1 s2.0 S2666352X23001097 Main

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Applications in Energy and Combustion Science 16 (2023) 100220

Contents lists available at ScienceDirect

Applications in Energy and Combustion Science


journal homepage: www.elsevier.com/locate/jaecs

Large eddy simulation investigation of ammonia spray characteristics under


flash and non-flash boiling conditions
Ziwei Huang, Haiou Wang ∗, Kun Luo, Jianren Fan
State Key Laboratory of Clean Energy Utilization, Zhejiang University, Hangzhou 310027, China

ARTICLE INFO ABSTRACT

Keywords: Ammonia is an ideal zero-carbon fuel for energy systems, and it can be used directly in a liquid state. However,
Liquid ammonia liquid ammonia is susceptible to flash boiling due to its unique physical and chemical properties, which
High-pressure injection brings challenges for liquid ammonia spray and combustion simulations. In the present study, high-pressure
Flash boiling
liquid ammonia injection under flash boiling and non-flash boiling conditions was investigated. Large eddy
Large eddy simulation
simulations were conducted in an Eulerian–Lagrangian framework. The aim is to explore in detail the influence
of the critical parameters on the characteristics of liquid ammonia spray. Under the flash boiling condition, the
effects of ambient pressure were examined. Comparisons of the measured and predicted spray penetration and
morphology demonstrate that the present simulations can reproduce liquid ammonia spray characteristics well.
Varying ambient pressures causes the ammonia spray to be at different superheat levels, resulting in significant
changes in spray features. Under the non-flash boiling condition, the effects of injection pressure and ambient
pressure were investigated. The higher injection pressure feature a higher gas velocity and improves the mixing
of ammonia and air. The ammonia spray propagates more rapidly and the spray width becomes wider at lower
ambient pressure.

1. Introduction systems. For gas turbine combustors, supplying ammonia in liquid state
to the combustor can get rid of the pre-vapor equipment, which reduces
Driven by the goal of carbon neutrality, it is imperative to develop the cost and start-up time compared to supplying ammonia in gaseous
zero-carbon combustion engines. Fossil fuels should be replaced by phase. Okafor et al. [17] reported the first successful experimental
carbon-free fuels such as hydrogen and ammonia to mitigate carbon study on direct combustion of liquid ammonia spray in a gas turbine-
emissions. Hydrogen is a promising clean fuel, and its combustion
like combustor, where the preheated swirling air was used to enhance
product is only water. However, issues associated with safe storage and
flame stability, and it was found that co-firing with methane improved
transportation of hydrogen bring challenge to its implementation [1].
the liquid ammonia spray combustion performance. The authors further
Alternatively, ammonia has a high vapor pressure, hence ammonia
is typically stored in liquid phase which facilitates its storage and investigated the emission characteristics of liquid ammonia combus-
transportation. Owing to the above advantages, the use of ammonia tion, indicating that two-stage rich-lean combustion can be employed to
as a fuel has attracted extensive attention [2–4]. However, utilization control the pollutant emission. In addition, the direct injection dual-fuel
of ammonia as a fuel in combustion engines faces challenges such strategy is also an effective method for liquid ammonia combustion.
as the relatively low flame speed [5] and high nitrogen oxide emis- Gross et al. [18] experimentally studied combustion and emission
sion [6–8]. In recent years, significant efforts have been made towards characteristics of a compression-ignition engine using the mixture of
overcoming the above challenges in ammonia combustion. Previous liquid ammonia and DME. Scharl et al. [19] reported the mixture
studies have shown that ammonia-air flames can be stabilized by strong formation and combustion behavior of ammonia sprays under high-
swirling flows [9,10], which generates recirculating flows that promote pressure direct-injection conditions, where ammonia spray flames are
transport of heat and active radicals [11]. Another effective method is stabilized by a diesel pilot injection. Somarathne et al. [20] numerically
blending active fuel such as hydrogen or methane to improve ammonia
investigates liquid ammonia spray combustion with hydrogen/air co-
combustion performance [12–16].
firing in a swirl combustor using large eddy simulation (LES) with finite
At present, most studies focused on gaseous ammonia combustion.
rate chemistry combustion models. The study successfully achieved
In fact, ammonia can be used directly in a liquid state for energy

∗ Corresponding author.
E-mail address: wanghaiou@zju.edu.cn (H. Wang).

https://doi.org/10.1016/j.jaecs.2023.100220
Received 19 July 2023; Received in revised form 28 October 2023; Accepted 29 October 2023
Available online 11 November 2023
2666-352X/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

stable liquid ammonia spray flames over a wide range of equiva- 2. Numerical methodology
lence ratios and the equivalence ratio corresponding to the minimum
emission was determined. 2.1. Governing equations of the gas phase
Although considerable advances have been made recently on liq-
uid ammonia combustion, most of the studies focused on practical In large eddy simulations, spatially filtered Navier–Stokes compress-
problems such as liquid ammonia spray flame stability and nitrogen ible equations of continuity, momentum, energy, and species mass frac-
oxide emission. However, important fundamental processes such as tion for non-reacting two-phase flows are solved. The set of transport
liquid ammonia spray evolution are not well understood. Liquid am- equations have the form as:
monia is prone to flash boiling due to its low boiling point, which 𝜕 𝜌̄
̄𝒖) = 𝑆̄̇𝜌
+ ∇ ⋅ (𝜌̃ (1)
significantly changes the morphological features of spray [21,22], and 𝜕𝑡
affects the fuel/air mixing characteristics. Pele et al. [23] reported ̄𝒖)
𝜕(𝜌̃
the first experimental study on ammonia spray characteristics under + ∇ ⋅ (𝜌̃𝑢) = −∇𝑝 + ∇ ⋅ (𝝉 − 𝝉 𝑆𝐺𝑆 ) + 𝑺̄̇𝒖
̄𝒖̃ (2)
𝜕𝑡
gasoline direct injection (GDI) engine-relevant conditions. The effects
𝜕(𝜌̄̃
ℎ) 𝐷𝑝
of ambient pressure and temperature on the characteristics of liquid ̄𝒖̃
+ ∇ ⋅ (𝜌̃ ̄ ̃
ℎ) − ∇ ⋅ (𝜌𝛼∇ℎ) = + 𝑆̄̇ℎ (3)
𝜕𝑡 𝐷𝑡
ammonia spray were investigated and compared with ethanol and
gasoline spray. The experimental results reveal that liquid ammonia ̃𝑘 )
𝜕(𝜌̄𝑌
+ ∇ ⋅ (𝜌̃̃𝑘 ) − ∇ ⋅ 𝜌(𝐷
̄𝒖𝑌 ̃𝑘 ) = 𝑆̄̇𝑘
̄ 𝑘 ∇𝑌 (4)
is more sensitive to changes in ambient density and temperature than 𝜕𝑡
other fuels. Subsequently, Cheng et al. [24] investigated evolution and where 𝜌, 𝒖 and 𝑝 are the density, velocity and pressure of the gas phase,
evaporation behaviors of ammonia sprays under different injection respectively. 𝝉 is the viscous stress tensor, and 𝝉 𝑆𝐺𝑆 is the sub-grid
pressure and air density condtions. The spray geometric parameters viscous stress tensor which is solved by Smagorinsky sub-grid model.
such as penetration length, cone angle and spray area were analyzed. Li ℎ is the specific enthalpy, and 𝛼 is the thermal diffusion coefficient. 𝑌𝑘
et al. [25] measured near-field and far-field features of ammonia spray and 𝐷𝑘 is the mass fraction and mass diffusivity coefficient of the 𝑘th
under different fuel temperature and ambient density conditions. The species, respectively. In current simulations, 𝛼 is equal to 𝐷𝑘 under the
experimental results are useful to understand the key parameters of the assumption that the Lewis number is unity. 𝑆̄̇𝜌 , 𝑺̄̇𝒖 , 𝑆̄̇ℎ and 𝑆̄̇𝑘 are the
breakup process of superheated ammonia spray and provide insightful corresponding source terms in the continuity, momentum, energy, and
data for developing simulation models of ammonia sprays. species mass fraction equations, respectively, which are calculated as
follows:
Spray formation and evolution is a complicated process, and nu-
1 ∑ 𝑑𝑚𝑝
𝑁
merical simulation is an effective method that can complement exper-
𝑆̄̇𝜌 = − (5)
imental studies of spray characteristics. Zhang et al. [26] numerically 𝛥𝑉 𝑖 𝑑𝑡
studied liquid ammonia direct injection spray characteristics under GDI
1 ∑ 𝑑𝑚𝑝 𝒖𝒑
𝑁
and diesel engine conditions. The aim was to explore the capabilities 𝑺̄̇𝒖 = − (6)
of various evaporation models developed within the Euler–Lagrange 𝛥𝑉 𝑖 𝑑𝑡
framework on the prediction of ammonia spray in a wide range of
1 ∑ 𝑑(𝑐𝑝 𝑚𝑝 𝑇𝑝 )
𝑁

ambient conditions. Li et al. [27] and Yousefi et al. [28] reported simple 𝑆̄̇ℎ = − (7)
𝛥𝑉 𝑖 𝑑𝑡
numerical simulations for ammonia/diesel dual-fuel engine, and the
numerical simulations were mainly carried out to support the experi- where 𝛥𝑉 and 𝑁 represent the volume of the cell and the number of
mental results and complement the analysis of the combustion process. droplets in the cell, respectively. 𝑚𝑝 , 𝑇𝑝 and 𝒖𝒑 is the mass, temperature
Pandal et al. [29] modeled ammonia spray in GDI engines, where and velocity of spray parcels, respectively. The value of 𝑆̄̇𝑘 is the same
as 𝑆̄̇𝜌 for the species of 𝑁𝐻3 , and it is zero for other species.
several conditions for flash and non-flash ammonia spray were selected,
and the simulation results were compared with the experiments in
terms of liquid and vapor penetration length, Sauter mean diameter 2.2. Governing equations of the liquid phase
and global spray morphology. An et al. [30] modeled ammonia spray
under different fuel temperatures and ambient pressure conditions The spray parcels are tracked using the Lagrangian particle tracking
using a novel phase-change model. The research focused on comparing (LPT) method. Under the Lagrangian framework, a group of droplets
with the same characteristics instead of each real droplet are tracked.
the performance of different phase-change models for predicting the
Therefore, the position 𝒙𝒑 , velocity 𝒖𝒑 , mass 𝑚𝑝 and temperature 𝑇𝑝 for
phase-change process of flash boiling ammonia spray.
the spray parcels are solved by the following equations [32]:
Despite of the above-mentioned studies, the phase-change process
d𝒙𝒑
of liquid ammonia spray is not well understood yet, which hinders = 𝒖𝒑 (8)
the utilization of liquid ammonia as a fuel in combustion engines. d𝑡
Most previous numerical studies have focused on the characteristics d𝒖𝒑
𝑚𝑝 = 𝑭𝒑 (9)
of ammonia spray in flash boiling state. The present study aims to d𝑡
explore the influence of key parameters, such as the ambient pressure
d𝑚𝑝
and injection pressure, on the characteristics of liquid ammonia spray = 𝑚̇ 𝑝 (10)
d𝑡
under the flash boiling and non-flash boiling conditions. In addition,
the Voronoi analysis [31] is introduced to understand the dispersive or d𝑇𝑝
𝑚𝑝 𝐶𝑝,𝑙 = 𝑚̇ 𝑝 ℎ(𝑇𝑝 ) + ℎ𝑐 𝐴𝑝 (𝑇 − 𝑇𝑝 ) (11)
clustering behavior of ammonia sprays. d𝑡
The remainder of this study is organized as follows. Firstly, govern- where 𝐶𝑝,𝑙 is the heat capacity of droplet and 𝑭 𝒑 is the Stokes drag
ing equations of the gas phase and liquid phase are introduced. Then, which can be determined as follows under the assumption of a solid
the numerical setup and simulation details are described. After that, sphere:
the results of ammonia sprays under various conditions are presented. 3𝜇𝐶𝐷 𝑅𝑒𝑝
𝑭𝒑 = (𝒖 − 𝒖𝒑 ) (12)
Finally, the conclusions of this study are drawn. 4𝜌𝑝 𝑑 2 𝑝

2
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Fig. 1. Pressure-enthalpy (𝑝–ℎ) diagram of liquid ammonia.

where 𝑅𝑒𝑝 is the droplet Reynolds number and 𝐶𝐷 is the drag coeffi-
cient which can be calculated as:
⎧ 24 ( 1 2∕3
)
⎪ 𝑅𝑒𝑝 1 + 6 𝑅𝑒𝑝 , 𝑅𝑒𝑝 ≤ 1000
𝐶𝐷 = ⎨ (13)
⎪ 0.424 , 𝑅𝑒𝑝 > 1000

Fig. 2. Computational domain used in the present study.
Liquid ammonia is prone to be superheated due to its low boiling
point. Hence, in current simulations, flash boiling phenomenon occurs
when high-pressure liquid ammonia is injected into low-pressure atmo- Table 1
Initial parameters for the numerical study.
sphere. In order to reasonably characterize the phase-change process of
Case 𝐴𝑚𝑏𝑖𝑒𝑛𝑡 𝑝𝑟𝑒𝑠𝑠𝑢𝑟𝑒 [MPa] 𝐼𝑛𝑗𝑒𝑐𝑡𝑖𝑜𝑛 𝑝𝑟𝑒𝑠𝑠𝑢𝑟𝑒 [MPa]
liquid ammonia in the flash boiling state, in addition to the equilibrium
evaporation model, the flash boiling model developed by Zuo et al. [33] PA0.1-PI100 0.1 100
PA0.5-PI100 0.5 100
is introduced. The total evaporation rate of superheated droplets can be PA0.9-PI100 0.9 100
determined using: PA4-PI50 4 50
d𝑚𝑝 PA4-PI75 4 75
= 𝑚̇ 𝑓 𝑙𝑎𝑠ℎ + 𝑚̇ 𝑡 (14) PA4-PI100 4 100
d𝑡 PA3-PI100 3 100
The flash boiling evaporation rate is estimated as: PA2-PI100 2 100
PA1-PI100 1 100
(𝑇𝑑 − 𝑇𝑏 )
𝑚̇ 𝑓 𝑙𝑎𝑠ℎ = 𝜋𝑑 2 𝑎𝑠 (15)
ℎ𝑣
where 𝑇𝑑 , 𝑇𝑏 and ℎ𝑣 is the internal temperature, boiling temperature
velocity and pressure solution is based on the compressible PIMPLE
and latent heat of vaporization for flash boiling droplets, respectively.
algorithm, which combines both the classical PISO and SIMPLE algo-
𝛼𝑠 denoted the overall heat transfer coefficient and can be determined
rithms to ensure both stability and accuracy. An implicit second-order
according the Adachi correlation [34]:
method is used for temporal discretization and second-order accu-
⎧ 0.26 racy is employed for spatial discretization. The governing equations
⎪760(𝑇𝑑 − 𝑇𝑏 ) , 0 ≤ 𝑇𝑑 − 𝑇𝑏 < 5
⎪ in the Lagrangian framework were integrated using the Euler-implicit
𝛼𝑠 = ⎨270(𝑇𝑑 − 𝑇𝑏 )2.33 , 5 ≤ 𝑇𝑑 − 𝑇𝑏 < 25 (16) scheme. For the sake of interphase two-way coupling, a cell inter-

⎪13800 × (𝑇𝑑 − 𝑇𝑏 )0.39 , 25 ≤ 𝑇𝑑 − 𝑇𝑏 polation scheme [37] is used to interpolate the gas phase quantities
⎩ to the particle location, whereas the two-way coupling terms were
The evaporation rate owing to external heat transfer between the interpolated using the semi-implicit scheme.
environment and the droplet can be calculated using: The simulation parameters of present work are summarized in Ta-
( ( )) ble 1. It is noted that the effect of ambient pressure on liquid ammonia
𝐾𝑆ℎ 1 ℎ𝑔 − ℎ𝑠 𝑚̇ 𝑓 𝑙𝑎𝑠ℎ
𝑚̇ 𝑡 = 𝜋𝑑 𝑚̇
ln 1 + 1+ (17) spray under flash boiling conditions are investigated by analyzing cases
𝐶𝑝 1 + 𝑓 𝑙𝑎𝑠ℎ ℎ𝑣 𝑚̇ 𝑡
𝑚̇ 𝑡 PA0.1-PI100, PA0.5-PI100 and PA0.9-PI100. Cases PA4-PI50, PA4-PI75
where, 𝐾 is the thermal conductivity of the droplet and 𝑆ℎ is Sher- and PA4-PI100 are performed to explore the effect of injection pressure
wood number. ℎ𝑔 and ℎ𝑠 are the gas phase enthalpy and surface on liquid ammonia spray under non-flash boiling conditions, while
enthalpy of droplet, respectively. The equation is solved using an cases PA4-PI100, PA3-PI100, PA2-PI100 and PA1-PI100 are studied to
iterative method [35] in OpenFOAM. examine the effect of ambient pressure on liquid ammonia spray under
non-flash boiling conditions.
3. Experimental and numerical setup The current configuration of high-pressure liquid ammonia sprays
was studied experimentally by Fang et al. [38]. In the experiment, a
The open-source software OpenFOAM [36] is used for the simula- single-hole diesel injector with a nozzle diameter of 0.168 mm was used
tions, which employs the finite volume method. The coupling of the to inject liquid ammonia with an injection pressure up to 100 MPa.

3
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Fig. 3. Comparison of spray penetration length obtained with different mesh resolutions.

The injection pressure measures the total amount of pressure required


to force the liquid fuel into the environment. The fuel selected for the
experiment is high-purity liquid ammonia, and the temperature of the
fuel is 300 K. The ambient pressure in the chamber varies between
0.1 MPa and 4 MPa, and the ambient temperature is 300 K. In order
to distinguish the flash boiling and non-flash boiling conditions more
clearly, the pressure-enthalpy (p-h) diagram of ammonia is introduced
in Fig. 1. It can be seen that when liquid ammonia is injected into a
pressure ambient below the saturation pressure, flash boiling occurs.
Fig. 2 shows the computational domain used in the present simu-
lations, which is a rectangular three-dimensional box of 70 × 70 × 85
𝑚𝑚3 . The whole domain was discretized with about 2.2 million hexa-
hedrons. The grid resolution is refined to 0.11 mm in the spray core
region and the grid growth rate in the spray injection direction is
set to 1.02 so that the liquid ammonia spray characteristics can be
reasonably captured. It is known that the mesh size may influence
the simulation results. Therefore, we have performed the verification
of grid independence. The case with an ambient pressure of 0.5 MPa
and an injection pressure of 100 MPa is analyzed as an example, and
the results are presented in Fig. 3. The spray penetration length is
defined as the maximum distance of 95% liquid mass to the injection
Fig. 4. Definition of liquid penetration.
location, which is a commonly used method of defining the spray
penetration length [39,40], as illustrated in Fig. 4. Three cases with
different grids are compared, and the grid numbers of which are 0.2,
2.2 and 4.4 million. The grid distributions in the central plane of and the Rosin-Rammler model was chosen for describing the droplet
various cases are displayed in Fig. 5. The grid resolution improves as size distributions. The maximum diameter, minimum diameter and
the grid number increases, as expected. As can be seen in Fig. 3, the mean diameter are set to 50 μm, 1 μm and 12 μm, respectively. The
simulation with a grid number of 0.2 million over-predicts the spray parameter selection is somewhat arbitrary as the experiment did not
penetration. The difference of the predictions between the simulations provide detailed information on droplet distribution in the near-field
with a grid number of 2.2 million and 4.4 million is minor. In order to region of the injection. It is noticed the grid size is about 2.2 times
balance the computational cost and the accuracy of the predictions, a of the maximum droplet diameter at the inlet. However, we note that
grid number of 2.2 million is employed in the present work. the diameters of large droplets decrease shortly after injection due to
As mentioned above, the experiment uses a single-hole injector, so secondary breakup, so that the ratio of the grid size to the droplet
the injection type chosen in the simulation is the cone injection model diameter is more than 5 in most regions of the computational domain.
In this model, the users specify the duration of injection, injection This ratio is comparable to those of many previous LES studies of
position, droplet flow rate, and spray cone angle. The initial droplet sprays [30,39,41,42].
velocity is calculated using the pressure-driven velocity method: The simulations were performed at the National Supercomputer
√ Center in Beijing, China. Each case under flash boiling conditions
2(𝑃𝑖𝑛𝑗 − 𝑃𝑎𝑚𝑏 )
𝑈 = 𝐶𝑑 (18) requires approximately 900 core hours and approximately 5000 core
𝜌 hours are required for each case under non-flash boiling conditions.
where 𝐶𝑑 is the discharge coefficient, 𝜌 represents the density of liquid
ammonia, and 𝑃𝑖𝑛𝑗 and 𝑃𝑎𝑚𝑏 are injection pressure and ambient pressure 4. Results and discussion
respectively.
For all cases in this study, the mass flow rate of the liquid fuel is In this section, the effects of ambient pressure on liquid ammonia
5×10−3 kg∕s and the injection angle is 30◦ . The distribution characteris- spray under flash boiling conditions were examined, while the effects of
tics of droplets are important for the Eulerian–Lagrangian simulations, injection pressure and ambient pressure on liquid ammonia spray were

4
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Fig. 5. Distributions of the grids in the central plane for various cases.

Fig. 6. Comparison of measured and predicted spray liquid penetration length in flash boiling conditions.

Fig. 7. Comparison of measured and predicted morphological features of ammonia spray at tASOI = 0.26 ms: (a) case PA0.1-PI100, (b) case PA0.5-PI100 and (c) case PA0.9-PI100.

studied under non-flash boiling conditions. For each subsection, the featuring the most intense flash boiling state. As can be seen from
spray liquid penetration length is first compared with the experiment Fig. 6, the predicted spray penetration length of case PA0.1-PI100 in
to quantitatively verify the simulation results. Then, the spray mor- the early stage of injection (before 0.15 ms) is higher than that of
phological characteristics under different conditions are shown, and the experiment. In fact, this phenomenon was also observed in case
the statistics at different axial locations are quantitatively analyzed. Fi- PA0.5-PI100 and case PA0.9-PI100. As the superheat level decreases,
nally, the phenomenon of particle preferential concentration for liquid the overpredicting of the spray penetration length in the early stage of
ammonia sprays is explored in terms of the Voronoi analysis. injection is reduced.
Recent numerical simulations [30] on liquid ammonia sprays show
4.1. Flash boiling spray under different ambient pressure conditions
that the modeling of flash boiling sprays tends to overestimate the spray
This subsection examines the effect of ambient pressure on the char- liquid penetration during the primary stage. Previous flash boiling
acteristics of superheated liquid ammonia spray. Fig. 6 demonstrates spray studies [43,44] indicate that flash boiling is a complex process
the comparison of measured and predicted spray liquid penetration including bubble nucleation, growth and breakup, the flow character-
length after start of injection (ASOI) of cases PA0.1-PI100, PA0.5-PI100 istics inside the nozzle are critical to the near-field spray structure.
and PA0.9-PI100. It can be seen that the results of the simulations Therefore, the present Lagrange-based spray model is not expected
are generally in good agreements with those of the experimental mea- to capture well the liquid ammonia spray characteristics in the near-
surements. The superheat level of case PA0.1-PI100 is the maximum, nozzle region. This indicates direct injection spray simulations under

5
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Fig. 8. The comparison of gaseous NH3 mass fraction (left) gas phase temperature(middle) and velocity (right) in different axial locations: 𝑧 = 5 mm, 𝑧 = 15 mm, 𝑧 = 30 mm at
𝑡𝐴𝑆𝑂𝐼 = 0.26 ms.

Fig. 9. Instantaneous distributions of gaseous ammonia mass fraction in a typical y-z plane at 𝑡𝐴𝑆𝑂𝐼 = 0.26 ms. The polygons denote the Voronoi diagram and the points indicate
liquid ammonia particles.

flash boiling conditions require more sophisticated models which can Fig. 8 shows a comparison of the gaseous NH3 mass fraction, gas
reasonably consider the various sub-processes of the spray. phase temperature and velocity profiles at different axial locations.
Fig. 7 shows a comparison of ammonia spray morphology at 𝑡𝐴𝑆𝑂𝐼 A lower ambient pressure corresponds to a higher superheat level,
= 0.26 ms for various cases. As can be seen, the overall spray mor- which induces an increase in the evaporation rate of the droplets.
phology can be qualitatively reproduced by the current simulations. Therefore, the gaseous NH3 mass fraction of case PA0.1-PI100 is always
Different ambient pressures cause the ammonia spray to be at different the maximum at different downstream locations. The distribution of
superheat levels, resulting in significant changes in spray morphology. the gas phase temperature is correlated with that of the gaseous NH3
Particularly, case PA0.1-PI100 with the highest superheat level features mass fraction. The higher evaporation rate indicates more intense heat
the largest penetration length, while case PA0.9-PI100 with the lowest transfer between the droplet and the environment, which results in a
superheat level has the smallest penetration length. lower gas phase temperature. Ammonia spray with a lower ambient

6
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Fig. 10. (a–c) Joint p.d.f.of gaseous ammonia mass fraction and 𝐴𝑐 ∕𝐴𝑐 , (e–f) the mean gas phase temperature 𝑇 conditionally averaged on gaseous ammonia mass fraction and
𝐴𝑐 ∕𝐴𝑐 for cases PA0.1-PI100, PA0.5-PI100 and PA0.9-PI100 at 𝑡𝐴𝑆𝑂𝐼 = 0.26 ms.

pressure is characterized by a higher velocity and more intensive


turbulence, which promotes the mixing of air and ammonia and leads
to a larger spray width, especially downstream of the spray. It is noted
that the magnitude of flash boiling evaporation rate term in Eq. (15)
dominates the total evaporation rate under the flash boiling conditions,
and the magnitude of this term decreases with decreasing superheat
level.
To understand the dispersive or clustering behavior of ammonia
spray, the Voronoi diagram is introduced. Voronoi diagram presents
a method of spatial data interpolation into polygons around each point
in such a way that each location from the area surrounding a given
point is closer to it than to any other point. The application of Voronoi
analysis can be found in a wide range of fields, including cell biology,
astronomy, medical diagnosis, and multiphase flows. In the context
of the current spray study, the area of a Voronoi cell, defined for a
Fig. 11. Comparison of measured and predicted spray liquid Penetration length at
node (i.e., location of a particle), depends on the separation distance different injection pressures.
relative to neighboring particles. Thus, a smaller Voronoi area refers
to a higher local droplet concentration. Accordingly, the calculation of
the Voronoi areas in a spray image can facilitate the identification of
4.2. Non-flash boiling spray under different injection pressure conditions
clusters of droplets. The method was detailed in [31]. Fig. 9 shows the
Voronoi diagram in a typical y-z plane of case PA0.1-PI100, which is
In this subsection, the spray characteristics of non-flash boiling
superimposed with the contours of gaseous NH3 mass fraction and gas
liquid ammonia spray under different injection pressures are explored.
phase temperature. It can be seen that the areas of the Voronoi cells
are small in the regions where particles are aggregated, which means Fig. 11 displays the comparison of measured and predicted spray liquid
the area of the Voronoi cell can be used to characterize the degree of penetration length of cases PA4-PI50, PA4-PI75 and PA4-PI100. It
particle preferential concentration. It is also shown that the particles can be seen that the predictions of the present simulations at various
clustering results in greater evaporation and therefore higher gaseous injection pressures are in good agreements with the experimental mea-
NH3 mass fractions, and consequently lower temperatures as a result of surement. The higher injection pressure leads to a longer penetration
the latent heat. length of ammonia spray. It is noted that compared to flash boiling
To explore the preferential concentration effects on the phase- ammonia spray, the penetration length of non-flash boiling spray in
change process of ammonia sprays, Fig. 10(a–c) show the joint proba- the initial injection stage increases more rapidly.
bility density functions (JPDF) of gaseous ammonia mass fraction and Fig. 12 shows a comparison of ammonia spray morphology at 𝑡𝐴𝑆𝑂𝐼
normalized areas of the Voronoi cells 𝐴𝑐 ∕𝐴𝑐 and Fig. 10(e–f) display the = 1.1 ms for various cases. It is clearly shown that spray morphology
gas phase temperature 𝑇 conditionally averaged on gaseous ammonia measured by the experiment can be reproduced well by the present
mass fraction and 𝐴𝑐 ∕𝐴𝑐 for various cases, where 𝐴𝑐 is average area simulations. The ammonia spray propagates more rapidly at higher
of the total Voronoi cells. It is found that the high value of JPDF injection pressures, which is consistent with the observations of the
corresponds to a small normalized area 𝐴𝑐 ∕𝐴𝑐 and high ammonia vapor experiment [38].
mass fraction. This implies that the concentration of ammonia vapor To explored the effects of injection pressures on non-flash boiling
is higher in the regions of particle preferential concentration. It is ammonia spray more quantitatively, Fig. 13 shows a comparison of the
clearly shown that the temperature of the gas phase is lower in the gaseous NH3 mass fraction, gas phase temperature and velocity profiles
regions of smaller 𝐴𝑐 ∕𝐴𝑐 , indicating that liquid ammonia evaporation at different axial locations for various cases. The gaseous NH3 mass
is significant in these regions. fraction of various cases is similar at z = 15 and 30 mm, while ammonia

7
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Fig. 12. Comparison of measured and predicted morphological features of ammonia spray at tASOI = 1.1 ms: (a) case PA4-PI50, (b) case PA4-PI75 and (c) case PA4-PI100.

Fig. 13. The comparison of gaseous NH3 mass fraction (left) gas phase temperature (middle) and velocity (right) in different axial locations: 𝑧 = 15 mm, 𝑧 = 30 mm, 𝑧 = 45 mm at
𝑡𝐴𝑆𝑂𝐼 = 1.1 ms.

concentration is higher for case PA4-PI100 with a higher injection those in flash boiling ammonia sprays. Again, it is shown that a lower
pressure at z = 45 mm. The distribution of gas phase temperature and gas phase temperature appears in the region of particle preferential
gaseous NH3 mass fraction correlates well. That is, the region with a concentration.
higher mass fraction of ammonia corresponds to a lower gas phase
temperature. It is noted that the gas phase temperature near the central 4.3. Non-flash boiling spray under different ambient pressure conditions
line of the non-flash boiling ammonia spray is much higher than that
of the flash boiling spray, which means that the cooling effects due The effect of ambient pressure on the characteristics of non-flash
to evaporation in the non-flash boiling ammonia sprays are not as
boiling ammonia spray is examined in this subsection. Fig. 15 displays a
significant as those in flash boiling ammonia sprays. It can be seen that
comparison of measured and predicted spray liquid penetration length
in the near-field, sprays with a higher pressure feature a higher gas
of cases PA4-PI100, PA3-PI100, PA2-PI100 and PA1-PI100. It can be
velocity. Further downstream, the gas phase velocity gradually reduces
seen that the predictions of the present simulations at various ambient
and the discrepancy between various cases becomes minor.
Fig. 14(a–c) show the joint probability density functions of am- pressures are in good agreements with the experimental measurement.
monia mass fraction and normalized areas 𝐴𝑐 ∕𝐴𝑐 and Fig. 14(e–f) It is found that the lower ambient pressure leads to a longer spray
display the gas phase temperature 𝑇 conditionally averaged on gaseous penetration length. Note that as the ambient pressure increases, the
ammonia mass fraction and 𝐴𝑐 ∕𝐴𝑐 for various cases. The results of density of the ambient air is large, which results in a larger drag force
JPDF for various cases are similar, which implies that the aggregation and resistance for spray propagation. Fig. 16 shows a comparison of
behavior of particles is not significantly affected by injection pressure. ammonia spray morphology at 𝑡𝐴𝑆𝑂𝐼 = 1.1 ms of various cases. For
The evaporation rate of liquid ammonia is higher in the regions of non-flash boiling ammonia sprays, varying ambient pressure affects the
particle preferential concentration. The cooling effects due to evapo- resistance of spray propagation, so that the ammonia spray propagates
ration of the non-flash boiling ammonia sprays are not as significant as more rapidly at a lower ambient pressure.

8
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Fig. 14. (a–c) Joint p.d.f.of gaseous ammonia mass fraction and 𝐴𝑐 ∕𝐴𝑐 , (d–f) the mean gas phase temperature 𝑇 conditionally averaged on ammonia mass fraction and 𝐴𝑐 ∕𝐴𝑐 for
cases PA4-PI50, PA4-PI75 and PA4-PI100 at 𝑡𝐴𝑆𝑂𝐼 = 1.1 ms.

is negatively correlated with 𝐴𝑐 ∕𝐴𝑐 , which means the concentration


of ammonia vapor is higher in the region of particle preferential
concentration. Case PA1-PI100 with a low ambient pressure shows
very low temperature in the region where the normalized Voronoi
area is small and the ammonia vapor mass fraction is high, while the
temperature decrease in the other cases is not as significant as that in
case PA1-PI100.

4.4. Comparison of flash boiling and non-flash boiling sprays

The flash boiling and non-flash boiling sprays are compared in


this sub-section. Specifically, case PA0.1-PI100 and case PA4-PI100 are
analyzed. Fig. 19 shows a comparison of flash boiling spray morphology
at 𝑡𝐴𝑆𝑂𝐼 = 0.26 ms and non-flash boiling spray morphology at 𝑡𝐴𝑆𝑂𝐼 =
1.1 ms. It is seen that there are evident differences between flash boiling
and non-flash boiling sprays, which are quantified below. Fig. 20
shows the scatter plots of droplet velocity versus the residence time
Fig. 15. Comparison of measured and predicted spray liquid penetration length at
different ambient pressures. for the two cases. The velocity of droplets conditionally averaged on
the residence time is also shown. The error bars correspond to the
conditional standard deviation. As can be seen, the flash boiling spray
is featured by a high initial velocity of the droplets, and the droplet
To gain further insights into the effect of ambient pressure on the
velocity decreases sharply as the residence time increases. In contrast,
characteristics of non-flash boiling ammonia sprays, Fig. 17 shows a
the droplet velocity of the non-flash boiling spray is lower initially
comparison of the gaseous NH3 mass fraction, gas phase temperature
and the variation of the droplet velocity as a function of the residence
and velocity profiles at different axial locations for various cases. It
time is more moderate. Additionally, the results of standard deviation
is clearly shown that the gaseous NH3 mass fraction reduces as the
ambient pressure rises. It is noted that the overall characteristics of case indicate that the flashing spray features a larger velocity fluctuations
PA1-PI100 show remarkable differences from the other three cases. The of the droplets, which results in a larger relative velocity between the
explanation behind this phenomenon is that case PA1-PI100 is close droplets and the surrounding fluid, and enhances the heat and mass
to the critical state between flash boiling and non-flash boiling, and a transfer processes between the gas and liquid phases.
much higher evaporation rate occurs in this case. The distribution of The evaporation rate is affected by the operating conditions. Fig. 21
the gas phase temperature and gaseous NH3 mass fraction correlates shows a comparison of scatter plots of the evaporation source terms
well. The higher mass fraction of ammonia indicates more intense heat plotted against the gas phase temperature between the two cases. It can
transfer between the droplet and the environment, which results in a be seen that the evaporation source term of the flash boiling spray is
lower gas phase temperature. Sprays with a lower ambient pressure much larger than that of non-flash boiling spray, which indicates that
correspond to a higher gas phase velocity. the evaporation of the liquid ammonia is more intensive in the flash
To understand the preferential concentration effects on the phase- boiling case.
change process of non-flash boiling ammonia sprays at various ambient Previous experimental studies on flash boiling ammonia sprays [25,
pressures, Fig. 18(a–d) show the joint probability density functions 38] have indicated that flash boiling sprays show evident spray ex-
of gaseous ammonia mass fraction and normalized Voronoi areas and pansion, which is analyzed quantitatively in the present work. Fig. 22
Fig. 18(e–h) display the gas phase temperature 𝑇 conditionally aver- shows the scatter plots of the radial distance of droplets versus the
aged on gaseous ammonia mass fraction and normalized Voronoi areas residence time for the two cases. The radial distance conditionally
for various cases. Again, it is found that gaseous ammonia mass fraction averaged on the residence time is also shown. The error bars correspond

9
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Fig. 16. Comparison of predicted morphological features of ammonia spray at tASOI = 1.1 ms: (a) case PA1-PI100, (b) case PA2-PI100, (c) case PA3-PI100 and (d) case PA4-PI100.

Fig. 17. The comparison of gaseous NH3 mass fraction (left) temperature (middle) and velocity (right) in different axial locations: 𝑧 = 5 mm, 𝑧 = 15 mm, 𝑧 = 30 mm at 𝑡𝐴𝑆𝑂𝐼 =
1.1 ms.

to the conditional standard deviation. As can be seen, the radial 5. Conclusions


distance of droplets in the flash boiling spray increases rapidly within a
small residence time of droplets, resulting in evident spray expansion. In the present study, high-pressure liquid ammonia injection under
In contrast, the radial expansion of the non-flash boiling spray is much flash boiling and non-flash boiling conditions were investigated using
weaker, consistent with Fig. 19. Large eddy simulation. The Eulerian and Lagrangian methods were
The radial expansion of the spray can be understood by analyzing employed for solving the gas and liquid phases, respectively. The effects
the dilation of the fluid. Fig. 23 shows the scatter plots of the fluid of ambient pressure on the spray characteristics of flash boiling liquid
dilation at the droplet location versus the gas phase temperature for ammonia were explored in detail, and the effects of injection pressure
the two cases. It can be seen that the fluid dilation of the flash boiling and ambient pressure on the ammonia spray were investigated under
spray case is larger than that of the non-flash boiling spray case, which the non-flash boiling condition. In addition, a comparison between
indicates that the fluid expansion is more obvious in the flash boiling flash boiling and non-flash boiling sprays was performed. The main
spray, which facilitates the radial expansion of the droplets. conclusions are summarized as follows:

10
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Fig. 18. (a–d) Joint p.d.f.of gaseous ammonia mass fraction and 𝐴𝑐 ∕𝐴𝑐 , (e–h) the mean gas phase temperature 𝑇 conditionally averaged on ammonia mass fraction and 𝐴𝑐 ∕𝐴𝑐
for cases PA1-PI100, PA2-PI100, PA3-PI100 and PA4-PI100 at 𝑡𝐴𝑆𝑂𝐼 = 1.1 ms.

Fig. 19. Comparison of flash boiling spray morphology at 𝑡𝐴𝑆𝑂𝐼 = 0.26 ms and non-flash boiling spray morphology at 𝑡𝐴𝑆𝑂𝐼 = 1.1 ms.

Fig. 20. The scatter plots of droplet velocity versus the residence time. The solid line corresponds to the velocity of droplets conditionally averaged on the residence time. The
error bars correspond to the conditional standard deviation.

11
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Fig. 21. Comparison of scatter plots of the evaporation source terms plotted against the gas phase temperature.

Fig. 22. The scatter plots of the radial distance of droplets versus the residence time. The solid line corresponds to the radial distance conditionally averaged on the residence
time. The error bars correspond to the conditional standard deviation.

Fig. 23. Comparison of scatter plots of the fluid dilation at the droplet location versus the gas phase temperature.

• The results of simulations are generally in good agreements with The higher injection pressure leads to a longer penetration length
the experimental measurements in terms of penetration length of ammonia spray. It is found that the penetration length of non-
and spray morphology. Different ambient pressures cause the flash boiling spray in the initial injection stage increases more
ammonia spray to be at different superheat levels, resulting in sig- rapidly compared to flash boiling ammonia spray.
nificant changes in spray characteristics. Gas phase temperature • The lower ambient pressure leads to a longer spray penetration
is lower in the regions of smaller normalized areas of the Voronoi length, and the ammonia vapor mass fraction reduces as the
cells, implying that evaporation preferentially occurs. ambient pressure rises. The higher ammonia vapor mass fraction
• Spray morphology under various injection pressure measured by indicates more intense heat transfer between the droplet and the
experiment can be reproduced well by the present simulations. environment, which results in a lower gas phase temperature.

12
Z. Huang et al. Applications in Energy and Combustion Science 16 (2023) 100220

Sprays with a lower ambient pressure correspond to a higher gas [15] Zhang R, Lin C, Wei H, Li J, Chen R, Pan J. Understanding the difference
phase velocity. in combustion and flame propagation characteristics between ammonia and
methane using an optical SI engine. Fuel 2022;324:124794.
• The flash boiling spray is featured by a high initial velocity of
[16] Li J, Zhang R, Pan J, Wei H, Shu G, Lin C. Ammonia and hydrogen blending
the droplets. In contrast, the droplet velocity of the non-flash effects on combustion stabilities in optical SI engines. Energy Convers Manage
boiling spray is lower and the variation of the droplet velocity as a 2023;280:116827.
function of the residence time is more moderate. The evaporation [17] Okafor E, Yamashita H, Hayakawa A, Somarathne KDK, Kudo T, Tsujimura T, et
source term of flash boiling spray is much larger than that of non- al. Flame stability and emissions characteristics of liquid ammonia spray co-fired
with methane in a single stage swirl combustor. Fuel 2020;287.
flash boiling spray. The radial distance of droplets in the flash [18] Gross C, Kong S-C. Performance characteristics of a compression-ignition engine
boiling spray increases rapidly within a short time after injection, using direct-injection ammonia–DME mixtures. Fuel 2013;103:1069–79.
resulting in evident spray expansion, while the radial expansion [19] Scharl V, Lackovic T, Sattelmayer T. Characterization of ammonia spray com-
of the non-flash boiling spray is much weaker. bustion and mixture formation under high-pressure, direct injection conditions.
Fuel 2023;333.
[20] Somarathne KDK, Yamashita H, Colson S, Okafor E, Hayakawa A, Kudo T,
Declaration of competing interest Kobayashi H. Liquid ammonia spray combustion and emission characteristics
with gaseous hydrogen/air co-firing. 2021.
The authors declare that they have no known competing finan- [21] Zeng W, Xu M, Zhang G, Zhang Y, Cleary D. Atomization and vaporization for
cial interests or personal relationships that could have appeared to flash-boiling multi-hole sprays with alcohol fuels. Fuel 2011;95.
[22] Wang B, Jiang C, Xu H, Badawy T. Microscopic characterization of isooc-
influence the work reported in this paper. tane spray in the near field under flash boiling condition. Appl Energy
2016;180:598–606.
Data availability [23] Ronan P, Mounaïm-Rousselle C, Brequigny P, Hespel C, Bellettre J. First study on
ammonia spray characteristics with a current GDI engine injector. Fuels 2021;2.
[24] Qiang C, Ojanen K, Diao Y, Kaario O, Larmi M. Dynamics of the ammonia spray
Data will be made available on request.
using high-speed schlieren imaging. 4, 2022,
[25] Li S, Li T, Wang N, Xinyi Z, Chen R, Ping Y. An investigation on near-field and
Acknowledgments far-field characteristics of superheated ammonia spray. Fuel 2022;324:124683.
[26] Zhang Y, Leilei X, Zhu Y, Xu S, Bai X-S. Numerical study on liquid ammonia
This work was supported by Natural Science Foundation of China direct injection spray characteristics under engine-relevant conditions. Appl
Energy 2023;334:120680.
(Grant Nos.: 52022091, 51976185) and the Fundamental Research [27] Li T, Xinyi Z, Wang N, Wang X, Chen R, Li S, et al. A comparison between
Funds for the Central Universities (Grant No. 2021FZZX001-10). low- and high-pressure injection dual-fuel modes of diesel-pilot-ignition ammonia
combustion engines. J Energy Inst 2022;102.
References [28] Yousefi A, Guo H, Dev S, Liko B, Lafrance S. Effects of ammonia energy fraction
and diesel injection timing on combustion and emissions of an ammonia/diesel
dual-fuel engine. Fuel 2022.
[1] Levinsky H. Why can’t we just burn hydrogen? Challenges when changing fuels
[29] Pandal A, Zembi J, Battistoni M, Hespel C, Ronan P, Brequigny P, et al. GDI
in an existing infrastructure. Prog Energy Combust Sci 2021;84:100907.
ammonia spray numerical simulation by means of OpenFOAM. 2023.
[2] Dimitriou P, Javaid R. A review of ammonia as a compression ignition engine
[30] An Z, Xing J, Kurose R. Numerical study on the phase change and spray
fuel. Int J Hydrogen Energy 2020;45.
characteristics of liquid ammonia flash spray. SSRN Electron J 2022.
[3] Cai T, Zhao D. Overview of autoignition and flame propagation properties for
[31] Aurenhammer F. Voronoi diagrams—a survey of a fundamental geometric data
ammonia combustion. AIAA J 2023;61:1–25.
structure. ACM Comput Surv 1991;23:345–405.
[4] Mercier A, Mounaïm-Rousselle C, Brequigny P, Bouriot J, Dumand C. Im-
[32] Wang Y, Cai R, Changxiao S, Luo K, Fan J. A priori and a posteriori studies of
provement of SI engine combustion with ammonia as fuel: Effect of ammonia
a novel spray flamelet tabulation methodology considering evaporation effects.
dissociation prior to combustion. Fuel Commun 2022;11:100058.
Fuel 2023;331:125892.
[5] Hayakawa A, Goto T, Mimoto R, Arakawa Y, Kudo T, Kobayashi H. Laminar
[33] Zuo B, Gomes A, Rutland C. Modelling superheated fuel sprays and vaproization.
burning velocity and Markstein length of ammonia/air premixed flames at
Int J Engine Res 2000;1:321–36.
various pressures. Fuel 2015;159.
[34] Adachi M, McDonell VG, Tanaka D, Senda J, Fujimoto H. Characterization of fuel
[6] Meng Z, An Z, Wei X, Wang J, Tan H. Emission analysis of the CH4/NH3/air
vapor concentration inside a flash boiling spray. SAE Technical Paper, 1997.
co-firing fuels in a model combustor. Fuel 2021;291:120135.
[35] Marchevsky I, Puzikova V. OpenFOAM iterative methods efficiency analysis for
[7] An Z, Meng Z, Zhang W, Runze M, Wei X, Wang J, et al. Emission predic-
linear systems solving. In: Proceedings of the institute for system programming
tion and analysis on CH4/NH3/air swirl flames with LES-FGM method. Fuel
of RAS. Vol. 24. 2013, p. 71–86.
2021;304:121370.
[36] Weller H, Tabor G, Jasak H, Fureby C. A tensorial approach to computa-
[8] Honzawa T, Kai R, Okada A, Valera-Medina A, Bowen P, Kurose R. Predictions
tional continuum mechanics using object orientated techniques. Comput Phys
of NO and CO emissions in ammonia/methane/air combustion by LES using a
1998;12:620–31. http://dx.doi.org/10.1063/1.168744.
non-adiabatic flamelet generated manifold. Energy 2019;186.
[37] Luo Y, Wen X, Wang H, Luo K, Fan J. Evaluation of different flamelet tabulation
[9] Hayakawa A, Arakawa Y, Mimoto R, Somarathne KDK, Kudo T, Kobayashi H.
methods for laminar spray combustion. Phys Fluids 2018;30:053603.
Experimental investigation of stabilization and emission characteristics of am-
[38] Fang Y, Ma X, Zhang Y, Li Y, Zhang K, Jiang C, et al. Experimental investigation
monia/air premixed flames in a swirl combustor. Int J Hydrogen Energy
of high-pressure liquid ammonia injection under non-flash boiling and flash
2017;42.
boiling conditions. Energies 2023;16:2843.
[10] Somarathne KDK, Hatakeyama S, Hayakawa A, Kobayashi H. Numerical study of
[39] Wehrfritz A, Vuorinen V, Kaario O, Larmi M. Large eddy simulation of high-
a low emission gas turbine like combustor for turbulent ammonia/air premixed
velocity fuel sprays: Studying mesh resolution and breakup model effects for
swirl flames with a secondary air injection at high pressure. Int J Hydrogen
spray A. Atomiz Sprays 2013;23(5).
Energy 2017;42:27388–99.
[40] Pandal A, Zembi J, Battistoni M, Hespel C, Pele R, Brequigny P, et al. GDI
[11] Syred N. A review of oscillation mechanisms and the role of the precessing vortex
ammonia spray numerical simulation by means of OpenFOAM. Tech. rep., 2023.
core (PVC) in swirl combustion systems. Prog Energy Combust Sci 2006;32.
[41] Salehi F, Ghiji M, Chen L. Large eddy simulation of high pressure spray with
[12] Meng Z, An Z, Wang L, Wei X, Jianayihan B, Wang J, et al. The regulation
the focus on injection pressure. Int J Heat Fluid Flow 2020;82:108551.
effect of methane and hydrogen on the emission characteristics of ammonia/air
[42] Kahila H, Wehrfritz A, Kaario O, Masouleh MG, Maes N, Somers B, et al. Large-
combustion in a model combustor. Int J Hydrogen Energy 2021;46:21013–25.
eddy simulation on the influence of injection pressure in reacting Spray A.
[13] Valera-Medina A, Marsh R, Runyon J, Pugh D, Beasley P, Hughes T, et al.
Combust Flame 2018;191:142–59.
Ammonia–methane combustion in tangential swirl burners for gas turbine power
[43] Wu S, Xu M, Hung DL, Pan H. In-nozzle flow investigation of flash boiling fuel
generation. Appl Energy 2016;185.
sprays. Appl Therm Eng 2017;117:644–51.
[14] Zhang R, Lin C, Wei H, Li J, Ding Y, Chen R, et al. Experimental investigation
[44] Zhang Y, Xu L, Zhu Y, Xu S, Bai X-S. Numerical study on liquid ammonia direct
on reactivity-controlled compression ignition (RCCI) combustion characteris-
injection spray characteristics under engine-relevant conditions. Appl Energy
tic of n-heptane/ammonia based on an optical engine. Int J Engine Res
2023;334:120680.
2022;24:146808742211244.

13

You might also like