Fuel 389 (2025) 134474

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Effects of ammonia and carbon dioxide concentration variation on

ammonia-diesel blended fuel combustion in CO2/O2 atmosphere
Yongfeng Liu a,* , Chenyang Yin a, Jiaying Pan b, Jin’ou Song b, Hua Sun c
a
Beijing Engineering Research Center of Monitoring for Construction Safety, Beijing University of Civil Engineering and Architecture, Beijing 100044, China
b
State Key Laboratory of Engines, Tianjin University, Tianjin 300072, China
c
School of Humanities, Beijing University of Civil Engineering and Architecture, Beijing 100044, China

A R T I C L E I N F O A B S T R A C T

Keywords: In a closed-cycle diesel engine (CCDE) with liquid oxygen, CO2 and O2 constitute the main components of the
Ammonia-diesel blended fuel gases within the cylinders. To investigate the effects of ammonia (NH3) and carbon dioxide (CO2) concentration
Quantum chemical variation on the combustion characteristics of ammonia-diesel blended fuel (ADBF) in CO2/O2 atmosphere, a
The effects of ammonia and carbon dioxide
new ammonia-diesel oxidation combustion (ADOC) model is proposed, which utilized quantum chemical
Rapid compression machine
calculation. Firstly, the structures of the reactants and transition states are optimized using the B3PLYP and CCSD
(T) methods, and the reaction mechanism is simplified using the DRGEP method. A simplified mechanism
comprising 348 species and 2168 reactions is obtained, which is integrated into a physical model in Converge to
simulate spray ignition and combustion processes. Secondly, rapid compression machine (RCM) platform is built
and different ammonia-diesel ratios (1:9, 2:8, 3:7 and 4:6) and CO2/O2 (35 %/65 %, 39 %/61 %, 43 %/57 %, 47
%/53 %, 51 %/49 % and 55 %/45 %) are measured. Furthermore, operational stability is verified according to
three top dead center pressures. Thirdly, reaction pathway for CO + HNO, orbit-based analysis of the Fukui
function, maximum in-cylinder pressures (MICPs), flame areas and propagation speeds are discussed between
simulation and experiment for different ammonia-diesel ratios and CO2/O2. The results showed that ADOC
model can effectively simulate MICPs of ADBF combustion under CO2/O2 atmosphere and maximum error is
6.53 % at 10 %NH3 + 90 %diesel (35 % CO2 / 65 % O2). END-2 is more likely to occurs than the formation of
END-1 in CO + HNO → NH + CO2. As the CO2 concentration increases (from 35 % to 43 %), both the flame
propagation speed and flame area significantly decrease. Conversely, with the increase in NH3 concentration, the
initial acceleration of flame propagation becomes more pronounced. Especially at a CO2 concentration of 43 %,
the flame propagation speed at an NH3 concentration of 30 % shows a substantial improvement compared to that
at 10 % NH3.

1. Introduction combustion while capturing and storing CO2 at low temperatures. Any
uncaptured CO2 is recirculated back into the combustion chamber
Ammonia emerges as a key focus of research due to its high energy through the exhaust gas recirculation tube, aiming to achieve zero car­
density and potential as a commercially viable, carbon-free hydrogen bon emissions. However, the intense combustion of diesel under high
carrier [1,2]. However, challenges arise from its low flame speed [3,4] concentrations of liquid oxygen can lead to problems like detonation. To
and high ignition temperature [5,6], which complicate ignition and mitigate this risk and improve combustion stability, ammonia-diesel
sustained combustion in internal combustion engines. To address these blended fuel (ADBF) is introduced as a solution.
limitations, ammonia is often blended with other fuels, such as Recent advancements in ammonia-diesel dual-fuel combustion can
hydrogen, diesel, and methane, to improve its combustion characteris­ be categorized into three main areas: ammonia concentration, com­
tics [7,8]. In recent years, the closed-cycle diesel engine (CCDE), which bustion stability, and emission control. Studies focusing on maximizing
uses liquid oxygen, attracts considerable attention [9,10]. The CCDE ammonia substitution, such as those by Nadimi et al. [11,12], demon­
operates by using liquid oxygen as the oxidizing agent to support diesel strate that ammonia can replace a significant portion of diesel’s energy

* Corresponding author.
E-mail address: [email protected] (Y. Liu).

https://doi.org/10.1016/j.fuel.2025.134474
Received 18 November 2024; Received in revised form 6 January 2025; Accepted 21 January 2025
Available online 10 February 2025
0016-2361/© 2025 Published by Elsevier Ltd.
Y. Liu et al. Fuel 389 (2025) 134474

input (up to 84.2 %), while improving indicated thermal efficiency focusing on ignition delay times and updating the detailed chemical
(ITE). Li et al. [13] extended this further, achieving 97 % diesel mechanism. The study found that ammonia is primarily consumed by O,
replacement in marine engines with high-pressure injection systems, H, and OH radicals, producing NH2 radicals. Some of the NH2 radicals
highlighting the potential for substantial reductions in CO2 and NH3 are converted to H2NO, which is further oxidized to HNO, eventually
emissions. In terms of combustion stability, Cai et al. [14] successfully leading to NO formation. HNO plays a crucial role as an intermediate in
implemented port fuel injection for ammonia combined with direct NH3 combustion and is a key pathway in NOx formation.
diesel injection in heavy-duty engines, showing that increasing Despite the significant advancements in these studies, the specific
ammonia concentration delays ignition and extends combustion dura­ reaction mechanisms of ADBF combustion have not been fully investi­
tion, but maintains stable operation. Similarly, Xu et al. [15] explored gated, especially in CCDE where the CO2 concentration is significantly
reaction controlled compression ignition (RCCI) for marine engines, higher than that in conventional air. This circumstance renders the
achieving up to 24 % diesel usage while reducing greenhouse gas impact of CO2 on ADBF combustion particularly significant. Conse­
emissions by 70 %, although they reported increases in CO and CO2 quently, there is an urgent need to further investigate how CO2 affects
emissions. Emission control has also been a key focus. Zhou et al. [16] ADBF combustion and to explore its interactions with ammonia. The
demonstrated that in low-speed marine engines, both premixed and present study introduces a novel Ammonia-Diesel Oxidative Combustion
high-pressure spray combustion modes can achieve ITEs exceeding 50 (ADOC) model designed to overcome the limitations of existing models.
%, with the latter significantly lowering NOx emissions compared to Distinct from previous investigations, the ADOC model accounts for the
diesel-only operation. Reiter and Kong [17] showed the feasibility of unique interplay between CO2 and ammonia combustion, offering a
using ammonia in a multi-cylinder turbocharged engine, while Yang more precise depiction of ADBF combustion behavior. The ADOC model
et al. [18] found that increasing ammonia energy concentration reduced captures intricate reaction pathways, including the role of HNO as an
CO and CO2 emissions but led to higher NH3, N2O, NO, and NO2 emis­ intermediate and the facilitative effect of CO2 on the generation of OH
sions, indicating a trade-off between improved efficiency and emission radicals. In this study, quantum chemistry, widely applied in combus­
control. tion research [30–33], is used to calculate the reaction between CO
From the studies mentioned above, it is evident that ADBF com­ (produced from CO2 pyrolysis) and HNO. The new pathway is incor­
bustion can reduce CO2 emissions to some extent. However, the specific porated into the mechanism to examine the effect of CO2 on the com­
reaction mechanisms have not been sufficiently explored. Additionally, bustion characteristics of ADBF.
in CCDE, the CO2 concentration is much higher than in conventional air,
making its impact significant and warranting further investigation into 2. ADOC model
how CO2 influences ADBF combustion. Previous research has primarily
focused on CO2 inhibiting the combustion of hydrocarbon fuels [19]. For 2.1. Quantum chemical calculation
example, Li et al. [20] observed that the reaction rate of CH4/CO2
mixtures was lower compared to that of CH4/air flames. Dong et al. [21] Density-functional calculations are performed using Gaussian 16.
found that CO2 inhibits the combustion of n-heptane, with the laminar The structure optimization and frequency calculation for the initial
burning velocity, adiabatic flame temperature, and net heat release rate reactant, CO, are carried out with the B3PLYP functional [34]. For the
decreasing as the CO2 doping ratio increases. Additionally, the peak open-shell radical HNO, structure optimization and frequency calcula­
values of certain intermediates are reduced, and both the physical and tions are conducted using the UB3PLYP functional [35–37], along with
chemical effects of CO2 contribute to inhibiting the combustion process, Grimme’s D3 dispersion correction [38] with Becke-Johnson damping
with the physical effects being more prominent. Zhang et al. [22] [39]. The 6–311++G (d, p) basis set is employed for all calculations
studied the impact of CO2 in air on the spontaneous combustion of n- [40–42]. To calculate single-point energies with higher accuracy, the
heptane by analyzing the ignition delay time of n-heptane/air mixtures coupled-cluster method CCSD(T) [43–46] is applied, where HNO is
at temperatures of 1050–1400 K, pressures of 2 and 10 atm, and treated in its open-shell form, UCCSD(T) [47], and the cc-pVQZ basis set
equivalence ratios of 1.0 and 0.5. The results indicated that CO2 natu­ is used for both HNO and the base group [48]. The level of calculation
rally inhibits the combustion of n-heptane, with its thermal effect being for transition states (TS) and intrinsic reaction coordinate (IRC) is kept
the primary factor in suppressing ignition. However, some studies have consistent with the aforementioned methods.
shown that CO2 can promote hydrocarbon combustion under certain The initial reactants, CO and NHO, are structurally optimized to
conditions. For instance, Liu et al. [23] captured images of flame determine their energy-minimum conformations. Following that, reac­
propagation at four different CO2 concentrations (air, 50 % CO2 + 50 % tive site predictions are made for CO and NHO, with the orbital weight-
O2, 43 % CO2 + 57 % O2, and 35 % CO2 + 65 % O2) in a constant volume based Fukui function and dual descriptors employed for CO. In molec­
combustion chamber (CVCC) and developed a QC model. Their findings ular systems with higher-order point group symmetry, frontier molec­
revealed that CO produced by CO2 pyrolysis at high temperatures en­ ular orbitals may exhibit energy-simplex or quasi-simplex properties,
hances the generation of OH radicals, thereby promoting combustion. which can limit the accuracy of conventional Fukui functions and dual
The effect of CO2 on hydrocarbon fuel combustion has been exten­ descriptors in these contexts. Pino-Rios [49] introduced the concept of
sively studied, but its influence on NH3 combustion remains less orbital weight Fukui functions and subsequently developed orbital
explored. Nevertheless, Ponnuthurai et al. [24] found that the reaction weight Fukui functions and dual descriptors by assigning weights to
between NH radicals and CO2 plays a significant role in determining NOx each molecular orbital. fw+ (r) describes the nucleophilicity of the mole­
formation rates under specific conditions. The academic community has cule, where the subscript represents the orbital weight, calculated as
amassed a wealth of research [25,26] on the combustion of NH3. Zhang shown in Eq. (1):
et al. [27] investigated the reaction pathways of NH3 at an equivalence
∑
∞
ratio of 1.0–1.1 and identified three primary routes for the conversion of fw+ (r) = wi |φi (r) |2 (1)
NH3 to NO: NH3 → NH2(NH) → HNO → NO, NH3 → NH2 → NH → NO, i=LUMO

and NH3 → NH2 → NH → N → NO. The pathway involving HNO as an

intermediate was found to be the predominant route for NO production. Where wi the track weight and its expression are as follows:
Wei et al. [28], through numerical simulations, studied the impact of [ ( ) ]
2
NH3 addition on NO formation in CH4 premixed combustion and re­ exp − μ−Δεi
ported that the role of the HNO pathway in NO formation increased with wi = [ ( ) ]
2
(2)
∑∞
i=LUMO exp −
μ− εi
higher NH3 concentrations. Dai et al. [29] investigated NH3 ignition in Δ
the temperature range of 1040–1210 K at pressures of 20–75 bar,

2
Y. Liu et al. Fuel 389 (2025) 134474

fw− (r) is the orbital weight Fukui function describing the electrophilicity specie B would cause the direct error in specie A’s productivity to sur­
of the molecule, as shown in Eq. (3): pass the set threshold, meaning specie B must be retained, which is
repeated to filter and retain all essential species and reactions in the
HOMO
∑ mechanism.
fw− (r) = wi |φi (r) |2 (3)
Since the DRG method considers only direct relationships between
i
species and overlooks the weakening of inter-species correlations as they
[ ( ) ]
2 propagate along reaction pathways, the resulting simplified mechanism
exp − μ−Δεi is often still too large for practical use. To address this, the DRGEP
wi = [ ( ) ]
2
(4) method, an extension of the DRG method, is employed in this paper.
∑HOMO
i exp − μ−Δεi DRGEP accounts for the effect of pathway length on species interactions.
In DRGEP, once specie A is determined to be retained, the total contri­
fw0 (r) is the orbital weight Fukui function describing the reaction of the bution to specie A can be calculated by introducing the ’R’ value for all
molecule with the radical.Δfw (r) is the orbit weight double descriptor species reachable from specie A. The corresponding formula is provided
and they are computed as shown in Eqs. (5) and (6) in the following equation:
[ + ] {* }
f (r) + fw− (r) RA (B) = maxy rAB (9)
fw0 (r) = w (5)
2
n− 1
∏
Δfw (r) = fw+ (r) − fw− (r) (6) rAB
*
= rx rx+1 (10)
x=1

The method described by document [50] for calculating the orbital Where, RA (B) denotes the total error caused by deleting specie B to
weight Fukui function and orbital weight dual descriptor is applicable specie A; y denotes all reaction paths between specie A and specie B; x
only to closed-shell systems. For the HNO radical, the general form of the
denotes the x-th reaction; n is the total number of reactions; and rAB *
Fukui function and dual descriptors was used. The most reactive sites
denotes the direct error caused by deleting specie B to specie A in a
between CO and HNO were identified, and the conformation at this site
certain path.
was used as the initial structure for the first transition state (TS1). An
In Eq. (9), the value is expressed as the product of direct errors be­
intrinsic reaction coordinate (IRC) scan of the optimized structure was
tween all species along the current pathway. According to Eq. (10), if
then conducted to determine the highest point on the potential energy
specie A is retained, specie B must also be retained if there is at least one
surface, as well as the associated reactants and products. The same
path from specie A to specie B with an ’R-value’, denoted as RA (B), that
approach was applied for the subsequent TS2 calculations.
exceeds the threshold. Finally, sensitivity analysis is applied to the
simplified mechanism generated by the DRGEP method to remove spe­
2.2. Mechanism construction cies and reactions that are closely coupled with retained species but have
low significance to the validation objective. The calculations are pro­
The reaction processes of n-C7H16 and NH3 in CO2/O2 atmosphere vided below:
are simulated using a 0-dimensional closed homogeneous reactor in dϕ
CHEMKIN PRO, under conditions matching the experimental setup. The = F[ϕ, t, v] (11)
dt
NUIG model [51], which encompasses all fundamental reactions of n-
alkanes at both low and high temperatures, is selected. Due to the high ∂ϕ
wi = (12)
computational demands, the detailed mechanism is simplified. The ∂vi
mechanism is simplified using the Direct Relationship Graph Based on
Where ϕ is the target parameter such as species molar fraction, IDT,
Error Propagation (DRGEP) method [52], which uses the molar con­
temperature distribution; t is the time in s, and is the corresponding
centration of key species as the validation target for the model, and the
chemical reaction rate in s− 1; i denotes the i-th primitive reaction; and is
error threshold is set at 10 %. The final simplified ADOC mechanism
the sensitivity coefficient.
includes 348 species and 2168 reactions. The simplification method
In Eq. (12), when the sensitivity coefficient is > 0, it indicates that
used is as follows: based on the Direct Relationship Graph (DRG) method
the i-th primitive reaction is promoted, and when the sensitivity coef­
[53], direct relationship graphs for all species in the detailed mechanism
ficient is < 0, it indicates that the i-th primitive reaction is inhibited. The
are constructed. Unimportant species are removed from the mechanism
logarithmic form of Eq. (12) indicates the extent to which the removal of
according to the defined validation target and threshold. The concept of
a reaction affects the target parameter.
direct error is introduced, defined by the contribution value of the
change in yield of specie A due to the removal of specie B. The formula is
provided in Eq. (7) below: 2.3. Physical model
∑n ⃒⃒ i ⃒
i=1 vA ωi δB
i ⃒
rAB = ∑n i (7) The reaction paths calculated in Section 2.1 are coupled with the
i=1 |vA ωi |
simplified mechanism obtained in Section 2.2 to obtain an ADOC
{ mechanism containing 348 species and 2168 reactions. The ADOC
1, If species B participates in reaction i
δiB = (8) mechanism is imported into Converge, and a physical model with the
0, others
same dimensions as the rapid compression machine (RCM) experiment
where i denotes the i-th reaction; n is the total number of reactions; viA was built. The rapid compression machine physical model adopts an
is the stoichiometric coefficient of species A in the i-th reaction; ωi is the adaptive mesh, which saves computational time by having fewer meshes
chemical reaction rate of the i-th reaction, mol/(L-s); and is the corre­ in the cylinder during compression, and the mesh in the cylinder auto­
lation coefficient. matically becomes denser during combustion to ensure computational
In Eq. (7), the denominator represents the total absolute contribution accuracy.
to the productivity of species A from all reactions involving species A in As shown in Fig. 1, the left side is the compressed and encrypted
the mechanism, while the numerator represents the absolute contribu­ mesh image, and the right side of Fig. 1 shows the flow distribution in
tion to the productivity of specie A from reactions involving specie B. If the cylinder, it is found that as the diesel is sprayed out from the right
the direct error, rAB , exceeds the threshold, it indicates that removing side injector, the gas in the cylinder is disturbed rapidly and two

3
Y. Liu et al. Fuel 389 (2025) 134474

supply air to the high-pressure storage tank; then, the air compressor is
used to compress air into the high-pressure storage tank as the driving
force; finally, when the oil pressure in the brake cylinder is relieved, the
compressed air pushes the driving piston, which finally moves to the
brake piston by means of the intact piston mechanism during the
mixture compression stage. During the mixture compression stage, the
piston rapidly compresses the mixture into the combustion chamber.
The RCM platform used in the experiment is equipped with a 40 mm
thick quartz window, capable of withstanding high pressures and
resisting the maximum experimental pressure (over 20 MPa). A high-
speed camera, the Photron SA-Z (with a 105 mm diameter AF Micro
Nikkor 1:2.8D lens), is used to photograph the flame through the quartz
window. The camera operates at a frequency of up to 220,000 fps and a
shutter speed of 3.15 μs, allowing for precise capture of flame front
Fig. 1. RCM model.
evolution. The aperture of the fuel injector is 0.18 mm, and the injection
pressure is 80 MPa. The basic parameters of the fast compression system
are provided in Table 2.
vortexes are formed with the center of the spray as the axis of symmetry,
The gases used in this experiment are NH3, CO2 and O2 with purities
which can help to see the direction of the flow and the magnitude of the
of 99.9999 %, 99.9999 %, and 99.999 % respectively. The experimental
velocity in the cylinder very well. On the right side of the cylinder is the
procedure is described as follows: first, a vacuum pump is used to
injector, in which the spray turbulent diffusion uses the O’Rourke
evacuate the cylinder, CO2, NH3 and O2 are proportioned according to
model, the droplet evaporation uses the Frossling model, and for the
the principle of Dalton’s Law of Partial Pressure and introduced into the
high Reynolds number case such as diesel spray, the KH-RT model is
cylinder for pre-mixing and heating, and then the piston is controlled to
chosen. Considering the computational accuracy and computational
compress the cylinder. As shown in Fig. 2, three sets of cylinder pressure
cost, the large eddy simulation (LES) is chosen for RCM model, which
curves are measured prior to the start of the experiment, and their
can accurately simulate the physicochemical processes in spray ignition
maximum pressures are all around 2.2 MPa. Therefore, after considering
and combustion. During the combustion process, the SAGE reactor is
the effect of the signal transmission delay, we decide to use 2 MPa as the
used to solve the multi-step reaction mechanism in a zero-dimensional
threshold for the injection trigger pressure, so that the diesel fuel can be
homogeneous environment by calculating the reaction rate of each
injected when the piston reaches top dead center.
step of the reaction, which is directly coupled with the CFD solver and
Before starting the formal experiment, three test runs are conducted
bring into the component transport equations for solving. The specific
under the conditions of an NH3-to-diesel energy ratio of 4:6, a CO2-to-O2
parameters of the RCM model are shown in Table 1.
molar ratio of 7:3, and an equivalence ratio of 0.667, as shown in Fig. 2.
Since the fuel and gas components remain the same and no ignition
3. Experiment
occurs under these conditions, the stability of RCM platform can be
tested more efficiently and safely. Fig. 3 shows that the top dead center
Fig. 2 shows the schematic structure of RCM, which consists of five
pressures and rates of pressure change in the three trials were nearly
main parts: high-pressure gas storage tank, pneumatic drive chamber,
identical. Combined with previous results from this apparatus, the un­
hydraulic damping chamber, compression chamber and combustion
certainty is estimated to be ± 20 %.
chamber. The basic operating principle is as follows: firstly, hydraulic oil
In the formal experiment, four ammonia-diesel ratios (10 %NH3 +
at an oil pressure of about 10 MPa is injected into the hydraulic brake
90 %diesel, 20 %NH3 + 80 %diesel, 30 %NH3 + 70 %diesel, 40 %NH3 +
cylinder to hold the brake piston; secondly, an air compressor is used to
60 %diesel) and six CO2/O2 ratios (35 %CO2/65 %O2, 39 %CO2/61 %
O2, 43 %CO2/57 %O2, 47 %CO2/53 %O2, 51 %CO2/49 %O2, 55 %CO2/
Table 1 49 %O2) are set. Flame development images are obtained using a high-
Spray calculation specific parameters. speed camera, and cylinder pressure curves are recorded through pres­
Discription Parameter Value sure sensors. Since the maximum in-cylinder pressure (MICP) can
approximately reflect the intensity of fuel combustion, particular
O’ Rourke Film Splash model Critical value for splashing 3330.0
​ Fraction splashed 0.7
attention in this study is given to discussing the MICPs under different
​ Rebound weber number 5.0 operating conditions. The specific experimental conditions are shown in
​ Separation constant 3.0 Table 3.

KH create child parcels Fraction of injected mass/parcel 0.05 4. Results and discussion
​ Shed mass constant 0.1
Model size constant 0.6
​
4.1. Reaction pathway for CO + HNO
​ Model velocity constant 0.188
​ Model breakup time constant 7.0
At temperatures exceeding 2000 ◦ C, carbon dioxide will decompose
to form carbon monoxide. The background gas in this study contains
RT model without breakup length Model breakup time constant 1.0
​ Model size constant 0.1 significantly higher levels of CO2 than would be found in a conventional
​ Model breakup length constant 99999.0 combustion environment, such as air. Consequently, the decomposition
effect of CO2 is a crucial factor that must be considered in this study, and
Discharge coefficient model Use correlation for Cv 1.0 the chemical effect of CO is correspondingly more significant. HNO is a
Injection Temporal type Sequential crucial free radical in the combustion of ammonia, and its high reactivity
​ Temperature 380 K makes it a priority for this study.
Li [54] performed structural optimization of CO and HNO using the
Nozzle Nozzle diameter 0.00018 m MP2 method. However, considering the insufficient accuracy of the MP2
​ Circular injection radius 4.5 × 10-5 method, the structures of CO and HNO are optimized using B3PLYP and
Spray cone angle 20.0 deg
UB3PLYP, respectively. As illustrated in Fig. 4, the H atom of HNO
​

4
Y. Liu et al. Fuel 389 (2025) 134474

Fig. 2. Rapid compression machine.

absorbs energy, which gradually decreases the C-N-O bond angle. This
Table 2
results in the formation of transition state TS2-2, with a relative energy
Basic parameters of RCM.
of 80.132 kcal/mol. As the C-N-O bond angle continues to decrease, the
Basic characteristic parameters Numerical value O, which were originally connected to the N, are attracted by the C to
Combustion chamber piston diameter/mm 75 form new C-O bonds. This process culminates in the formation of the
Drive cylinder piston diameter/mm 240 final product, END-2, which has a relative energy of 59.527 kcal/mol.
Clearance/mm 12
The energy required for MID in Path 1 is 27.9 kcal/mol, which is
Compression ratio 20
Driving air pressure/MPa 1
considerably higher than that required for Path 2. Consequently, the
Compression time/ms 40 probability of the reaction occurring through Path 2 is higher, with an
energy barrier of 6.986 kcal/mol. Furthermore, the end product END-2
in Path 2 is also capable of further cleavage reactions in high tempera­
attacks the C atom of CO, resulting in the formation of a saddle point on ture environments. At a temperature of 2000 K, the C-N bond in END-2 is
the conformation potential energy surface, is taken as the first transition broken, resulting in the spontaneous cleavage of the molecule into an
state (TS1) of the reaction. The C-N–H angle is 104.56◦ , the O-C-N angle iminium radical (NH) and CO2. The potential energy and reaction rate
is 131.56◦ , and the relative energy is 77.015 kcal/mol. As the distance are presented in Table 4. It can be noticed that as the temperature in­
between the C and N decreases, the C-N double bond is formed. This is creases, the energy barrier of the reaction decreases gradually and the
accompanied by a change in the bond length of the N-O bond, which rate becomes correspondingly large. In summary, within a cylinder with
initially has a length of 1.24 Å but subsequently increases to 1.35 Å. a high concentration of CO2, it is highly likely that the pyrolysis of CO2
Concurrently, the C-N–H angle increases from 104.56◦ to 121.18◦ , while will produce a large amount of CO. Subsequently, CO reacts with HNO to
the O-C-N angle increases from 131.56◦ to 174.74◦ . The relative energy form CO2 and NH radicals, as represented by CO + HNO = CO2 + NH.
of this intermediate state which is called MID is 73.146 kcal/mol. This process can be carried out to some extent in a cyclic manner,
A search for MID yield two potential pathways. Path 1: MID absorbs slightly promoting the combustion of ADBF.
energy, and the H originally connected to the N gradually move closer to
the C, during which the transition state TS2-1 is formed. The relative
4.2. Orbit-based analysis of the Fukui function
energy of TS2-1 is 101.046 kcal/mol. Following the alteration of the
bond lengths and angles, the H form a new C–H bond with the C,
4.2.1. The analysis of fw0 for CO
resulting in the formation of the final substance, END-1. Path 2: MID
A Fukui function based on orbital weights was employed to analyze

5
Y. Liu et al. Fuel 389 (2025) 134474

Fig. 3. Cylinder pressure during test runs.

and predict the optimal location for nucleophilic or electrophilic re­ 4.2.2. The analysis of Δfw for CO
agents to attack the molecule. Fig. 5(a) depicts the visualization of fw0 Fig. 6(a) depicts the visualization of Δfw with an isosurface of 0.03 a.
with an isosurface of 0.03 a.u.. The green region indicates positive u., with positive values in green and negative values in blue. Δfw can
values. The symbol fw0 is employed to describe the capacity of a molecule indicate both nucleophilic and electrophilic reaction sites. In general,
to act as both an electron acceptor and a donor in a free radical reaction. the greater the positive value of Δfw , the more likely the site is to be
The value indicates the tendency of a site in a molecule to gain and lose nucleophilic; conversely, the greater the negative value of Δfw , the more
electrons simultaneously. The higher the value, the more active the site likely the site is to be electrophilic. It can be observed that the green area
is in the free radical reaction. It can be demonstrated that the region offw0 at the C-atom end of CO is considerably larger than that at the O-atom
= 0.03 a.u. at the C-atom end of CO is considerably larger than the O- part, which indicates that electrophilic reactions are more likely to occur
atom portion. This indicates that the C-atom end is more susceptible to at the C-atom end. As illustrated in Fig. 6(b), which depicts the distri­
pro-radical reactions. As illustrated in Fig. 5(b), which provides a more bution of contours in greater detail, the red area represents the region
detailed representation of the contour distribution, the red regions with a positive Δfw − value, while the blue area represents the region
indicate areas where fw0 is positive, while the blue regions indicate re­ with a negative Δfw − value. Fig. 6 shows an increased presence of red
regions near the O-atom, indicating higher nucleophilic activity. In
gions where fw0 is negative. From Fig. 5(b), it can be observed that there
contrast, the blue regions near the lower end of the C atom suggest that
are a greater number of red regions in proximity to the carbon atom,
the C atom is more active in electrophilic reactions.
suggesting that these atoms exhibit heightened activity in free radical
Similarly, quantitative analyses were conducted for Δfw of CO, as
reactions. This suggests that the carbon atom may serve as a potential
illustrated in Fig. 6(c). The red spheres represent the exceptionally large
key reaction site in free radical reactions. This hypothesis could explain
values, while the blue spheres represent the exceptionally small values.
the site of the CO and HNO radical reactions in section 4.1, which appear
The larger values indicate that the location is more active in nucleophilic
to be on and below the C-atom.
reactions, while the smaller values indicate that the location is more
Furthermore, the fw0 of CO was quantitatively analyzed as illustrated
active in electrophilic reactions. At the apex of the O and C atoms, two
in Fig. 5(c). The red spheres represent the very large values, while the
particularly diminutive values are observed, namely − 0.0006267 a.u.
blue spheres represent the very small values. The larger values indicate
and − 0.0048134 a.u., respectively. Of these, the latter is more modest in
that the position is more active in the free radical reaction. At the top of
magnitude and is situated on the C-atom. This suggests that electrophilic
the O-and C-atoms, there are two extremely small values (0.0003269 a.
reactions are more probable on the C-atom. The upper part of the C-atom
u. and 0.0024190 a.u., respectively), which indicate that this position is
and the middle part of the O-atom exhibit extremely large value points
unlikely to undergo free radical reactions. In the lower half of the C-
(0.0018036 a.u. and 0.0007383 a.u., respectively). It can be demon­
atom and in the upper half of the O-atom, there are points of great
strated that the absolute value of the minima at the top of the C atom is
magnitude (0.0039529 a.u. and 0.0006145 a.u., respectively). The
considerably larger than the maxima in the middle of the C-atom. This
greater magnitude on the C-atom indicates that the lower half of the C-
numerical evidence suggests that the C-atom position of CO is more
atom has a greater ability to gain and lose electrons and is more likely to
susceptible to electrophilic reactions. In conclusion, the reaction be­
undergo a pro-radical reaction.
tween CO and HNO radicals indicates that CO is more likely to gain

6
Y. Liu et al. Fuel 389 (2025) 134474

Table 3
specific experimental conditions.
CO2/O2 Ammonia- Molar mass Molar Equivalence Injection
content diesel ratios of mass of ratio duration/
（energy ammonia diesel μs
ratio） gas fuel

35 % 1:9 1.02 E-3 2.38 E- 1.5 4093

4
​ 2:8 1.99 E-3 2.07 E- 1.5 3588
4
​ 3:7 2.85 E-3 1.73 E- 1.5 3036
4
​ 4:6 3.66 E-3 1.43 E- 1.5 2550
4

39 % 1:9 0.95 E-3 2.22 E- 1.5 3836

4
​ 2:8 1.85 E-3 1.93 E- 1.5 3360
4
​ 3:7 2.71 E-3 1.65 E- 1.5 2903
4 Fig. 4. CO + HNO reaction path and potential energy surface.
​ 4:6 3.53 E-3 1.38 E- 1.5 2464
4
Table 4
43 % 1:9 0.90 E-3 2.12 E- 1.5 3665 Relative potential energy and reaction rate of END-2 → NH + CO2.
4
Temperature Relative potential Energy barrier Reaction rate /k
2:8 1.72 E-3 1.79 E- 1.5 3131
(kcal*mol− 1) (s− 1*M− 1)
​
(K) energy
4
(kcal*mol− 1)
​ 3:7 2.53 E-3 1.54 E- 1.5 2725
4 END22 CO2 +
​ 4:6 3.30 E-3 1.29 E- 1.5 2321 NH
4
1000 32.89 42.27 9.38 1.54E + 13
1100 35.64 42.87 7.23 7.67E + 13
47 % 1:9 0.81 E-3 1.91 E- 1.5 3322 1200 38.37 43.46 5.08 2.96E + 14
4 1300 41.08 44.03 2.95 9.34E + 14
​ 2:8 1.63 E-3 1.69 E- 1.5 2979 1400 43.77 44.60 0.83 2.52E + 15
4 1500 46.44 45.16 − 1.28 5.99E + 15
​ 3:7 2.40 E-3 1.45 E- 1.5 2591 1600 49.09 45.71 − 3.38 1.28E + 16
4 1700 51.72 46.26 − 5.46 2.52E + 16
​ 4:6 3.12 E-3 1.22 E- 1.5 2207 1800 54.34 46.80 − 7.54 4.62E + 16
4 1900 56.94 47.33 − 9.61 7.97E + 16
2000 59.53 47.86 − 11.67 1.31E + 17
2100 62.10 48.39 − 13.71 2.05E + 17
51 % 1:9 0.77 E-3 1.80 E- 1.5 3150 2200 64.66 48.91 − 15.75 3.08E + 17
4
​ 2:8 1.54 E-3 1.60 E- 1.5 2826
4
pressure trough forms, around 7.6 MPa. This phenomenon may be due to
​ 3:7 2.26 E-3 1.37 E- 1.5 2458
4 the dilution effect of CO2, which affects the combustion process, sup­
​ 4:6 2.94 E-3 1.15 E- 1.5 2093 pressing combustion intensity and causing a pressure drop. As the CO2
4 concentration increases to 55 %, the in-cylinder pressure drops again to
about 6.5 MPa, indicating that a high CO2 concentration has a signifi­
55 % 1:9 0.72 E-3 1.69 E- 1.5 2979 cant inhibitory effect on combustion, possibly because excessive CO2
4 reduces the available oxygen concentration, leading to incomplete
2:8 1.40 E-3 1.46 E- 1.5 2598
​
combustion and a reduction in cylinder pressure. The increase in NH3
4
​ 3:7 2.04 E-3 1.23 E- 1.5 2236 concentration (from 10 % to 40 %) results in a significant decrease in the
4 MICP. Specifically, when the NH3 concentration is 40 %, the MICP drops
​ 4:6 2.67 E-3 1.04 E- 1.5 1921 to around 5.8 MPa. This could be because ammonia has a lower calorific
4 value and a slower combustion reaction, resulting in incomplete com­
bustion and thus affecting the generation of in-cylinder pressure. The
unpaired electrons from HNO radicals. study by Xu et al. [55] also mentions that the low heating value and
relatively slow combustion rate of NH3 can significantly inhibit the
combustion process. At low NH3 concentrations (10 %), the in-cylinder
4.3. Maximum inside cylinder pressures pressure is higher, especially when the CO2 concentration is 43 %, where
the in-cylinder pressure reaches its peak at about 8.67 MPa. This in­
Fig. 7 illustrates the effects of different CO2 and NH3 concentrations dicates that at low NH3 concentrations, the combustion efficiency is
on MICP. The MICP exhibits a clear nonlinear trend with changes in CO2 higher, the flame propagation speed is faster, and the pressure genera­
and NH3 concentrations. From the Fig., it can be seen that as the CO2 tion capacity is stronger.
concentration increases (from 35 % to 55 %), the MICP generally first In Fig. 7, the red dashed circles highlight four peak values, which
rises and then falls. When the CO2 concentration is around 43 %, the occur at 10 %NH3 + 90 %diesel (35 %CO2/65 %O2), 10 %NH3 + 90 %
MICP reaches its peak, approximately 8.67 MPa. At this point, the diesel (43 %CO2/57 %O2), 30 %NH3 + 70 %diesel (43 %CO2/57 %O2),
combustion process is most complete, with a fast flame propagation and 40 %NH3 + 60 %diesel (35 %CO2/65 %O2), respectively. Under the
speed and high pressure. At a CO2 concentration of 39 %, a distinct

7
Y. Liu et al. Fuel 389 (2025) 134474

Fig. 5. Orbit-based Fukui functions fw0 for CO.

Fig. 6. Orbit-based Fukui functions Δfw for CO.

condition of 10 %NH3 + 90 %diesel (35 %CO2/65 %O2), the MICP combustion reaction, resulting in a faster flame propagation speed and
reaches 8.33 MPa. At this condition, the CO2 concentration is relatively greater energy release, leading to a peak in-cylinder pressure.
low, the dilution effect is weak, and there is sufficient oxygen during the Under the condition of 10 %NH3 + 90 %diesel (43 %CO2/57 %O2),
combustion process, allowing the combustion of n-heptane to proceed the MICP reaches the highest value of 8.67 MPa. This is likely because,
fully. Meanwhile, the NH3 concentration is low, comprising only a small as the CO2 concentration increases to 43 %, some of the CO2 decomposes
portion of the fuel, so NH3 has a minimal inhibitory effect on combus­ into CO at high temperatures and reacts with HNO to generate NH
tion. The NH radicals produced during combustion can enhance the radicals, which help to enhance the combustion process. At the same

8
Y. Liu et al. Fuel 389 (2025) 134474

moderate CO2 concentration and low NH3 concentration can maintain

high in-cylinder pressure. Despite the presence of the dilution effect, the
radical reactions and lower NH3 concentration help to sustain combus­
tion intensity.
When the NH3 concentration increases to 30 %, the lower calorific
value and slower combustion rate of NH3 begin to inhibit the overall
combustion process. However, the CO and NH radicals generated from
CO2 decomposition still enhance the combustion reaction, so at 30 %
NH3 + 70 %diesel (43 %CO2/57 %O2), the peak pressure remains at
8.34 MPa. As the NH3 concentration increases to 40 %, the inhibitory
effect of NH3 on combustion dominates the process. Therefore, although
a peak is still observed at 40 %NH3 + 60 %diesel (35 %CO2/65 %O2),
the MICP drops to 7.90 MPa.
As shown in Fig. 8, the solid line represents the experimental values,
and the dashed line represents the simulated values. The ADOC model,
which considers the reaction between CO and HNO, can accurately
predict the trend of MICP with changes in CO2 concentration. When the
CO2 concentration increases from 35 % to 39 %, the ADOC model can
describe the inhibitory effect of CO2 on diesel combustion. As the NH3
concentration increases, the dilution effect of CO2 is gradually replaced
by the inhibitory effect of NH3, resulting in a gradual decrease in the
slope within this range. When the CO2 concentration increases from 39
Fig. 7. Experimental values of maximum in-cylinder pressures. % to 43 %, the MICP reaches a peak, and the ADOC model can effectively
describe this phenomenon. As the CO2 concentration continues to in­
time, the NH3 concentration remains low at 10 %, so the negative impact crease from 43 % to 55 %, both the experimental and simulated values
on combustion is minimal. Although the dilution effect of CO2 is show that the MICP gradually decreases. Across all conditions, the
stronger in this range, it does not significantly weaken the combustion ADOC model shows a maximum error of 6.53 %, and an average error of
intensity, and overall combustion efficiency remains high. Therefore, 3.87 % compared to the experimental values.

Fig. 8. Comparison between simulated and experimental values of MICP.

9
Y. Liu et al. Fuel 389 (2025) 134474

4.4. Flame area and propagation speed because, during the early stages of combustion, the fuel and oxygen are
well-mixed, accelerating the combustion process and causing the flame
Fig. 9(a) shows the flame propagation process from 14.95 ms to to spread quickly. In the false-color images, the color change at the edges
15.85 ms after ignition under the condition of 10 % NH3 + 35 % CO2. also reflects changes in flame intensity and propagation speed. The
The images present the dynamic changes of the flame from three per­ flame edge extends from blue regions outward, accompanied by a
spectives: real images, false-color images, and flame edge extraction. At temperature rise, indicating that the flame is propagating quickly during
14.95 ms, the flame has just been ignited, with a small flame area and these moments. From 15.65 ms to 15.85 ms, although the flame is still
low brightness. As time progresses, the flame gradually enlarges, espe­ spreading, the propagation speed slows down, the flame edge becomes
cially between 15.35 ms and 15.65 ms, where the flame area expands smoother, and the flame approaches a stable state, with combustion
significantly, almost covering the circular experimental area. This re­ gradually entering a balanced phase. According to the study by Nadimi
flects the acceleration of the combustion reaction, and the increasing et al. [56], as the NH3 content increases, the combustion mode gradually
flame area indicates that the combustion is entering a more intense shifts from diffusion combustion, typical of pure diesel combustion, to a
phase. In the false-color images, the thermal intensity of the flame can be premixed combustion mode under dual-fuel conditions. This transition
observed more clearly: it gradually transitions from blue regions to red results in a shorter combustion duration and an earlier combustion
regions, indicating that as combustion continues, the flame intensity phase, which aligns with the rapid acceleration of flame propagation
increases, and the temperature rises. The expansion of the red regions observed at low NH3 concentrations in the experiment. Under these
shows the rapid spread of the flame within the experimental area. By conditions, although the dilution effect of CO2 is present, the relatively
15.85 ms, the flame nearly fills the entire field of view, reaching its low concentration of NH3 minimizes its inhibitory effect on the com­
maximum area, suggesting that combustion is approaching a fully bustion process, thereby maintaining a higher combustion intensity.
propagated state. As shown in Fig. 9(b), under the condition of 43 % CO2 + 57 % O2,
From the flame edge extraction results, it can be seen clearly that the the ignition time for the mixture of 10 % NH3 + 90 % diesel is relatively
flame edge gradually expands outward, indicating that the flame prop­ late. After false-color processing, a weak blob-like flame can be observed
agation speed is accelerating. Particularly between 15.35 ms and 15.55 at 24.50 ms, at which point the cylinder pressure rises rapidly. By 25.15
ms, the flame edge expansion speed increases significantly, showing a ms, the area of the blob-like flame increases, and its brightness slightly
noticeable acceleration in the flame’s outward propagation. This is enhances. At 25.80 ms, the flame’s brightness significantly increases,

Fig. 9. The flame propagation.

10
Y. Liu et al. Fuel 389 (2025) 134474

Fig. 9. (continued).

appearing red in the false-color image, and the flame can be detected by color image shows a significant expansion of the red high-temperature
the edge detection program at this point. From 26.45 ms onwards, the regions. Between 14.25 ms and 15.15 ms, the flame area gradually ap­
flame’s basic shape tends to stabilize, and the ignition of residual spray proaches the boundary of the combustion chamber, with the red region
is clearly visible in the center of the image, forming a triangular area. continuing to expand. However, the basic shape of the flame has stabi­
The primary flame combustion region is located in the lower-left part of lized, and in the center of the combustion chamber, the ignition of re­
the combustion chamber, where the flame appears light blue, indicating sidual spray is clearly visible, indicating that the combustion process is
that the fuel mixture is relatively well-prepared in this area and the gradually entering a high-temperature phase.
combustion temperature is higher. As shown in Fig. 9(d), it illustrates the flame propagation process of
As shown in Fig. 9(c), it illustrates the flame propagation process of 40 % NH3 + 60 % diesel in a 35 % CO2/65 % O2 atmosphere. In the early
30 % NH3 + 70 % diesel in a 43 % CO2/57 % O2 atmosphere. At 12.45 stages (from 16.95 ms to 17.35 ms), the flame propagation is relatively
ms, the flame has just been ignited, with a very small flame area, and the slow, with a small flame area in a “butterfly shape.” Combustion has just
false-color image shows a relatively uniform low-temperature region begun, and the false-color image mainly shows blue regions, indicating
(blue). By 13.05 ms, the flame begins to rapidly expand, and the tem­ relatively low temperatures at this stage. After 17.75 ms, the flame
perature starts to rise. Blue flames appear in the real image, and a small propagation speed significantly accelerates, and the flame temperature
amount of red high-temperature areas also emerge in the false-color rises rapidly, with a marked increase in red regions, indicating that
image. At 13.65 ms, the flame area further increases, and the false- combustion has reached a high-temperature zone, and the temperature

11
Y. Liu et al. Fuel 389 (2025) 134474

distribution within the flame becomes more uniform. The high- occurring. Under the condition of 30 % NH3 + 70 % diesel (43 % CO2/
temperature region expands rapidly, and by 18.95 ms, the flame 57 % O2), the flame area increases from 0 mm2 at 12.45 ms to 2404.31
nearly fills the entire combustion chamber. The false-color image shows mm2 at 15.15 ms, with the flame propagation speed reaching 1614.65
the red region covering the flame, indicating that combustion has mm2/ms and peaking at 1743.44 mm2/ms at 13.65 ms. However, it then
reached its peak and the intensity is at its maximum. rapidly declines, eventually reaching − 73.59 mm2/ms at 15.15 ms, with
Using OpenCV to calculate the flame area and flame propagation an average flame propagation speed of 890.48 mm2/ms. This condition
speed from flame edge detection images, the results are shown in Fig. 10. shows faster initial flame propagation, but as the flame approaches its
Under the condition of 10 % NH3 + 90 % diesel (35 % CO2/65 % O2), at maximum area, the propagation speed rapidly decreases. This indicates
14.95 ms, the flame area is relatively small, only 3.53 mm2, while at that a higher proportion of NH3 helps the initial combustion, but the
15.85 ms, the flame area rapidly expands to 3819.07 mm2. During this inhibitory effect of CO2 on combustion remains significant. Under the
period, the flame propagation speed increases from 364.29 mm2/ms to condition of 40 % NH3 + 60 % diesel (35 % CO2/65 % O2), the flame
11629.79 mm2/ms, with the highest speed observed at 15.65 ms. This area increases from 1.77 mm2 at 16.95 ms to 4288.01 mm2 at 18.95 ms,
indicates that the flame undergoes a rapid acceleration phase in the showing relatively significant flame area growth. The flame propagation
early stages of propagation, and as the flame expands, the propagation speed reaches 819.10 mm2/ms at 17.35 ms and peaks at 3257.63 mm2/
speed gradually decreases, but the overall average propagation speed ms at 18.15 ms, then gradually decreases to 1483.65 mm2/ms at 18.95
remains at 4977.65 mm2/ms. This trend suggests that under this specific ms. The average flame propagation speed is 2143.12 mm2/ms.
condition, the initial acceleration of the flame is quite pronounced, In summary, as the CO2 concentration increases (from 35 % to 43 %),
followed by a stabilization of the expansion speed. This may be related the flame propagation speed significantly decreases, and in some cases,
to the fuel mixture ratio and the composition of the oxidizer. The higher negative values for flame propagation (flame area reduction) are
proportion of CO2 may suppress the combustion rate, leading to faster observed. This indicates that CO2 reduces the energy for flame propa­
flame expansion at higher temperatures. Under the condition of 10 % gation by diluting oxygen and absorbing heat. As the NH3 concentration
NH3 + 90 % diesel (43 % CO2/57 % O2), the flame propagation speed increases, the initial acceleration of flame propagation becomes more
shows more complex fluctuations. From 24.50 ms to 27.75 ms, the flame pronounced, especially under the condition of 40 % NH3, where the
propagation speed increases from 3.40 mm2/ms to 1511.50 mm2/ms, flame propagation speed is relatively high, and the flame area increases
but then gradually decreases to − 174.59 mm2/ms (negative values more rapidly. This suggests that NH3 provides more combustion energy
indicate a reduction in flame area). The average flame propagation in the early stages, promoting flame propagation. However, when the
speed is 394.146 mm2/ms. This suggests that in an environment with a CO2 content reaches 43 %, the average flame speed increases with the
higher proportion of CO2, flame propagation may be more strongly rise in NH3 content. According to the reaction pathways discussed in
inhibited, with phenomena such as partial backflow or flame extinction Section 4.1 and the study by Shrestha et al. [57], CO2 may partially

Fig. 10. Flame Area and Flame Propagation Speed.

12
Y. Liu et al. Fuel 389 (2025) 134474

decompose into CO at high temperatures, which can subsequently react [5] Chu X, et al. High-temperature auto-ignition characteristics of NH3-H2-CH4. Fuel
2024;365:131228.
with NH and HNO free radicals, thereby enhancing the combustion
[6] Wen M, et al. Optical study of combustion stability in dual fuel approach using
process. ammonia and high reactivity fuel. Energ Conver Manage 2024;319:118910.
[7] Langella, G., et al. Ammonia as a fuel for internal combustion engines: Latest
advances and future challenges. In: Journal of Physics: Conference Series. Vol.
5. Conclusions
2385. No. 1. IOP Publishing, 2022.
[8] Valera-Medina A, Xiao H, Owen-Jones M, David WI, Bowen PJ. Ammonia for
The ammonia-diesel oxidation combustion (ADOC) model proposed power. Prog Energ Combust Sci 2018;69:63–102.
[9] Wang L, et al. Effect of CO2 and its concentration variation on the ignition and
in this study effectively simulates the combustion process of ammonia-
combustion of diesel surrogate fuel: optical experiments and numerical
diesel blended fuel (ADBF) in a CO2/O2 environment. The analysis re­ simulations. J Energy Inst 2023;110:101350.
veals the significant role of HNO in this environment. Experimental [10] Liu Y, et al. The third body effect of carbon dioxide on N-heptane ignition delay
results also provide insights into the effects of CO2 and NH3 on the characteristics under O2/CO2 conditions. Combust Sci Technol 2022;194(14):
2817–35.
combustion characteristics of ADBF. The specific conclusions are as [11] Nadimi E, et al. Effects of using ammonia as a primary fuel on engine performance
follows: and emissions in an ammonia/biodiesel dual-fuel CI engine. Int J Energy Res 2022;
46(11):15347–61.
[12] Nadimi E, Przybyła G, Lewandowski MT, et al. Effects of ammonia on combustion,
(1) The ADOC model effectively simulates ADBF combustion in a emissions, and performance of the ammonia/diesel dual-fuel compression ignition
CO2/O2 environment. The maximum is 6.53 % at 10 % NH3 + 90 engine. J Energy Inst 2023;107:101158.
% diesel (35 % CO2/65 % O2), with an average error of 3.83 %. [13] Li T, et al. A comparison between low-and high-pressure injection dual-fuel modes
of diesel-pilot-ignition ammonia combustion engines. J Energy Inst 2022;102:
(2) In the reaction between CO and HNO radicals, CO is prone to gain 362–73.
unpaired electrons from HNO radicals. In the reaction pathway of [14] Cai K, et al. Combustion behaviors and unregular emission characteristics in an
CO + HNO, the formation of END-2 is easier than the formation of ammonia–diesel engine. Energies 2023;16(19):7004.
[15] Xu L, et al. Performance and emission characteristics of an ammonia/diesel dual-
END-1, with an energy barrier of 80.132 kcal/mol.
fuel marine engine. Renew Sustain Energy Rev 2023;185:113631.
(3) With the increase in CO2 concentration (from 35 % to 43 %), a [16] Zhou X, et al. Pilot diesel-ignited ammonia dual fuel low-speed marine engines: a
significant reduction in flame propagation speed is observed. At a comparative analysis of ammonia premixed and high-pressure spray combustion
modes with CFD simulation. Renew Sustain Energy Rev 2023;173:113108.
CO2 concentration of 35 %, the initial acceleration of flame
[17] Reiter AJ, Kong S-C. Demonstration of compression-ignition engine combustion
propagation becomes increasingly evident as the NH3 concen­ using ammonia in reducing greenhouse gas emissions. Energy Fuel 2008;22(5):
tration rises. Notably, at a CO2 concentration of 43 %, the flame 2963–71.
propagation speed at an NH3 concentration of 30 % shows a [18] Liu H, et al. Effects of flame temperature on PAHs and soot evolution in partially
premixed and diffusion flames of a diesel surrogate. Energy Fuel 2019;33(11):
substantial improvement compared to that at 10 % NH3, high­ 11821–9.
lighting the pronounced effect of increased NH3 levels under [19] Yang C, et al. Effects of ammonia energy fractions, diesel injection timings, and
higher CO2 conditions. loads on combustion and emission characteristics of PFI-DI ammonia-diesel
engines. Int J Engine Res 2024;25(4):743–57.
[20] Li M, et al. Investigation of methane oxy-fuel combustion in a swirl-stabilised gas
CRediT authorship contribution statement turbine model combustor. Energies 2017;10(5):648.
[21] Dong W, et al. Effects of carbon dioxide on the combustion characteristics of the
laminar premixed n-heptane/air flames at elevated pressures. J Energy Inst 2021;
Yongfeng Liu: Funding acquisition, Conceptualization. Chenyang 99:127–36.
Yin: Writing – original draft, Validation, Methodology. Jiaying Pan: [22] Zhang D, et al. Experimental and numerical investigation of vitiation effects on the
Data curation. Jin’ou Song: Methodology. Hua Sun: Writing – review & auto-ignition of n-heptane at high temperatures. Energy 2019;174:922–31.
[23] Wang L, et al. A phenomenological model of diesel combustion characteristics
editing. under CO2/O2 atmosphere. Fuel Process Technol 2022;229:107167.
[24] Gokulakrishnan P, et al. NOx formation from ammonia, and its effects on oxy-
combustion of hydrocarbon fuels under supercritical-CO2 conditions. Applic
Declaration of competing interest Energy Combust Sci 2023;13:100110.
[25] Wen M, et al. Study on combustion stability and flame development of ammonia/n-
The authors declare that they have no known competing financial heptane dual fuel using multiple optical diagnostics and chemical kinetic analyses.
J Clean Prod 2023;428:139412.
interests or personal relationships that could have appeared to influence [26] Shi Y, et al. Study on the combustion characteristics of NH3 and the inhibition
the work reported in this paper. characteristics of R134a on NH3. Combust Flame 2024;264:113438.
[27] Zhang J, et al. The impact of hydrogen addition and OH concentration on NO
emissions in high-pressure NH3/air combustion. Int J Hydrogen Energy 2024;54:
Acknowledgments 1017–28.
[28] Wei Z, et al. The coupling influence of chemical/physical effects of NH3 on the NO
The authors thank for the financial support provided by the National formation in the premixed CH4-NH3 flames. J Energy Inst 2024;114:101614.
[29] Dai L, et al. Experimental and numerical analysis of the autoignition behavior of
Natural Science Foundation of China (No. 52376091), State Key Labo­ NH3 and NH3/H2 mixtures at high pressure. Combust Flame 2020;215:134–44.
ratory of Engines (K2023-04), The Cultivation project Funds for Beijing [30] Shmakov AG, et al. Development of the detailed mechanism of pyrolysis and
University of Civil Engineering and Architecture(X24030). combustion of triphenyl phosphate: New quantum chemistry calculations and
experimental data on structure of the H2/O2/Ar flame doped with TPP. Combust
Flame 2024;266:113534.
Data availability [31] Diao S, Li H, Minggao Yu. Atomic insights into the combustion mechanism of DME/
NH3 mixtures: a combined ReaxFF-MD and DFT study. Int J Hydrogen Energy
2024;80:743–53.
The data that has been used is confidential.
[32] Zhang Y, et al. Analysis of oxidation pathways for characteristic groups in coal
spontaneous combustion. Energy 2022;254:124211.
References [33] Chen P, et al. Effect of ammonia on N migration and transformation characteristics
during coal pyrolysis: quantum chemical calculations and pyrolysis experiments.
Fuel 2024;361:130762.
[1] Alnajideen M, et al. Ammonia combustion and emissions in practical applications:
[34] Stephens PJ, et al. Ab initio calculation of vibrational absorption and circular
a review. Carbon Neutrality 2024;3(1):1–45.
dichroism spectra using density functional force fields. J Phys Chem 1994;98(45):
[2] Liu H, et al. A perspective on the overarching role of hydrogen, ammonia, and
11623–7.
methanol carbon-neutral fuels towards net zero emission in the next three decades.
[35] Becke AD. Density-functional thermochemistry. I. The effect of the exchange-only
Energies 2022;16(1):280.
gradient correction. J Chem Phys 1992;96(3):2155–60.
[3] Berwal P, Kumar S. Effect of Various Fuel Blends on the Laminar Burning Velocity
[36] Lee C, Yang W, Parr RG. Development of the Colle-Salvetti correlation-energy
of Ammonia–Air Mixtures. In: Ammonia and Hydrogen for Green Energy Transition.
formula into a functional of the electron density. Phys Rev B 1988;37(2):785.
Singapore: Springer Nature Singapore; 2024. p. 39–70.
[37] Vosko SH, Wilk L, Nusair M. Accurate spin-dependent electron liquid correlation
[4] Faghih M, et al. Effect of radiation on laminar flame speed determination in
energies for local spin density calculations: a critical analysis. Can J Phys 1980;58
spherically propagating NH3-air, NH3/CH4-air and NH3/H2-air flames at normal
(8):1200–11.
temperature and pressure. Combust Flame 2023;257:113030.

13
Y. Liu et al. Fuel 389 (2025) 134474

[38] Grimme S, et al. A consistent and accurate ab initio parametrization of density [48] Woon DE, Dunning Jr TH. Gaussian basis sets for use in correlated molecular
functional dispersion correction (DFT-D) for the 94 elements H-Pu. J Chem Phys calculations. III. The atoms aluminum through argon. J Chem Phys 1993;98(2):
2010;132:15. 1358–71.
[39] Grimme S, Ehrlich S, Goerigk L. Effect of the damping function in dispersion [49] Pino-Rios R, et al. Proposal of a simple and effective local reactivity descriptor
corrected density functional theory. J Comput Chem 2011;32(7):1456–65. through a topological analysis of an orbital-weighted fukui function. J Comput
[40] Binning Jr RC, Curtiss LA. Compact contracted basis sets for third-row atoms: Ga- Chem 2017;38(8):481–8.
Kr. J Comput Chem 1990;11(10):1206–16. [50] Pino-Rios R, et al. Orbital-weighted dual descriptor for the study of local reactivity
[41] McGrath MP, Radom L. Extension of Gaussian-1 (G1) theory to bromine-containing of systems with (quasi-) degenerate states. Chem A Eur J 2019;123(49):10556–62.
molecules. J Chem Phys 1991;94(1):511–6. [51] Wu Y, et al. Understanding the antagonistic effect of methanol as a component in
[42] Curtiss LA, et al. Extension of Gaussian-2 theory to molecules containing third-row surrogate fuel models: a case study of methanol/n-heptane mixtures. Combust
atoms Ga-Kr. J Chem Phys 1995;103(14):6104–13. Flame 2021;226:229–42.
[43] Čížek J. On the use of the cluster expansion and the technique of diagrams in [52] Niemeyer KE, Sung CJ. On the importance of graph search algorithms for DRGEP-
calculations of correlation effects in atoms and molecules. Adv Chem Phys 1969; based mechanism reduction methods. Combust Flame 2011;158(8):1439–43.
14:35–89. [53] Yu C, et al. Methane/air auto-ignition based on global quasi-linearization (GQL)
[44] Purvis GD, Bartlett RJ. A full coupled-cluster singles and doubles model: the and directed relation graph (DRG): implementation and comparison. Combust Sci
inclusion of disconnected triples. J Chem Phys 1982;76(4):1910–8. Technol 2020;192(9):1802–24.
[45] Scuseria GE, Janssen CL, Schaefer HF. An efficient reformulation of the closed-shell [54] Li AY. Theoretical investigation of hydrogen bonds between CO and HNF2, H2NF,
coupled cluster single and double excitation (CCSD) equations. J Chem Phys 1988; and HNO. Chem A Eur J 2006;110(37):10805–16.
89(12):7382–7. [55] Xu L, et al. A skeletal chemical kinetic mechanism for ammonia/n-heptane
[46] Scuseria GE, Schaefer HF. Is coupled cluster singles and doubles (CCSD) more combustion. Fuel 2023;331:125830.
computationally intensive than quadratic configuration interaction (QCISD)? [56] Nadimi E, et al. Effects of ammonia on combustion, emissions, and performance of
J Chem Phys 1989;90(7):3700–3. the ammonia/diesel dual-fuel compression ignition engine. J Energy Inst 2023;
[47] Watts JD, Gauss J, Bartlett RJ. Coupled-cluster methods with noniterative triple 107:101158.
excitations for restricted open-shell Hartree-Fock and other general single [57] Shrestha KP, et al. Detailed kinetic mechanism for the oxidation of ammonia
determinant reference functions. Energies and analytical gradients. J Chem Phys including the formation and reduction of nitrogen oxides. Energy Fuel 2018;32
1993;98(11):8718–33. (10):10202–17.

14