Direct Numerical Simulation of a Reacting Turbulent Hydrogen/Ammonia/Nitrogen Jet in an Air Crossflow at 5 Bar

Abstract

The article aims to analyze the fluid dynamics and combustion characteristics of a non-premixed flame burning a fuel mixture derived from ammonia partial decomposition injected in an air crossflow. Nominal pressure (5 bar) and inlet air temperature (750 K) conditions are typical of micro-gas turbines. The effects of strain on the maximum flame temperature and $\mathrm{NO}$ generation in laminar non-premixed counter-flow flames are initially explored. Then, the whole three-dimensional fluid dynamic problem is investigated by setting up a numerical experiment: it consists of a Direct Numerical Simulation, based on accurate transport, chemical, and numerical models. The flow topology of the specific reacting jet in crossflow configuration is described in terms of its main turbulent structures, like shear layers, ring, and horse-shoe vortices, as well as of its leeward recirculation region anchoring the flame. The reacting region is characterized by providing instantaneous spatial distributions of temperature, heat release, and some transported chemical species, including $\mathrm{NO}$, and calculating the Flame Index to identify non-premixed and premixed combustion local conditions. The latter is quantified by looking at the distribution of the volume fraction associated with a certain Flame Index versus the Flame Index and at the distribution of the average values of both the Heat Release Rate and $\mathrm{NO}$ versus the Flame Index and the mixture fraction.

1. Introduction

In recent years, the pursuit of alternatives to fossil fuels has greatly expanded. Among these, ammonia has emerged as a noteworthy substance for storing hydrogen [1,2], which can be easily broken down and used as a sustainable fuel for both transportation fuel cells and power generation.

Technically speaking, ${\mathrm{NH}}_3$ offers considerable advantages when compared to hydrogen in terms of cost and convenience as a fuel because of its higher density, simple production, simplified storage, handling, and distribution; furthermore, it benefits from an already well-established infrastructure, thus displaying potential for commercial viability. Notably, the hydrogen density of ${\mathrm{NH}}_3$ is four times higher than that of the most advanced methods involving metal hydrides [3]. Although ammonia is toxic, it poses a considerably reduced fire risk compared to hydrogen and gasoline [4,5]. Moreover, the cost per unit of stored energy for ammonia is more cost-effective than that of hydrogen [2,6]. Additionally, ammonia has demonstrated its merit as a viable alternative for hydrocarbon fuels in gas turbine engines, showcasing the potential for achieving high efficiencies [7].

Nonetheless, there are notable combustion difficulties linked to the use of ${\mathrm{NH}}_3$. These challenges stem from its low flame speed, temperature, and reactivity, as well as its potential for substantial ${\mathrm{NO}}_{\mathrm{x}}$ production [8,9,10,11]. To tackle these challenges, supplementary fuels such as ${\mathrm{H}}_2$ have shown improvements in enhancing and stabilizing operational performance [12]. However, conventional combustor designs require fundamental reassessment and development to reduce ${\mathrm{NO}}_{\mathrm{x}}$ emissions before ${\mathrm{NH}}_3$ can be seriously regarded for extended use in cleaner and more efficient technologies, like gas turbines [13,14].

Several studies have indicated that a blend of approximately 30% ${\mathrm{H}}_2$ or ${\mathrm{CH}}_4$ with 70% ${\mathrm{NH}}_3$ can enhance combustion stability. Such a blending strategy can be adopted to overcome the problem of ${\mathrm{NH}}_3$ combustion poor reactivity and to manage mixtures with combustion properties close to methane, useful features for gas turbine applications. Raising the ammonia fraction in the fuel blend promotes lean blowout but diminishes the tendency to flashback. This latter impact is particularly noticeable when the volume fraction of ammonia in the fuel blend surpasses 0.7. Under these conditions, boosting the equivalence ratio at a constant bulk velocity prevents flashback, leading to rich blowout instead, resulting in a significantly broader spectrum of stable equivalence ratios [15].

In the realm of gas turbine operations, Valera-Medina et al. [13] have presented studies involving a generic swirl burner fueled by various ammonia/methane blends: although reduced ${\mathrm{NO}}_{\mathrm{x}}$ and CO emissions were found at high equivalence ratios (>1.10), it was also stressed that conventional injection methods are unsuitable for ammonia-based blends. Hayakawa et al. [16] have explored reaction enhancement and stability of ammonia flames by adding hydrogen, demonstrating the potential for efficient ${\mathrm{NH}}_3$ combustion using ammonia blends. Research conducted by Pugh et al. [17], Zhang et al. [18], and Franco et al. [19] has examined the product gas characteristics of ammonia/hydrogen flames within industrially relevant swirl burner flame configurations.

Sorrentino et al. [20] demonstrated the feasibility of MILD (Moderate or Intense Low-oxygen Dilution) combustion for pure ammonia, showing good stability and emissions performance when the reactor temperature exceeded 1300 K. Following studies [21] have shown that introducing water into MILD ammonia combustion holds promise in enhancing both ${\mathrm{NO}}_{\mathrm{x}}$ emissions and stability, especially under fuel-lean conditions.

Experiments conducted on an industrial micro-gas turbine system have suggested a two-stage rich-lean combustion setup for pure ammonia [22], ammonia/methane blends [23], ammonia/hydrogen/nitrogen blends [24], and liquid ammonia [25] combustion. Numerical studies suggest that optimizing the equivalence ratio in the first (rich) combustion stage is essential for achieving low emissions at the outlet [26]. Furthermore, chemical kinetics investigations demonstrated that using a staged combustion (consisting of a rich stage, followed by a relaxation stage before the final lean stage) coupled with high-pressure conditions (>20 bar), it is fundamentally possible to achieve ${\mathrm{NO}}_{\mathrm{x}}$ emissions as low as in natural gas turbines [27].

Due to the convenience of distributing and storing ammonia compared to hydrogen, it is plausible to consider that in many instances, when relying on co-firing ammonia with hydrogen, the necessary hydrogen can be locally generated by breaking down a portion of the available ammonia. Several researchers have investigated the structure of partially cracked ammonia-air flames [28], leading to the development of thermal cracking processes. Additionally, other investigators [29,30] have recorded successful power generation through catalytic processes using the cracked ammonia molecule. However, cracking 2 moles of ammonia produces 3 moles of hydrogen and 1 mole of nitrogen. The nitrogen can either be separated or, to avoid separation costs, introduced into the burner alongside hydrogen and ammonia. The nitrogen introduced into the burner will serve as a diluent, causing chemical effects and most likely impacting emissions and flame stabilization. Khateeb’s research [15,31] demonstrated that the introduction of nitrogen into the ammonia-hydrogen blend decreases NO emissions, promotes lean blowout, and retards flashback. Furthermore, the computational investigation in [32] unveiled that for nearly equivalent adiabatic flame temperature, thickness, and propagation velocities, premixed laminar ${\mathrm{NH}}_3/{\mathrm{H}}_2/{\mathrm{N}}_2$-air flames demonstrate significantly greater resistance to strain-induced extinction compared to laminar ${\mathrm{CH}}_4$-air flames [33].

Since ${\mathrm{NO}}_{\mathrm{x}}$ formation times are of the order of milliseconds, reducing the flame length as it is typically done in micro-mixing combustion systems could be a successful strategy for reducing it. Such technology is currently under investigation for hydrogen combustion [34,35,36]. In this work, the authors would like to check if it can be extended to ammonia/hydrogen blends by investigating a non-premixed jet flame of ${\mathrm{NH}}_3/{\mathrm{H}}_2/{\mathrm{N}}_2$ injected in an air crossflow at 5 bar, simulated in the Direct Numerical Simulation (DNS) context. Despite the simplicity of the geometric configuration, its fluid dynamics are complex and make it possible to highlight the effects of turbulent mixing and strain on flame structure and stabilization and on $\mathrm{NO}$ formation.

The paper is organized as follows: Section 2 quantifies $\mathrm{NO}$ formation in laminar diffusion flames with the fuel having the same composition of the DNS jet mixture at different strain rates; Section 3 describes the DNS configuration; Section 4 presents the results, highlighting the flow and flame structures, characterizing the combustion mode by means of the Flame Index and mixture fraction, and quantifying the pollutants production; conclusions follow in Section 5.

2. Laminar Diffusion Flames

In a high-velocity diffusion flame, fuel and oxidant are subjected, in the region where they mix, to unsteady strain due to turbulent vortices. This section investigates the effect of strain-rate variation on laminar non-premixed counter-flow flames with the fuel side having the same composition mixture as the jet in the DNS numerical experiment. In particular, a numerical study with OpenSMOKE++ software [37] (https://www.opensmokepp.polimi.it/, accessed on 19 November 2023) is conducted to measure the maximum flame temperature and $\mathrm{NO}$ concentration as a function of the strain rate up to the quenching value. The species transport equations incorporated thermal diffusion (Soret effect), and individual species mixture-averaged diffusion coefficients were taken into account [38]. The counter-flow flame in its axisymmetric configuration is shown in the panel of Figure 1. The global strain rate χ is obtained as follows [39]:

$$\mathsf{\chi }=\frac{2|u_{\mathrm{O}}|}{\mathrm{L}}\left[1+\frac{|u_{\mathrm{F}}|}{|u_{\mathrm{O}}|}\sqrt{\frac{{\rho }_{\mathrm{F}}}{{\rho }_{\mathrm{O}}}}\right],$$

(1)

where $\rho$ is the density, u the velocity of fuel (F) and oxidizer (O) fluxes separated by the distance L.

The adaptive grid refinement technique was employed during the simulations. The relative and absolute errors of the nonlinear equations iterative process during the simulations were fixed to ${10}^{-6}$ and ${10}^{-12}$, respectively. In ammonia-hydrogen and air flames, the flame is positioned towards the oxidizer side of the stagnation plane. Extinction strain rates were determined by initially simulating a stable low-strain flame and gradually increasing the strain rate in a series of simulations until reaching the point of extinction. The kinetic scheme used matches the DNS simulation.

Figure 1 shows temperature and $\mathrm{NO}$ concentration trend as a function of the strain rate for different chemical mechanisms. The trends are the same for all mechanisms; the Gotama one shows a slight deviation at high strains; it was adopted in the present DNS since it is specifically improved for laminar burning velocities under a fuel-rich flame, i.e., up to an equivalence ratio of 1.8, and pressure in the range 0.1–0.5 MPa [40]. Despite the decreasing temperature with increasing strain rate up to the quenching, $\mathrm{NO}$ concentration increases. In fact, the $\mathrm{NO}$ Thermal formation path becomes less important as the temperature decreases, while the weight of the fuel pathway increases, as shown in Figure 2, which depicts the normalized $\mathrm{NO}$ formation rate for each reaction distinctive of different pathways [41]. The panel of Figure 2 depicts the relative importance [42] of each path in the $\mathrm{NO}$ formation for two different strains. So even though temperature decreases as strain increases, and thermal path loses importance in the $\mathrm{NO}$ formation, 70% of it is formed via the fuel path reaching ∼10,000 ppm before quenching.

3. The Numerical Experiment and Its Computational Setup

The three-dimensional compressible Navier–Stokes equations are solved to study the combustion behavior of a reactive transverse jet of NH${}_3$/H${}_2$/N${}_2$ fuel and a crossflow of heated non-vitiated air at the pressure of 5 bar.

In principle, for applications, it would be convenient to have a fuel blend with nominal flame properties identical to methane while gaining the positive effects of hydrogen [43]. Such a mixture is adopted for the present case: 40% NH${}_3$, 45% H${}_2$, and 15% N${}_2$ by volume. This mixture is discharged into a crossflow of air through a round nozzle with $\mathrm{d}=1\mathrm{mm}$, positioned flush with a smooth solid surface. The jet center aligns with the spanwise symmetric plane, and it is located along the streamwise direction at $\mathrm{x}=0$. For the sake of clarity, the origin of the system coincides with the jet center. Nominal conditions of the inlet flows are summarized in

Table 1. Flow inlets conditions.

	Jet Flow	Crossflow
Species	NH${}_3$/H${}_2$/N${}_2$	O${}_2$/N${}_2$
Composition, (% by vol.)	40/45/15	21/79
Pressure, (bar)	5	5
Velocity, (m/s)	150	50
Temperature, (K)	750	850
Density, (kg/m${}^3$)	0.9558	2.0411
Viscosity, (kg/m/s)	2.8437 × 10${}^{-5}$	3.8510 × 10${}^{-5}$

. The crossflow air temperature is ${\mathrm{T}}_{\mathrm{cf}}=850\mathrm{K}$ (a usual value at the end of the compression stage in stationary gas turbines), while the fuel jet is at ${\mathrm{T}}_{\mathrm{j}}=750\mathrm{K}$ (a value coherent with the thermal cracking of ammonia). The crossflow freestream velocity ${\mathrm{u}}_{\mathrm{cf}}=50\mathrm{m}/\mathrm{s}$ and the jet velocity ${\mathrm{u}}_{\mathrm{j}}=150\mathrm{m}/\mathrm{s}$ result in a velocity ratio $\mathrm{r}=3$ and a momentum flux ratio $\mathrm{J}=1.97$, and the jet Reynolds number is ${\mathrm{Re}}_{\mathrm{j}}=\mathsf{\rho }{\mathrm{u}}_{\mathrm{j}}\mathrm{d}/$$\mathsf{\mu }\approx 5042$ (see

Table 2. Some characteristics of the numerical experiment.

Jet exit diameter, d (m)	0.001
Domain extent, L${}_{\mathrm{x}}$ × L${}_{\mathrm{y}}$ × L${}_{\mathrm{z}}$	25d × 20d × 14d
Grid size, N${}_{\mathrm{x}}$ × N${}_{\mathrm{y}}$ × N${}_{\mathrm{z}}$	1008 × 1008 × 672
Velocity ratio, $\mathrm{r}={\mathrm{u}}_{\mathrm{j}}/{\mathrm{u}}_{\mathrm{cf}}$	3
Momentum flux ratio, $\mathrm{J}=$ $\mathsf{\rho }$${}_{\mathrm{j}}{\mathrm{u}}_{\mathrm{j}}^2/$$\mathsf{\rho }$${}_{\mathrm{cf}}{\mathrm{u}}_{\mathrm{cf}}^2$	1.97
Jet Reynolds number, Re${}_{\mathrm{j}}$	5042

).

The boundary conditions are set as follows. The incoming air crossflow is introduced into the computational domain through an inflow boundary condition (where velocity, temperature, and composition are fixed, while pressure and density evolve with the flow inside the computational domain). The velocity fluctuations prescribed at the inlet of the domain are of the order of 5% of the mean inflow air velocity. No turbulent velocity fluctuations are imposed on the jet [44]. This assumption arises from the jet’s relatively low Reynolds number, which has a minor influence on the jet features with respect to the significant impact exerted by the turbulence from the incoming crossflow and their interaction. At the lower wall boundary ($\mathrm{y}=0$), a no-slip adiabatic solid surface condition is imposed, while the upper wall ($\mathrm{y}=0.02\mathrm{m}$) is treated as an Eulerian (slip) adiabatic solid surface. The flame leaves the domain through a non-reflecting outlet boundary condition. Periodicity is enforced in the spanwise direction. A schematic of the applied boundary conditions is exposed in Figure 3.

The physical dimensions of the DNS domain are ${\mathrm{L}}_{\mathrm{x}}\times {\mathrm{L}}_{\mathrm{y}}\times {\mathrm{L}}_{\mathrm{z}}=25\mathrm{d}\times 20\mathrm{d}\times 14\mathrm{d}$, respectively, in the streamwise, wall-normal, and spanwise directions. The choice of grid spacing aims to adequately resolve both the flame and turbulence structures. A grid with $\Delta$x = $\Delta$y = $\Delta$z $\approx 10$ µm is employed in the region defined by $\mathrm{x}/\mathrm{d}\in [-1,9]$, $\mathrm{y}/\mathrm{d}\in [1,8]$, and $\mathrm{z}/\mathrm{d}\in [-1.5,1.5]$, encompassing most of the flame regions. Algebraic stretching is then applied in all three spatial directions, reaching a maximum grid size of 130 µm. The first cell height in the wall-normal direction is set to 3 µm, ensuring ${\mathrm{y}}^{+}=1$ at the lower wall. The Kolmogorov scale, denoted as $\mathsf{\eta }$ = ${\mathrm{L}}_{\mathrm{t}}/{\mathrm{Re}}_{\mathrm{T},\mathrm{cf}}^{0.75}$, exhibits a minimum value of 54 µm. Here, the used length scale is ${\mathrm{L}}_{\mathrm{t}}=0.02$ m and ${\mathrm{Re}}_{\mathrm{T},\mathrm{cf}}$ is the crossflow turbulent Reynolds number, where the imposed velocity fluctuation was adopted, i.e., the 5% of the crossflow inlet velocity. The dissipative length scale calculated through the jet parameters seems to be not of practical usage in this study since, as discussed further below, no turbulent velocity fluctuations are applied on the jet inlet boundary condition, as motivated by the study of Kolla [44]. The same reasoning can be made about the flame zone, where the increase in viscosity with temperature raises the value of $\mathsf{\eta }$. The final mesh has a total number of grid points that is ${\mathrm{N}}_{\mathrm{x}}\times {\mathrm{N}}_{\mathrm{y}}\times {\mathrm{N}}_{\mathrm{z}}=1008\times 1008\times 672$, resulting in approximately 683 M cells.

Molecular transport in the multicomponent mixture is modeled using the Hirschfelder and Curtiss approximate formula [45], taking into account the Soret thermo-diffusive effect and the pressure gradient diffusion. At run-time, binary diffusion coefficients and the n-th species thermo-diffusion coefficients are computed using kinetic theory expressions. Dynamic viscosity and thermal conductivity for individual chemical species are pre-calculated using software libraries [46,47]. Mixture-average properties are estimated using Wilke’s formula with Bird’s correction for viscosity [48,49], and Mathur’s expression for thermal conductivity [48,50]. Preferential diffusion is modeled based on the Hirschfelder and Curtiss law [45]. It is worth pointing out that this study does not take into consideration radiative heat transfer effects.

The chemical reactions are modeled using the Gotama [51] chemical kinetics mechanism specifically designed for ammonia, blended with hydrogen, and combustion in air. This mechanism encompasses 26 distinct species and 119 elementary reaction steps.

The numerical simulations are conducted using the in-house parallel code HeaRT and ENEA’s supercomputing facility CRESCO [52]. The code solves the compressible Navier–Stokes equations, discretized using staggered finite-difference schemes. Each chemical species has its individual transport equation. Diffusive fluxes are calculated using a second-order accurate centered scheme; convective terms are modeled through the $AUS{M}^{+}$-$up$ method [53], coupled with a third/fifth-order accurate $WENO$ interpolation to minimize spurious oscillations. Extended non-reflecting boundary conditions [54,55,56] are employed at open boundaries to consider the impact of variable transport properties [57] and local heat release [58]. At the airflow inlet, a synthetic turbulence generator is utilized [59].

4. Results

4.1. Flow-Field Organization and Features

In this section, a brief overview of the fluid dynamics of the present jet in crossflow configuration (JICF) is provided, based on the analysis of instantaneous quantities from DNS. Figure 4 shows some iso-surfaces of the ${\mathrm{NH}}_3$ concentration and temperature, demonstrating the complex topology of the configuration. Looking at these iso-surfaces, the large vortical structures of the windward shear layer developing from the fuel jet are clearly visible: they produce mixing with the crossflow air stream by increasing the contact surfaces of the two streams. More downstream, such coherent structures develop to smaller turbulent scales, enhancing mixing at finer levels. The projection of the Heat Release Rate (HRR) of the central x−y plane is also depicted on the periodic plane and that of the y−z plane at x = 7d on the outlet plane. High HRR and temperature are observed within the leeward branch of the flow; this aspect will be considered in more detail later in this section. Moreover, in the periodic plane of the same picture, the stoichiometric mixture fraction iso-line is also shown (white line). The mixture fraction is calculated following the study of Tang [60] on a non-premixed NH${}_3$/H${}_2$/N${}_2$ jet flame as follows:

$$\mathsf{\xi }=\frac{(Y_H-Y_{H,2})/2W_H-(Y_O-Y_{O,2})/W_O}{(Y_{H,1}-Y_{H,2})/2W_H-(Y_{O,1}-Y_{O,2})/W_O},$$

(2)

where ${\mathrm{Y}}_{\mathrm{O}\left(\mathrm{H}\right)}$ and ${\mathrm{W}}_{\mathrm{O}\left(\mathrm{H}\right)}$ are the local elemental mass fractions and atomic masses of the species O and H. In the same expression, the subscripts 1 and 2 indicate the fuel jet and the air crossflow, respectively. Although the stoichiometric conditions are reached on both the windward and leeward sides of the fuel jet, no flame develops on the first region due to too short convective times and high strain rates, while the flame can develop around the stoichiometric iso-line in the latter region due to the longer residence times.

Figure 5 shows a snapshot of the vorticity magnitude ($\mathsf{\omega }$) projected on one of the periodic plane: the spatial distribution sheds light on the complex interaction between the crossflow and the transverse jet. The flow field is characterized by the formation and expansion of the shear layer, which then rolls up into a clear pattern of Kelvin-Helmholtz hydrodynamic instabilities. These vortical structures are believed to play a crucial role in governing key JICF processes like mixing and penetration [61]. As the jet evolves, these rolling vortices break into smaller and smaller structures as a result of the energy cascade phenomenon and are finally convected downstream, out of the domain, while being elongated in the streamwise direction. Vortex pairing is also visible, occurring on both the leeward and the windward side of the jet. These instantaneous turbulent structures are also represented three-dimensionally in the same figure, identified through the Q-criterion and colored by temperature. The formation of ring-like vortices (RLV) in the shear layer is clearly visible, as well as the horse-shoe vortex (HSV), which is in front of the windward side exit of the jet as a result of the interaction with the crossflow. This vortex further develops around the jet and dissipates far downstream. The overall described configuration is typical of this JICF problem with nominally $\mathrm{r}>1$. In fact, when the velocity ratio is considerably lower than the unity, another flow configuration with the presence of hairpin vortices is more likely to occur [62].

Figure 6 displays the instantaneous spatial distribution of the velocity components in the streamwise direction (${\mathrm{u}}_{\mathrm{x}}$) and the wall-normal direction (${\mathrm{u}}_{\mathrm{y}}$) within the central x−y plane. The distribution of the streamwise velocity is closely intertwined with the behavior of the crossflow and the transverse jet. Specifically, as the crossflow encounters the jet, it triggers the formation of shear layer vortices and leads to a notable reduction in the magnitude of ${\mathrm{u}}_{\mathrm{x}}$ immediately ahead of the jet, resulting in a recirculation zone that develops into an HSV. The figure also highlights an increase in ${\mathrm{u}}_{\mathrm{x}}$ values along the jet trajectory. Additionally, negative values of ${\mathrm{u}}_{\mathrm{x}}$ are found in the immediate vicinity of the jet on its leeward side, indicative of the entrainment of the crossflow and the presence of a recirculation zone. This recirculation zone effectively captures the high-temperature products generated by combustion on the leeward side, therefore contributing to the stabilization of the flame. The strongly different behavior of the windward and leeward regions of the fuel jet is also clear from the intensity of the vortex shedding in the related shear layers: the one on the leeward side of the jet is less pronounced compared to the windward side, attributable to an upsurge in flow viscosity around the flame. Conversely, the wall-normal velocity distribution indicates that the jet, characterized by significant values of ${\mathrm{u}}_{\mathrm{y}}$, is notably affected by the crossflow. This emphasizes the role of the crossflow in augmenting the mixing between the jet and the surrounding flow. From Figure 6b, the fuel jet penetration can also be deduced: its potential core reaches nearly ∼5 mm in the y-direction (nearly one quarter of the channel height), while its plume goes slightly above one third of the channel height.

Figure 7 shows the temperature field in six x−z planes at different distances from the bottom wall, demonstrating the wake behind the fuel jet. At shorter distances, the flame is limited to a small area and does not appear continuous, showing some dispersed hot spots asymmetrically released downstream. However, at greater distances from the wall, the reacting area significantly expands due to the accelerated rate of crossflow fluid entrainment [63,64] and gas expansion caused by the chemical reactions. Moreover, the temperature field clearly shows that the flame is well anchored to the leeward side of the jet.

A different view of the temperature field is provided in Figure 8, referring to the central x−y plane. The windward flame is absent and not situated within the shear layer, where the crossflow interacts with the jet. This is confirmed by the HRR (see Figure 4), which reaches its maximum values on the leeward side. The leeward branch of the flame is anchored within the shear layer near the jet exit and a broader reaction zone can be observed, undergoing to gas expansion with strong density gradients and with temperatures above 2400 K. In fact, the flame is situated in an area with lower airflow velocity and, consequently, lower strain rates. This region is characterized by longer residence times, allowing radicals and temperature to reach their highest levels. These results agree with those reported in the study of Grout [65], where the flame was anchored on the leeward side of the jet, coinciding with a region of low velocity and near stoichiometric mixture conditions. There was no distinct flame branch observed on the windward side, contrary to the study of Lyra [66], where a H${}_2$ JICF with hotter vitiated air was considered.

To close the overview of the topology of this reacting jet in crossflow, Figure 9a,b shows the instantaneous spatial evolution of the jet mixture fuel components, ${\mathrm{H}}_2$ and ${\mathrm{NH}}_3$, respectively. As the jet evolves and interacts with the crossflow, the fuel is consumed due to the combustion process, thus producing radicals and more stable combustion products. Figure 9c,d report the H and NO instantaneous field, respectively. It is worth noticing the high level of NO emissions, exceeding 20,000 ppm, and the strong correlation with high-temperature regions.

4.2. Analysis of the Combustion Modes

The turbulent structures deriving from the complex interaction between the fuel jet and the air jet in crossflow produce an intense mixing of reactants. Hence, combustion is likely to happen in both non-premixed and premixed modes, depending on the flow region.

The Flame Index helps in identifying such combustion modes. In particular, the Takeno index is calculated in its revised fuel multispecies formulation as $\mathrm{FI}=\nabla {\mathrm{Y}}_{\mathrm{F}}·\nabla {\mathrm{Y}}_{{\mathrm{O}}_2}/\left(\right|\nabla {\mathrm{Y}}_{\mathrm{F}}\left|\right|\nabla {\mathrm{Y}}_{{\mathrm{O}}_2}\left|\right)$, $\nabla {\mathrm{Y}}_{\mathrm{F}}$ being the gradient of the fuel species (${\mathrm{Y}}_{\mathrm{F}}={\mathrm{Y}}_{{\mathrm{NH}}_3}+{\mathrm{Y}}_{{\mathrm{H}}_2}$) [67]. Figure 10 reports an instantaneous Flame Index distribution in the central longitudinal x−y plane. It can be observed that $\mathrm{FI}\in [-1,1]$: this means that, although the overall flame is non-premixed, in fact, reacting structures are associated with both non-premixed ($\mathrm{FI}<0$) and premixed ($\mathrm{FI}>0$) combustion modes. At the same time, it must be stressed that no flame is exhibited on the windward side of the fuel jet, as shown in Figure 8. Hence, from Figure 10 it is hard to say which of the combustion modes is the most probable and relevant for flame anchoring.

At first, to understand which of the two combustion modes is the most relevant in terms of occupied volume, the volume fraction distribution in the computational domain as a function of the Flame Index $\mathrm{FI}\in [-1,1]$ was calculated, i.e., $\mathrm{V}\left(\mathrm{FI}\right)={\sum }_{\mathrm{n}=1}^{{\mathrm{N}}_{\mathrm{i}}}{\mathrm{V}}_{\mathrm{n}}\left(\mathrm{FI}\right)/{\mathrm{V}}_{\mathrm{Tot}}$, ${\mathrm{N}}_{\mathrm{i}}$ being the number of volumes at a fixed value of FI. Such a distribution is shown in Figure 11a, and it reveals that most of the volume can be attributed to the non-premixed mode. However, this analysis is based only on the mixing of reactants and does not consider the effects of local convective times and strains that can negatively affect the local ignition of the potentially reactive mixture. Hence, as already observed for the spatial distribution of FI in Figure 10, this analysis is not sufficient to draw conclusions on the most effective combustion mode.

From the previous analysis, to understand which of the combustion modes is the most relevant in this nominally non-premixed configuration, the Flame Index must be correlated with the local heat release. Figure 11b reports the mean Heat Release Rate distribution as a function of the Flame Index and mixture fraction $\mathsf{\xi }$. It is observed that the maximum heat release is found in regions near the stoichiometric mixture fraction ($\mathsf{\xi }=0.14$), but, unexpectedly, for positive values of the Flame Index associated with premixed combustion. These regions are located several diameters downstream of the fuel jet where reactants had time to mix. The heat released from the diffusion flame is located immediately downstream of the fuel jet, and is mainly responsible for the flame anchoring. Here, vortices of the jet diameter size contribute to bringing part of the crossflow air immediately downstream of the injection region and to intensively stretch the newly formed diffusion flames, thus reducing the peak value of the heat release in comparison to the premixed regions.

What was found from the combustion modes analysis has interesting implications on $\mathrm{NO}$ emissions. Figure 11c shows the mean NO pollutant concentration in dry ppm (dppm) as a function of the Flame Index and mixture fraction $\mathsf{\xi }$. As expected, it is mostly formed in the near stoichiometric regions, but the maximum values attain to the non-premixed combustion mode ($\mathrm{FI}<0$). Since the present jet in crossflow configuration exhibits such diffusion flames in zones characterized by high strain, this seems in agreement with findings from laminar diffusion flames where higher NO concentrations correspond to higher strains. The NO concentrations are well above the permitted limits; this is due to the presence of hydrogen in the fuel, which enhances thermal $\mathrm{NO}$ by increasing the maximum temperatures and fuel-$\mathrm{NO}$ by increasing the concentration of O and H radicals. This suggests working with ${\mathrm{H}}_2$ concentrations in fuel mixtures lower than the present one. Further reduction of $\mathrm{NO}$ can be achieved by increasing the nominal pressure [27]. This will be the aim of future work.

5. Conclusions

The article investigated the turbulent non-premixed combustion of a fuel mixture derived from ${\mathrm{NH}}_3$ partial decomposition in an air crossflow at 5 bar by means of DNS.

Preliminary chemical kinetics simulations of laminar flames in a non-premixed counter-flow reactor demonstrated that the maximum temperature decreases as the applied strain increases, while the maximum $\mathrm{NO}$ increases: in fact, under higher strains, the fuel-$\mathrm{NO}$ production is enhanced, while the thermal-$\mathrm{NO}$ becomes less important.

The three-dimensional DNS instantaneous flow field was analyzed, demonstrating the complex structure of this ideally simple jet in crossflow configuration. The flame develops and anchors on the leeward side of the fuel jet. The flow topology was described, demonstrating the different intensities of the windward and leeward shear layers, the latter having shed vortices with shorter lifetimes due to the damping effect of the reacting region. The recirculation zone behind the jet anchors the flame. The main turbulent structures, like ring and horse-shoe vortices, were identified. The analysis provided temperature, heat release, and main transported chemical species instantaneous spatial distributions.

More insights were given investigating the local combustion modes experienced in the flame, i.e., non-premixed and premixed. Although the configuration is globally non-premixed, the Flame Index distribution demonstrated both non-premixed (FI < 0) and premixed (FI > 0) combustion modes, locally. The non-premixed mode is the most relevant from a volumetric point of view, but the premixed one is the most relevant in terms of heat release. The local premixed combustion volumes belong to a low-strain region on the leeward side of the jet. Coherently, the $\mathrm{NO}$ concentration is higher in the local non-premixed combustion volumes belonging to a region of high strain, in agreement with the findings from laminar frames.

The results show that the jet in crossflow configuration, nominally non-premixed, can be potentially adopted to generate a non-negligible premixed combustion region. How to control the actual mixing and the stoichiometry of the desired premixed reacting region becomes the next critical step for the exploitation of the investigated strategy in applications. Reducing the hydrogen content in the fuel mixture can be of help: in fact, increasing the ignition delay time provides more time for turbulent mixing to act before combustion takes place in a non-premixed mode. Lower hydrogen content also helps in lowering $\mathrm{NO}$ emission; further reduction can be achieved by operating with higher nominal pressures. Moreover, aiming at a ${\mathrm{NH}}_3/{\mathrm{H}}_2/{\mathrm{N}}_2$ fuel blend with less hydrogen reduces the cracking requirements of ammonia.