Reactive Structures of Ammonia MILD Combustion in Diffusion Ignition Processes

Reactive structures have been analyzed, when ammonia is used as a fuel, in a steady 1D counterflow diffusion flame layer, mimicking diffusion ignition processes. The characterization has been carried out in a wide range of feeding parameters under Moderate or Intense Low-oxygen Dilution (MILD) combustion conditions. Both the Hot-Fuel-Diluted-Fuel (HFDF) and Hot-Oxidant-Diluted-Fuel (HODF) configurations were studied to analyze the main effects of the inlet feeding conditions on the oxidative structures. The reaction zone has been analyzed in terms of temperature and heat release profiles in the mixture fraction space for various ranges of inlet parameters, using a standard code and a validated chemical kinetic scheme. Several features of the reaction zone have been recognized as reported also in previous works for hydrocarbon flames. They were used as discriminative for the achievement of various combustion regimes. In particular, the flame thickening process and the absence of correlation between the maximum heat release and the stoichiometric mixture fraction were analyzed to build maps of behaviors. The latter were reported on an inlet preheating level-temperature increase plane for fixed values of the bulk strain rate and system pressures. Another relevant feature previously reported with hydrocarbons in the literature, in Hot Diluted Diffusion Ignition (HDDI) processes under MILD conditions, was the pyrolysis depression. The latter characteristic has been not observed when ammonia is used as a fuel, for the operative conditions here investigated. Indeed, the heat release profiles do not show the presence of negative heat release regions. The results obtained for the HFDF configuration are strongly dependent on the system pressure level. Finally, the HODF condition has been also analyzed for ammonia at the atmospheric pressure. Boundaries of the combustion regimes and reactive structure features showed several differences between HFDF and HODF cases with respect to the inlet parameters.


INTRODUCTION
In the context of new technological solutions able of simultaneously meeting thermal efficiency needs and pollutant emission restrictions, while responding to the flexibility with respect to operating conditions and fuel properties, Moderate or Intense Low-oxygen Dilution (MILD) Combustion represents a promising solution, able to accommodate the complexity and variability of energy mix, by avoiding drastic changes in the configuration of conventional plants (Cavaliere et al., 2008).
A paradigm shift is realized in the context of such combustion systems: the process is studied as a chemical reactor, that needs to be optimized for efficiency, eco-sustainability and fuel flexibility. Oxidation process relies on local ignition phenomena due to the presence of hot gases/fresh reactants mixtures that realizes low oxygen levels (Perpignan et al., 2018). Moreover, high dilution levels prevent the stabilization of diffusive and/or deflagrative structures in an environment where self-ignition occurs. As a consequence, the reactive structure strongly differs from conventional combustion cases, showing homogeneity of temperatures and scalars inside the reactor volume, the absence of a visible flame front and reduced pollutant emissions (such as CO, nitrogen oxides and soot) (Weber et al., 2005). The oxidative regions, under highly diluted and preheated conditions, attain a distributed ignition regime with small gradients of scalar quantities (Minamoto et al., 2014). It gives the opportunity to use a broad range of fuels, being a very good candidate for low-calorific-value fuels (Huang et al., 2014), hydrogen-based and industrial wastes (Derudi and Rota, 2019), as well as for liquid (Ye et al., 2015) and solid fuels (Saha et al., 2017).
Recently, few works showed also the possibility of using carbon free fuels such as ammonia or ammonia/hydrogen blends in advanced combustion technologies with promising results (Li et al., 2019;Sorrentino et al., 2019;Valera-Medina et al., 2019;Ariemma et al., 2020).
In addition, the elementary processes that are attained in MILD conditions differ from those involved in conventional systems, due to strong overlapping between mixing and chemistry effects (Galletti et al., 2007). Flue gas entrainment induces high local diluent levels so that chemical kinetics become slow enough to operate on time scales comparable to mixing ones (Özdemir and Peters, 2001).
Reactive structures obtained under highly diluted and/or preheated reactants conditions were rigorously analyzed in the literature for laminar diffusive layers in 1D configurations (de Joannon et al., 2009(de Joannon et al., , 2012 and under the occurrence of multiple mixing layers (Sorrentino et al., 2017). Part of this interest is due to the relevance of these processes in pre-chamber jet ignition (Sidey and Mastorakos, 2018) as well as in inter/intra gas turbine combustion chamber (Nemitallah et al., 2018), but mostly in furnaces, gasification (Kwiatkowski et al., 2013) and incineration applications where low-heating value fuels (Song et al., 2013) are burned in a hot products environment.
The analysis of the combustion regimes that are realized when a fuel flow is impinged toward an oxidant stream in diffusion ignition processes, has to consider all the possible boundary conditions to include MILD combustion feeding parameters. In particular, in the Hot Diluted Diffusion Ignition (HDDI) process, both the fuel and oxidant streams are diluted and pre-heated at different levels, resulting in several configurations which may lead to diverse reactive structures. Several experiments confirmed the establishment of peculiar stabilization regimes in 2D systems, mimicking hot and diluted reactive conditions (Oldenhof et al., 2010;Ye et al., 2018). In such cases the oxidation process is sustained through multiple ignition kernels that are transported by the mean flow.
Diffusion-controlled reactive structures were also characterized in laminar coflow diffusion flames, in order to investigate the salient details of non-premixed MILD combustion systems and to provide insights into the chemical structure of such regime. Comprehensive experimental and computational studies of the heat release structures and their impact on laminar flames have been carried out by different groups to characterize the main features of the oxidative regions Sepman et al., 2013;Najafi et al., 2021).
In such a framework, a systematic classification of the phenomenologies that are attained under different boundary conditions was rigorously reported in the literature (de Joannon et al., 2009;de Joannon et al., 2012).
Specifically, the Hot-Fuel-Diluted-Fuel (HFDF) feeding case, mimicked local conditions that can occur in several applications such as the coal combustion, low heating value fuel oxidation derived from processes or systems with internal flue gas recirculation systems (Kwiatkowski and Mastorakos, 2016).
Another interesting case is the Hot-Oxidant-Diluted-Fuel (HODF) configuration, that is relevant in gas turbine applications with low heating value fuels (Maruta et al., 2000) and incineration processes for the destruction of volatile organic compounds (VOC).
Such configurations were previously analyzed for low molecular weight paraffins (de Joannon et al., 2009), as well as biogas (Chen and Zheng, 2011) and hydrogen (Chen et al., 2012).
The main results of such literature works were that, in MILD combustion conditions, the reactive structures are profoundly different with respect to the ones observed in conventional diffusion flames. In particular, the heat release profiles are widened in the mixture fraction space with the simultaneous disappearance of the pyrolytic region. At the same time, the location of the oxidation region in the Z-space is not correlated to the stoichiometric mixture fraction (Z st ). The latter behavior is more evident in the HODF case where the reactive structure is moved toward the hot side, whereas the Z st follows an opposite trend. Such oxidative structures supported the occurrence of a Diffusion-Ignition process (Sorrentino et al., 2020).
The peculiar microstructures that are realized in MILD reactive processes, are reported in the literature as "ignidiffusive structures" because they are characterized by a distributed diffusion-controlled local auto-ignition (Sorrentino et al., 2020). They revealed unconventional features of the oxidation region when high dilution and preheating levels of the reactants are established. This structure corresponds to the apparent macroscopic uniformity of the combustion and gives rise to a highly stable reactive process.
The above literature survey, concerning the categorization of the reactive structures that were obtained in diffusion ignition processes, are limited in discussing the performance of traditional fossil fuels under these novel combustion conditions. Up to date the corresponding characterization of such oxidative layers with sustainable energy carriers, such as ammonia or their blends, in innovative combustion regimes is still quite sparse. To the authors' best knowledge, until now very few papers dealt with this issue (Ku et al., 2018;Colson et al., 2020), despite they were focused more on the extinction characteristics of ammonia flames and their blends.
Thus, the previous arguments clearly showed that the characterization of the reactive ignidiffusive structures, in highly diluted and/or preheated conditions, provides key information on the sustainability and potentiality of MILD Combustion with new fuels for several technological applications.
Although it is a promising way to utilize ammonia more in novel combustion technologies, the necessary knowledge concerning the oxidation regimes that can be reached in the utilization of such new energy vector, is highly desired before it could be widely implemented with confidence in its performance (Valera-Medina et al., 2021).
In this framework, the present work aims to extend the categorization of the reactive structures in the 1D diffusive configuration, by investigating two feeding cases corresponding to the HFDF and HODF configurations when ammonia is used as a fuel.
Such elementary process was chosen to provide a systematic classification of the phenomenologies that are attained with ammonia under MILD conditions and that were rigorously recognized in the literatures by several combustion groups with fossil fuels. Several emerging features involving dilution and preheating effects on the reactive structures can be recognized only when elementary configurations are adopted, to decouple the chemistry effects from the fluid-dynamics ones.
The investigation was performed numerically by following a similar approach to the ones that were carried out in previous papers by the same group on HODF (de Joannon et al., 2009) and HFDF (de Joannon et al., 2012) configurations. Major emphasis has been placed on the HFDF case that resembles the majority of the technological applications. For this configuration, sensitivity analysis has been carried out to show the influence of the pressure on the system reactivity.
In particular, the methodology deals with detailed numerical simulations in 1D diffusion layers where the thermochemical patterns were obtained in a large set of feeding parameters. The main combustion regimes and their boundaries are reported and the most relevant differences between the oxidative regions are investigated as a function of the process parameters. Thus, the obtained results are compared between the two different configurations and the discussion on their implications extends the conceptual framework that was previously outlined.
Diffusion ignition elementary processes mimic in a proper way the local conditions that occur in MILD burners where internal recirculation of burned products influences both on the local and global scales the ignition/oxidation patterns.
The main originality of the present work is the investigation and characterization of the reactive structures of ammonia MILD in the diffusion ignition process. It provides a systematic classification of the phenomenologies that are attained with ammonia under MILD conditions and that were rigorously recognized in the literatures by several research groups with carbon fuels.
In particular, the influences of: 1) the preheated temperature of the fuel or oxidizer flows, 2) the diluent concentration in the fuel flow, 3) the system pressure, on the oxidative structure features of ammonia under peculiar combustion conditions are investigated with the aid of numerical simulations.
It is found that the sustainability of MILD combustion, fueled with ammonia under highly diluted conditions, is highly dependent on the preheating level of the reactants and system pressure.

CONFIGURATIONS AND NUMERICAL TOOLS
The reactive structures that are attained in a steady 1D diffusion ignition process were analyzed through the opposed counterflowing jets configuration. A sketch of the system is reported in Figure 1 where it represents both the HFDF and the HODF.
It is worth to note that such canonical configuration aims to mimic the reactive structures that can be obtained in a real system, both on the macro and micro scales.
Specifically, a diluted ammonia-nitrogen stream (NH 3 /N 2 ), with an inlet temperature (T 0 ), fuel molar fraction (X f ) and velocity (V 0 ) counter-diffuses toward an air jet with a certain preheating level (T in ). In the HFDF configuration the fuel flow is preheated at a fixed level whereas T in 300 K. On the opposite, in the HODF case T 0 300 K whereas T in is higher than the ambient temperature. The momentum of both the streams was fixed in such a way that their kinetic energies are equals. In this manner the stagnation point is always located at the center of the mixing layer for frozen conditions, in each case here considered (Mastorakos et al., 1995). The study was carried out by means of numerical computations with OPPDIFF module of Chemkin-Pro package (CHEMKIN-PRO 15131, 2013).
Thus the boundary conditions here used were related to the inlet jets velocity (V 0 , V in ), feeding temperatures (T 0 , T in ) and streams mole fraction (X f for the fuel jet and air composition for the oxidizer jet). System pressure (P) was also fixed for each computation at P 1 Atm or 20 Atm to evaluate the effect of such parameter.
1D counterflow flame model that solves the governing equations based on a similarity solution, while neglecting buoyancy in the momentum equation, were solved in the onedimensional physical space (Kee et al., 1989) with thermal diffusion, multi-component diffusion with species Lewis number modelled according to the transport database, and a constant imposed rate of strain, K 0 . The mixture-averaged transport model was also compared with the multi-component one to verify that differences in both the thermal and chemical reactive structures were negligible.
The steady time step size was chosen to be 10 -6 s while the relative and absolute tolerances were 10 -5 and 10 -9 , respectively.
Computations were tested to assure sufficient spatial resolution at the reaction zone and that solutions were independent on the grid size. The non-uniformly spaced grids of up to 250 points were found to be sufficient to resolve the reaction zone.
The bulk strain rate of the fuel flow (K 0 ) is equal to V 0 /D. Several reaction mechanisms reported in the literature for ammonia combustion were evaluated. In particular, Manna et al. (2020) reported an experimental characterization of NH 3 oxidation and pyrolysis processes in diluted conditions in two model reactors (JSFR and LFR). The effect of several diluents, such as argon or water, were evaluated. Numerical simulations were performed with different kinetic mechanisms available in literature to value their capability to describe ammonia chemistry.
In particular the authors compared the numerical results obtained with the "Konnov" (Konnov and Ruyck, 2000), "Glarborg" (Glarborg et al., 2018), "Song" (Song et al., 2016) and "Nakamura" (Nakamura et al., 2017) mechanisms. Among them the Nakamura kinetics model (Nakamura et al., 2017) was reported as the one with the best performance that predicted in a consistent way the experimental trends. Such findings were also confirmed by another work of the same research group for different dilution levels and diluents nature . The Nakamura mechanisms is a detailed kinetics (Nakamura et al., 2017) that was validated for ammonia/air on the basis of experimental tests in a micro flow reactor with controlled temperature profiles, at atmospheric pressure, at several equivalence ratios (Φ 0.8, 1.0, 1.2) and temperatures lower than 1,400 K. The mechanism consists of 38 species and 232 reactions." The characterization of the oxidative structures has been carried out in the manuscript by analyzing the temperature (T) and heat release rate ( _ h) profiles as a function of the mixture fraction (Z).
For conventional feeding conditions of both the air and ammonia jets (T in T 0 300 K with no dilution), temperatures and heat release profiles were reported in Figure 1 with black and grey lines, respectively. In particular, when both the streams are undiluted, the system temperature (solid black line) quickly passes from 300 K (at Z 0) up to T max 1,930 K around the stoichiometric value of Z Z st 0.135. Subsequently, the temperature has a smooth reduction as the mixture fraction increases, attaining T 0 300 K at Z 1.
The corresponding heat release trend (grey curve) sharply increases up to _ h max that is located nearby both the T max and Z st values.
The _ h after reaching its maximum value then gradually decreases down to zero. Such profile shows the prevalence of exothermic reactions that mainly lead to the formation of intermediate or to products of oxidation.
It is very important to note that the heat release profile with pure ammonia shows only positive values, denoting for such conditions the absence of the pyrolysis region that is usually recognized with hydrocarbon flames in the richest negative heat release regions . Such behavior is mainly related to the non-carbon nature of the fuel and the predominance of exothermic reactions that overcome the endothermic ones .
Detailed numerical computations were performed over a wide range of parameters such as T 0 , T in , dilution of the fuel jet (X f ) and system pressure. Results were reported in the following sections for both the HFDF and HODF cases.

RESULTS AND DISCUSSION
Reactive structures in one-dimensional diffusive layer are strongly influenced by dilution and/or preheating levels of both the fuel and oxidant flows. As explained in the previous sections, two configurations were considered in this work, on the basis of the inlet streams boundary conditions. In particular, the HFDF condition referred to the case where the fuel inlet flow is diluted and preheated at fixed levels. Moreover, the effect of system pressure variation on the oxidative structures for the pure ammonia cases was also reported.
Afterwards the HODF condition was investigated at ambient pressure and for the pure ammonia case to show the main differences with respect to the HFDF case.

Hot Fuel Diluted Fuel
The reactive structures that were obtained in the HFDF case have been reported in Figure 2 where temperature (upper part) and heat release (lower part), numerically evaluated for a fuel preheating level of 1,800 K, K 0 50 s −1 and p 1 Atm, were showed as a function of the mixture fraction on diagrams parametric in X f , which ranges from 1 to 0.01. Solid lines represent both T and _ h variables and their colors change as a function of the fuel dilution level (X f ). Frozen line is also reported in the temperatures diagram with a dashed line. Solid circles in the heat release profiles indicate the stoichiometric mixture fraction (Z st ) values for each X f level.
The bulk strain rate value here used (K 0 50 s −1 ) was chosen in accordance with the previous studies on diffusion ignition processes with hydrocarbons from the same research group (de Joannon et al., 2009;de Joannon et al., 2012), in order to compare the results here obtained with the methane ones.
At X f 1, i.e., for no fuel dilution cases, the T profile is very close to the one showed for conventional conditions in Figure 1. In particular, due to the fuel preheating level, T values reaches a higher maximum value of 2,400 K for Z 0.153, then decreases due to thermal diffusion. Even the heat release profile at X f 1 is very similar to the one obtained for the standard case. It increases up to a first relative maximum at Z 0.089, then it attains a second absolute maximum which occurs at Z 0.156, slightly higher than Z st , that cause the enlargement of the oxidative region toward higher Z values. Afterwards the heat release decreases down to zero, thus no pyrolysis region was observed due to the absence of negative _ h value zones. Such behavior is mainly ascribable to the non-carbon nature of the fuel, in contrast to the prevalence of endothermic zones in some part of the heat release profile that were observed for hydrocarbons in previous works (de Joannon et al., 2012;Sorrentino et al., 2020).
Thus, oxidative structure stabilization occurs in a narrow mixture fraction region (between Z 0.08 and Z 0.16) that includes the two maxima where the most part of the conversion occurs.
Fuel content decrease amplifies the differences with respect to standard conditions. For X f 0.5, the maximum temperature values occurs at higher mixture fractions (Z 0.32), which is nearby to Z Xf 0.5 st 0.305. Subsequently the temperature decrease is somewhat smoother than that observed without dilution. In particular, the maximum temperature increase is about 1,600 K, lower than 1,900 K, reached in the undiluted case. As a consequence, the oxidative region in the related heat release profile is shifted and widened. When compared to the undiluted case, even in this case, two positive maxima were observed with a less pronounced shoulder at Z 0.2. The latest behavior is further emphasized when the dilution is increased. In fact, as reported in Figure 2B, the first maximum in the heat release profile disappears when X f is decreased from 0.3 to 0.15 (purple line).
The alteration of the reactive structure becomes more complicated when the fuel dilution is further increased. The oxidative region is progressively shifted toward higher mixture fraction locations, following the shift of Z st with the X f decreasing. Specifically, two main effects of the dilution increase can be appreciated. The first one is the evolution of the left shoulder in correspondence of the first heat release maximum. The first peak results to be embedded in the remaining structure for X f values around 0.2. At the same time, a substantial widening of the oxidation region in the Z-space occurs, by outlining a modification of the reactive structure. The dilution increasing allows to shift the region of the space where ignition occurs towards higher Z locations identified by higher frozen temperatures. _ h profiles maximum intensity is strongly decreased when passing from X f 0.3 to X f 0.01, in contrast to what happens for lower dilution levels (X f > 0.3).
The preheating level of T 0 1,800 K was chosen as exemplificative of the different reactive structures that are achievable when ammonia content in the NH 3 /N 2 mixture is decreased down to X f 0.01. In such a manner it is possible to analyze the oxidative structure also for very high dilution levels.
Indeed, in the case of lower preheating levels (T 0 1,400 and T 0 1,600 K) such oxidative structures are not sustained for highly diluted conditions (X f < 0.1), thus extinction occurs.
Temperature and heat release profiles were reported in Figure 3 for lower preheating levels (T 0 1,400 and T 0 1,600 K) to show the sensitivity of the diffusion ignition process with respect to T 0 .
As previously said, extinction process occurs at T 0 1,600 K for N 2 dilution levels higher than 90% whereas such critical value decreases to 70% when T 0 1,400 K.
Such kind of behaviors were recognized on the basis of both the temperature and heat release rate profiles with T values very close to the frozen line and _ h values that are approximately zero. On the basis of the heat release rate (HRR) profiles, such as the one reported in Figure 2B or Figures 3A,B,D global categorization of the combustion regimes occurring under the different feeding conditions was performed, by carrying out numerical simulations in the T 0 range between 800 and 2,000 K in the whole interval of ammonia dilution (from X f 1 to X f 0.01). Such kind of analysis pursued an approach that was already followed in previous works of the same group for hydrocarbon fuels, related to HFDF and HODF feeding operative conditions (de Joannon et al., 2009). Consequently, a map of behavior for the system reactivity was achieved as a function of the inlet parameters, i.e., T 0 and X f (fuel dilution level). The latter one is reported on the map by taking into account the maximum temperature increment ΔT for the inlet feeding conditions. Thus the map reported in Figure 4 was obtained in the T 0 -ΔT plane for T in 300 K, K 0 50 s −1 and p 1 Atm. The fuel preheating temperature ranged from 800 to 2,000 K, while ΔT ranged from 600 to 1,800 K.
The map showed in Figure 4 reported mainly four combustion regimes that are in some aspects similar to the ones obtained for methane in previous papers (de Joannon et al., 2009;de Joannon et al., 2012).
The boundaries between each zone have been obtained by investigating the shape or magnitude of heat release and/or T curves for each T 0 value. The lowest T 0 and ΔT values were obtained in the lower-left region of the map. Such zone was classified with the "No Combustion" condition, where the heat released is zero and T profiles are coincident with the frozen one in the entire range of Z values. For ΔT in the range between 850 and 1,600 K no combustion conditions were obtained up to T 0 T ign , that is the characteristic ignition temperature of ammonia at p 1 Atm. Such regime without heat released widens up to T 0 1,800 K for ΔT lower than 850 K. In the T 0 interval between 800 K and T ign a temperature increment (ΔT) higher than 1,600 K is able to sustain a "feedback" combustion regime, where the heat release is supported by species and heat obtained from the reaction front, such as the one detected for conventional systems. In such cases, the flame stabilization process is the one typical of diffusion flames. The region where oxidation occurs is centered close to the Z st value in the heat release profile and it presents two maxima. It is worth to underline that, in contrast to hydrocarbon flames, ammonia diffusion ignition does not present pyrolytic conditions with negative heat release regions (Xiao et al., 2020).
For T 0 higher than the T ign value, the feedback combustion regime gradually reaches the high temperature combustion one. In such region, the reactive region structure is broadened on a wider mixture fraction interval but they exhibit very similar features to the ones obtained in the feedback regime.
By decreasing the ΔT from the High Temperature Combustion region (i.e., by increasing the fuel dilution), a further broadening of the oxidative structure was observed and the MILD Combustion region is reached, where a single maximum is detected in the heat release profiles and the maximum temperature increment in the mixture fraction space is lower than T ign . Another notable feature of the _ h profiles is associated with the position of Z hmax (location of the maximum _ h value in the Z space) with respect to Z st . In particular, in the MILD regime such two values do not coincide, as shown for example by heat release profiles of Figure 2B for X f < 0.15.
In this regard, the identification of MILD diffusion ignition conditions for ammonia combustion on the basis of temperature and heat release profiles appears to be more challenging with respect to hydrocarbons flames. In fact, in the latter cases MILD regimes for methane were characterized in previous works by three outstanding characteristics (de Joannon et al., 2012): -Pyrolysis depression -Flame thickening -No correlation of the stoichiometric and maximum heat release conditions They are notable signatures of the occurrence of MILD in HDDI processes (Sorrentino et al., 2020).
In this context, it is worth to note that the first characteristics (Pyrolysis depression) cannot be used for ammonia as a feature to identify the transition between conventional and MILD regimes. In fact, as stated above, the heat release profile with pure ammonia shows only positive values, denoting for such conditions the absence of the pyrolysis region that is usually recognized with hydrocarbon flames in the richest negative heat release regions. Such behavior is mainly related to the non-carbon nature of the fuel and is also strongly dependent on the bulk strain rate value.
Thus the second and the third characteristics have to be used as features that mark the transition to MILD conditions, as confirmed also by means of several DNS studies (Doan et al., Another key point that needs to be remarked is that the third characteristic, i.e., the absence of correlation between Z st and the maximum values of heat release, is obtained only in a limited part of the MILD domain, thereby indicating that the HDDI process is obtained in a subset of the MILD region. The flame thickening process, related to the widening of the heat release profiles for high dilution levels (X f < 0.2) marked the gradual transition of the reactive structure toward a distributed oxidation process where the "front-like" feature (Swaminathan, 2019) of the reactive region is not recognized (Ye et al., 2017).
The effect of strain rate on the reactive structures and heat release (HRR) profiles in MILD and conventional flames has been discussed in some literature works for hydrocarbon fuels . Oxidative structures are sustained under MILD conditions also at very high strain rates and, with extensive dilution and preheating, do not exhibit typical extinction behaviors (Mastorakos et al., 1995). The dissolution of an extinction strain rate in MILD conditions was already observed in previous findings for methane/air cases when reporting the heat release rate as a function of the strain rate .
Such behaviors were confirmed when ammonia is used as a fuel in the present study. In particular, numerical simulations were performed for the HFDF conditions of Figure 2, by changing the bulk strain rate value (K 0 ) and evaluating the maximum value of the heat release rate in the mixture fraction space (heat release max).
Simulations were reported in Figure 5 for T f 1,800 K and three X f values. X f 1 and X f 0.3 are related to cases in the high temperature combustion zone of the map in Figure 4, with a reactive region structure that is very close to a conventional combustion case. On the other hand, the case at X f 0.05 is related to a MILD condition, as reported previously.
The peak value of HRR for each case at varying bulk strain rates is shown in Figure 5 as a function of K 0 . Remarkable differences in the extinction behaviors between the cases at X f 1, X f 0.3 and the MILD one (X f 0.05) are evident.
The reactive structures at fuel mole fraction of 1 or 0.3 cannot stabilize at strain rates above 800 s −1 whereas the MILD one, as seen in Figure 5, depicted a continuous transition from high to low values of the max HRR and do not show evidence of abrupt extinction behaviors.
MILD conditions showed considerable heat released also for strain rates higher than the conventional ones and they do not extinguish up to 3,000 s −1 . Such features confirmed a strong extinction resilience of MILD regimes.
At very low strain rate values (lower than 25 s −1 ), the negative heat release region appears for fuel preheating levels higher than 1,500 K, suggesting a transition from MILD to conventional combustion behavior, due to the absence of the pyrolysis depression feature as reported above.
Particularly, the minimum heat release rate becomes zero at a strain rate of 10 s −1 for T f 1,500 K whereas it is 18 s −1 at T f 1,600 K and 23 s −1 for T f 1,800 K. Those results of the heat release profiles at very low strain rate values are not presented here for brevity.
The modifications of primary kinetic pathways between the different regimes reported above in Figure 4 were investigated by evaluating the temperature sensitivity to reaction rates as a function of the distance for two X f values (1 and 0.01) and T f 1,800 K. In particular in Figure 6 the temperature sensitivity coefficients are reported for the undiluted condition (X f 1) and a highly diluted case (X f 0.01). The latter condition is related to a MILD combustion case whereas the former pertains to a conventional combustion regime.
Results of temperature sensitivity in Figure 6A quantitatively demonstrate how sensitive temperature is to the reaction rate of the reaction H+O 2 O+OH (branching reaction) that has the largest positive sensitivity coefficient, indicating the largest effect on increasing the temperature of the system. Reaction N 2 H 2 +N NNH+H 2 has the largest negative sensitivity coefficient and affects the temperature decrease the most. The temperature sensitivity analysis reflects strong differences between the X f 0.01 and the undiluted case (X f 1). As shown in Figure 6B, reaction H+O 2 O+OH no longer has the largest positive impact on temperature. In the conventional combustion case, temperature is mostly influenced by fuelspecific reactions such as the following ones: NH 3 +M NH 2 +H+M and N 2 H 2 NNH+H. Such results suggest that the chemical pathways of ammonia in diffusion ignition processes show different features in the various cases and provide some motivations behind the different reactive structures that were observed when passing from undiluted to highly diluted ammonia streams.

Effect of the System Pressure on Pure Ammonia
The reactive structure characterization for pure ammonia, previously obtained at P 1 Atm, was carried out numerically also for a high pressure condition (P 20 Atm) to show the pressure effect on the system reactivity.
In particular, as reported in Figure 7, Temperature and heat release profiles retain their main characteristics, even if slight differences are achieved when compared to results previously reported in Figure 2 for T 0 1,800 K and K 0 50 s −1 . At 20 Atm, temperature profiles exhibit similar trends to the ones obtained at 1 Atm by showing slightly higher T values. The main differences are obtained in the heat release profiles, showing always a single maximum when the system is operated at 20 Atm. This feature can be also recognized at P 10 Atm but such results were not reported here for brevity.
Another relevant characteristics of the reactive structure can be revealed from the heat release intensity which is almost two order of magnitude higher for P 20 Atm with respect to the atmospheric cases.
The signatures of MILD Combustion previously identified, i.e., flame thickening and absence of correlation between Z st and heat release, are also in this case attained for X f < 0.15 when dilution levels are very high.
As showed for the atmospheric pressure case, no correlation between the stoichiometric mixture fraction and Z hmax was achieved only in a narrow subset of the MILD region.
Slight differences at higher pressures were also observed with regard to reactive structures broadening. In particular, at P 20 Atm the heat release profiles exhibit slightly thinner oxidation structures than the ones at atmospheric conditions. This kind of behaviour is mainly related to the shortening of the chemical time scale at higher pressure levels with higher Damköhler numbers for fixed dilution levels, also in MILD conditions. Reverse-flow configurations that were investigated in the literature  reported similar behaviors of the Damkohler number at higher pressures in MILD conditions. In this case the diffusion ignition process is realized locally in the burner, due to the dependence the oxidant mole fractions on both the feedings conditions and recirculation levels.
Such results that were obtained with ammonia in a simple 1D configuration can provide useful insights on the application of MILD Combustion to higher pressures (Lückerath et al., 2008)  with sustainable energy carriers in gas turbine systems (Ye et al., 2015). On the basis of the temperature and heat release profiles that were obtained with ammonia for T 0 that ranged from 800 to 2,000 K, a map of combustion regimes for P 20 Atm is shown in Figure 8. It distinctly reports that the ranges of feeding conditions that correspond to the various oxidation regions is modified to some extent while maintaining the features of the four main regimes.
By comparing the map at P 20 Atm of Figure 8 with the case at ambient pressure, showed in Figure 4, some differences can be pointed out concerning the extension of the combustion regimes. Specifically, an enlargement of both the feedback and high temperature combustion zones at higher pressures is obtained in the T 0 -ΔT plane. Such behavior is mainly ascribable to both the lowering of the ignition temperature and increasing of the Damköhler number at high pressures (Khalil et al., 2019). On the other hand, both the No Combustion and MILD regimes shrink in the ΔT (i.e., fuel dilution) space.

Hot Oxidant Diluted Fuel
MILD combustion features show relevant modification in the reactive structures, with respect to the HFDF case, when diffusion ignition occurs in a mixing layer where the diluted fuel diffuses into the non-diluted hot oxidizer. Results in this section have been reported for the HODF configuration when ammonia is used as fuel, at atmospheric system pressure. Reactive structures in the HODF case have been obtained to point out the main combustion features and their differences with respect to the HFDF ones.
In the HODF configuration both the air pre-heating and the fuel dilution levels are key parameters in altering the oxidative structures as a function of their magnitude. In such a case, system temperatures (upper part) and heat release rates (lower part) were showed in Figure 9. In accordance with the HFDF results reported in Hot Fuel Diluted Fuel section, they were computed for an oxidant pre-heating temperature (T in ) of 1,800 K, V 0 100 cm/s and p 1 Atm. T and _ h profiles were reported as a function of the mixture fraction and each curve is related to a fixed value of X f , which ranges from 1 to 0.05. Both the temperature and heat release profiles are illustrated with solid lines with different colors as a function of X f . On the same figure, frozen temperature is reported with a dashed line whereas solid circles represent the Z st location pertaining to a specific X f . When no dilution is taken into account (X f 1) the temperature profile that was obtained exhibits a similar aspect to the one attained under conventional conditions in Figure 1, albeit different magnitudes for temperatures and _ h values were observed. In particular, due to the preheating, T attains a maximum value of 2,620 K for Z 0.16, then steeply decreases toward the frozen line.
The heat release profile at X f 1 has an abrupt increase up to first relative maximum at Z 0.08, then it increases in the direction of a second absolute maximum which occurs at Z 0.16, marginally higher than Z Xf 1 st 0.141. Such circumstances imply a slight enlargement of the oxidation zone toward higher Z with a subsequent decrease of the _ h profile down to zero. Also in this case, as noted in the HFDF case, no pyrolysis behaviors were observed in the heat release profile, in contrast with the presence of pyrolytic zones obtained with hydrocarbon fuels.
The dilution of the fuel flow emphasizes the differences with respect to conventional conditions. For X f 0.5 the T max values are moved toward higher mixture fraction (Z 0.308), nearby the Z Xf 0.5 st 0.305. The consequent temperature decrease is very smooth toward the frozen line when compared to what happens with no dilution. As a result, the reactive structure associated with the _ h profile is moved and broadens. It again shows two positive maxima with the first shoulder that has a very small maximum intensity at Z 0.173.
The dilution increasing points out complex behaviors related to the oxidative structure evolution. In particular, at X f 0.3 (purple curve), the first heat release peak disappears and is embedded in the remaining structure whereas the second one broadens and is shifted toward higher Z values.
T max values follows a similar behavior of those related to the heat release, revealing the maximum temperatures increment at higher mixture fractions. Another relevant effect related to the dilution increase is the location of the maximum heat release intensity with respect to Z st . Specifically, for X f 0.3, Z hmax is lower than Z st (Z 0.43).
Fuel dilution levels higher than 70% (i.e., X f than 0.3) lead to a different behavior of the oxidative structure that moves in the opposite direction with respect to the Z st increase with the dilution.
In particular, Z h max 0.398 at X f 0.4, whereas Z Xf 0.2 st 0.5. At the same time the heat release profile broadens and widens and its maximum intensity is moved toward the hot side (preheated air). For higher dilution levels (X f 0.15), the stoichiometric mixture fraction is located outside the oxidation region, as clearly showed by the _ h profiles for X f 0.15 (black curve). The absence of correlation between the reactive structure location and the stoichiometric mixture fraction is further emphasized with the dilution level increasing, i.e., X f < 0.15, showing a profoundly difference in the oxidative structure when compared to diffusive combustion. The latter feature, together with the broadening of the oxidation layer entail the occurrence of MILD Combustion conditions. On the ground of the previous analysis, it is worth to note that the different characteristics of both the temperatures and heat release rate profiles are related to diverse oxidation regimes that occur as a function of the feeding parameters. They are very useful in obtaining an exhaustive categorization of the ammonia diffusion ignition process, through a map of behavior on a T in -ΔT plane, where the ΔT values are directly related to the fuel dilution levels. In accordance with the previous results reported in Hot Fuel Diluted Fuel section for the HFDF, the map was obtained for K 0 50 s −1 and p 1 Atm. Fuel is fed at ambient temperature whereas T in was varied from 800 to 2,000 K. Ignition temperature is also reported with a dashed line (T ign ).
The combustion regimes map of Figure 10 identifies four main zones. The lower left part describes the "no combustion" region, where the temperatures and residence times of the inlet two jets do not permit the achievement of a stable oxidation regime. For very low dilution levels (X f > 0.5) and T in < 1,400 K, the "feedback combustion" zone describe conditions where conventional diffusion flames are stabilized, i.e., the maximum reactivity is located nearby the stoichiometric conditions. Furthermore, the upper right part of the map bounded the "high temperature combustion" zone, where very high T max values are obtained because of the air pre-heating. In such a case the reactive structures are very similar to the one obtained under the feedback combustion mode. In both the regions only the oxidative region is identified in the _ h profiles whereas no FIGURE 9 | Temperature (A) and heat release (B) profiles for pure ammonia, parametric in X f , in the HODF configuration for T in 1,800 K, K 0 50 s −1 , P 1 Atm.
Frontiers in Energy Research | www.frontiersin.org October 2021 | Volume 9 | Article 649141 pyrolytic zones are detected. In such cases the heat release shows two maxima in the oxidative region. Finally, the "MILD" region occurs for T in > 1,500 K and X f values lower than 0.4 (ΔT < 1,450 K) and it delineated conditions for which distributed features of the oxidation layer are recognized. Specifically, in such regime a broadening process of the heat release profile is visible in the Z-space and the maximum heat release rate occurs at Z values very low with respect to the stoichiometric ones, in contrast to what happens in both the feedback and high temperature regions.
As a final remark it is worth to underline the main differences between HFDF and HODF configurations for ammonia combustion at P 1 Atm.
In both the cases the feedback combustion region is quite narrow, including inlet temperatures lower than 1,400 K. Similar behaviors were obtained also for the no combustion zone that is slightly wider for the HFDF case. Conversely, the high temperature combustion region is wider for the HFDF case than that under Hot-Oxidant-Diluted-Fuel conditions. Indeed, in the latter configuration, the MILD region is broadened for higher ΔT values when compared with the HFDF one. This means that MILD conditions are stabilized for lower fuel dilution levels when the oxidant flow is preheated with respect to the fuel preheating configuration.

CONCLUSION
The present study reported a categorization of the steady oxidative structures that occurred when ammonia is used as a fuel in diffusion ignition processes. The methodological approach here used was similar to that followed in previous works with hydrocarbon fuels.
The article provided a systematic classification of the phenomenologies that are attained with ammonia under MILD conditions and that were rigorously recognized in the literatures by several research groups with carbon fuels. Up to date the characterization of oxidative layers with sustainable energy carriers, such as ammonia or their blends, in MILD combustion regimes is still quite sparse. To the authors' best knowledge, until now very few papers dealt with this issue.
As a general consideration, it is important to note that the ammonia diffusion ignition process, in both the HFDF and HODF configurations, showed similar results and extend the ones obtained with hydrocarbons in previous literature studies, thus implying that the feeding parameters are reasonable choices for the obtainment of MILD combustion.
Moreover, it showed peculiar features that diversify the occurrence of MILD Combustion with ammonia with respect to fossil fuels.
Maps of behavior were obtained on the basis of the oxidative structure peculiarities that were recognized from temperature and heat release profiles. Such regime maps corresponds to local conditions that may be established in MILD combustion chambers. Specifically, inlet temperature of fuel or oxidizer flows and maximum temperature increment (i.e., the inlet dilution level of the fuel) were used as main feeding parameters to classify the combustion regimes and identify the inlet reactants conditions. Four main regions were identified in the T 0 -ΔT plane as a function of both the temperature and heat release profile features.
Relevant differences in the reactive structures were obtained by comparing the HFDF results for pure ammonia at higher pressure with those achieved at atmospheric pressure.
In particular, at P 20 Atm an enlargement of both the feedback and high temperature combustion zones was revealed with respect to P 1 Atm. Conversely, both the No Combustion and MILD regimes shrink in the ΔT space. It means that the stabilization of MILD regimes at higher pressures requires lower minimum preheating levels but higher minimum dilution levels.
Finally, a comparison between the HFDF and HODF configurations for NH 3 combustion was carried out. In the HFDF case the High temperature combustion zone is wider than that under HODF conditions. Furthermore, in the latter configuration, the MILD part widens for higher ΔT values when compared with the HFDF one. The main information derived FIGURE 10 | Map of behavior for pure ammonia in the HODF configuration at P 1 atm.
Frontiers in Energy Research | www.frontiersin.org October 2021 | Volume 9 | Article 649141 from such analysis is that MILD conditions can be stabilized for lower fuel dilution levels when the oxidant flow is preheated, with respect to the fuel preheating case. The characterization of the reactive ignidiffusive structures, in highly diluted and/or preheated conditions, provided key information on the sustainability and potentiality of MILD Combustion with new fuels for several technological applications.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
Sorrentino and Ariemma were involved in the bibliographic research to include in the manuscript the most important literature contributions on the diffusion ignition process in MILD Combustion. They have also strongly contributed in writing several parts of the article. Sabia and de Joannon gave their support in the theoretical background regarding categorization and classification of MILD Combustion processes. Ragucci gave important insights and hints in the physical description of diffusion ignition and its connection with dimensionless numbers. He also strongly revised the English language in the manuscript.