection, analysis, and interpretation of data, in the writing of the report, and in the decision to submit the article for publication.