International Journal of Hydrogen Energy 137 (2025) 925–938

Contents lists available at ScienceDirect

International Journal of Hydrogen Energy

journal homepage: [Link]/locate/he

Hydrogen-fuelled internal combustion engines: Direct Injection versus

Port-Fuel Injection
Fabián Musy , Rafael Ortiz , Inmaculada Ortiz , Alfredo Ortiz *
Chemical & Biomolecular Engineering Department, Universidad de Cantabria, Av. Los Castros 46, 39005, Santander, Spain

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

Handling Editor: Ibrahim Dincer The road-transport is one of the major contributors to greenhouse global gas (GHG) emissions, where hydrogen
(H2) combustion engines can play a crucial role in the path towards the sector’s decarbonization goal. This study
Keywords: focuses on comparing the performance and emissions of port-fuel injection (PFI) and direct injection (DI), in a
Internal combustion engine spark ignited combustion engine when is fuelled by hydrogen and other noteworthy fuels like methane and coke
Hydrogen
oven gas (COG). Computational fluid dynamic simulations are performed at optimal spark advance and air-fuel
Direct injection
ratio (λ) for engine speeds between 2000 and 5000 rpm. Analysis reveals that brake power increases by 40% for
Port-fuel injection
CFD simulation DI, attributed to 30.6% enhanced volumetric efficiency, while the sNOx are reduced by 36% compared to PFI at
Hydrogen combustion optimal λ = 1.5 for hydrogen. Additionally, H2 results in 71.8% and 67.2% reduction in fuel consumption
compared to methane and COG respectively, since the H2 lower heating value per unit of mass is higher.

by-product, which consists predominantly of a H2–CH4 blend, as

detailed in Table 1 [3–7].
1. Introduction The H2 molecule is the only one free of carbon atoms, offering a clean
combustion in terms of GHG. Additionally, the lower heating value
The new wave of hydrogen related studies and technological devel- (LHV), which represents the energy released during fuel combustion per
opment has mainly shaken Europe due to the eagerness of achieving the unit of mass, is nearly three times higher than conventional fuels, CH4
goal of net zero greenhouse gases (GHGs) emissions by 2050. In this and COG, proving the high potential of H2 as clean energy vector.
sense, transport sector represents the 25% of GHGs in the European However, it has the lowest density as a consequence of the lower mo-
Union (EU), and hence the European Green Deal demands a 90% lecular weight, and hence, power density is penalized in gas phase, being
emissions reduction by 2050 [1]. measured per unit of volume. Finally, regarding the higher laminar
To address the problem of emissions related to road transport, key flame speed (SL ) of H2, it helps improve the combustion efficiency,
actors in energy transition such as public authorities and private motor throughout the improvement in chemical reactivity [8].
companies are gradually transitioning from conventional technologies Afterwards, different types of configurations which combine the
as internal combustion engines (ICEs) to electric vehicles and alternative ignition mechanism and injection strategies are reproduced in H2-ICEs
fuels. In this sense, H2 fuelled internal combustion engines (H2-ICEs) and such as spark ignition (SI) coupled with direct or port-fuel injection (DI
fuel cell electric vehicles (FCEV) are suitable options to reduce the or PFI), and compression ignition (CI) [7].
environmental impacts arising from conventional vehicles. Moreover, Port-fuel injection with spark ignition (SI-PFI) is the most common
H2-ICEs may ease the transition towards low carbon mobility in a more strategy, which is based on injecting H2 into the intake manifold instead
economical way thanks to the upgrading of current ICEs without of directly into the combustion chamber [9–14]. In H2 SI-PFI systems the
requiring the presence of batteries and fuel cells. However, even though variation on the air-fuel ratio (λ) exerts great influence on engine per-
the use of H2 as fuel achieves zero CO2 emissions during combustion, it formance. The relative λ factor, represents the proportion of air per unit
has associated nitrogen oxides (NOx) emissions [1,2]. of mass of fuel in the mixture and is expressed as the coefficient of the
Regarding the application of H2 in ICEs, the H2 molecule exhibits real air-fuel ratio (λreal ) obtained in the simulations or bench test and the
distinct intrinsic physicochemical properties compared with conven- stoichiometric factor (λSt ), as defined in Eq. (1) [9].
tional fuels and other alternatives, including CH4 and coke oven gas
(COG), a residual stream arising from coke production and valorised as

* Corresponding author.
E-mail address: ortizal@[Link] (A. Ortiz).

[Link]
Received 27 February 2024; Received in revised form 24 May 2024; Accepted 9 July 2024
Available online 16 July 2024
0360-3199/© 2024 The Authors. Published by Elsevier Ltd on behalf of Hydrogen Energy Publications LLC. This is an open access article under the CC BY-NC-ND
license ([Link]
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

Nomenclature ρi Density (kg/m3)

ρa.i Air density at atmospheric conditions (g/cm3)
A Area (cm2) ηe Volumetric efficiency (%)
BMEP Brake mean effective pressure (bar) ṁ Mass flow (g/s)
BSFC Brake specific fuel consumption (g/kWh) ṁe Air mass inlet flow (g/h)
BTE Brake thermal efficiency (%) ma Real air mass (g)
CAD Crank angle degrees (◦ ) ṁf Mass flow of fuel injected per cycle (g/s)
CHR Chemical heat release (J) mfull Total mass of fuel injected per cycle (g)
dBTDC Degrees before top dead centre (◦ ) λ Relative air-fuel ratio
dATDC Degrees after top dead centre (◦ ) λSt Stoichiometric air-fuel ratio
dBBDC Degrees before bottom dead centre (◦ ) λreal Real air-fuel ratio
dABDC Degrees after bottom dead centre (◦ ) ηR Number of crank revolutions for each power stroke per
EOI End of injection (◦ ) cylinder
HR (TA ) Reagent enthalpy (J)
HP (TA ) Product enthalpy (J) Abbreviations
HP Horse power (HP) CFD Computational fluid dynamics
LHV Lower heating value (MJ/kg) CI Compression ignition
MBT Maximum brake torque (N⋅m) COG Coke oven gas
N Engine speed (rpm) CR Compression ratio
PW Pulse width (◦ ) DI Direct injection
SA Spark advance (◦ ) EU European union
SL Laminar flame speed (m/s) FCEV Fuel cell electric vehicles
sNOx Specific nitrogen oxides (g/kWh) GHGs Greenhouse gases
SOI Start of injection (◦ ) H2-ICE Hydrogen fuelled internal combustion engine
T Torque (N⋅m) HPDI High pressure direct injection
TDC Top dead centre (◦ ) ICE Internal combustion engine
Vd Cylinder volume displacement (L) PFI Port-fuel injection
vi Injection velocity (m/s) rpm Revolutions per minute
Wc,i Work delivered per cycle (J/cycle) SI Spark ignition
WHT Wall heat transfer (J) SI-DI Spark ignition – direct injection
p Pressure (bar) SI-PFI Spark ignition – port-fuel injection
P Power (kW) WOT Wide open throttle
PInj Injection Pressure (bar)

λreal range between 4.8 and 66.7 kW for engines volumes from 0.27 L (mo-
λ= (1)
λSt torbikes) to 3.984 L (heavy-duty vehicles), and BTE values in the range
22.7–38% [18].
Thus, according to Verhelst et al., operating at λ equal to 1.4, a 50%
On the other side, DI configuration is based on injecting H2 directly
net power output is achieved in comparison with the stoichiometric
into the combustion cylinder while the compression stroke takes place,
(1:1) air-fuel proportion (λ = 1). This enhancement is attributed to the
improving the engine performance by avoiding the displacement of air.
synergistic effects of a turbocharger and an exhaust gas recirculation
This results in a higher volumetric efficiency and thus, in higher power
(EGR) system [15]. Furthermore, in the research conducted by Tang
density than PFI systems [19].
et al. a 65% of power output obtained in the same engine powered by
Regarding the main differences between the PFI and the DI systems,
gasoline can be obtained through a more precise optimization of the λ
in the latter, the injector system is positioned in the head, allowing the
factor [16]. Despite the lower power output compared to gasoline en-
transport of the fuel directly to the cylinder [8]. Thus, several studies
gines, Tang et al. demonstrates that the engine brake thermal efficiency
have demonstrated that high pressure direct injection (HPDI) can ach-
(BTE), defined as the ratio of the mechanical energy extracted from the
ieve similar efficiencies to diesel engines. Mathias et al. show that
ICE to the fuel consumed, can be enhanced to 38%, surpassing that of
operating under ultra-lean conditions (λ = 3.3), BTE reaches a 35% in
gasoline engines [17].
most tests [20]. Mohammadi et al., also achieves a maximum BTE of
In addition, in PFI H2-ICEs fuelled by H2, the power outputs usually

Table 1
Properties of fuels at standard conditions [3–7].
Parameter Unit Gasoline Diesel Hydrogen Methane Coke Oven Gas

Chemical Formula C4–C12 C8–C25 H2 CH4 COG

Composition (C, H, O, N) Mass-% 84,16,0,0 86,14,0,0 0,100,0,0 25,75,0,0 24,46,16,14
Lower heating value (LHV) MJ/kg 43.9 42.5 119.9 45.8 39.9
Density kg/m3 730–780 830 0.089 0.72 0.41
Stoichiometric air/fuel ratio kg/kg 14.7 14.5 34.4 17.2 12.72
Flammability limits Vol-% 1.0–7.6 0.6–5.5 4.0–76.0 5.3–15.0 4.4–34.0
Laminar flame speed in air m/s 0.37–0.43 0.37–0.43 1.85 0.38 0.68–0.88
Auto-ignition temperature K ≈623 ≈523 853 813 ≈673

926
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

35% operating at lean conditions (φ = 0.5; λ = 2) [21]. Finally, Oikawa 2. ICE design and CFD simulation
et al. obtained a wide range of BTE results due to the different DI ap-
proaches studied achieving a maximum of 45% [22]. 2.1. Geometrical design
Likewise, due to the shorter residence time of the fuel inside the
engine, thermal efficiency values between 34 and 45.5% are obtained, An experimental bench of a naturally aspirated Volkswagen Polo 1.4
using engine volumes ranging from 0.66 (motorbikes) to 7.75 L (heavy- L SI-PFI internal combustion engine with a compression ratio (CR)
duty vehicles), even increasing power by up to 20 % compared to PFI 10.5:1 was tested and computationally reproduced in a previous work
[17,20–27]. reported by our research group [29]. In that work, the engine was first
Thus, in pursuit of enhanced engine performance, this configuration adapted to be fuelled by gaseous fuels. Those modifications mainly
has adopted a range of injection strategies, encompassing optimized include a replacement of the plastic inlet manifold by a cast manifold
duration and timing, alongside multi-injection strategies utilizing two or and the substitution of gasoline injectors by H2 injectors in stainless steel
more pulses per cycle [8]. 316L, preventing gas-fuel leakages.
Furthermore, the adoption of direct injection with spark ignition (SI- In the is study, the original Volkswagen engine was computationally
DI) techniques effectively eradicated combustion anomalies such as modified to switch from PFI to DI strategy. Firstly, the new designed
knock and backfiring. In these engines, direct H2 injection promotes a geometry was reproduced on Autodesk Inventor as can be seen in Fig. 1,
more homogeneous distribution of air-fuel mixture, thereby mitigating highlighting the new shape of the intake manifold and the location of the
the occurrence of backfiring. This approach, coupled with H2 injection direct injector.
at the beginning of compression stroke, prevents knock phenomena [20, The main changes in comparison with the PFI geometry model
28]. involved the removal of the gas inlet on the intake manifold, giving
Regarding NOx, DI engines result in a 20% reduction compared to geometrical continuity. This modification allows better flow of gas and
NOx levels observed in diesel engines operating at full load for avoids the creation of turbulences. Subsequently, the location of the
maximum power output [5,8]. In addition, they exhibit a strong direct injector on the head of the engine is defined to avoid the inter-
dependence on fuel load, showing that the results of early injection ruption of the valve and piston movement paths. Furthermore, the DI
strategies increase NOx emissions, and hence, late injection is preferable system was intentionally designed to employ identical CR and
at partial and low loads [8]. displacement volume (Vd ), ensuring a comparison under consistent
According to the adaptation of the ICE to be directly powered by H2, conditions.
some minor modifications are required, for instance, changes in the The injector was designed to spray fuel through 5 nozzles positioned
hardware side, such as the outplacement of fuel injectors, fuel lines, 72◦ apart and inclined 50◦ relative to the vertical axis of the geometry.
charcoal filters and fuel rail and the optimization of the injection system The number and arrangement of the nozzles were optimized to enhance
paying special attention to the aspects of sealing and lubrication, with fuel dispersion within the cylinder, minimize injection duration, and to
the aim of achieve an adequate combustion reaction [27,29–32]. expand the start of injection (SOI) range. Consequently, the injector was
Moreover, the injection strategy has a significant impact on DI per- positioned slightly off-centre from the cylinder head to accommodate
formance and NOx formation. In this context, there are two main types the spark plug as in a real engine. In this case, the spark plug is tilted
of DI systems; low-pressure direct injection (LPDI) and high-pressure with regard to the vertical axe.
direct injection (HPDI). The type of injection has to be incorporated Therefore, the spark and the H2 sprayed will intersect close to the
with a well-defined injection strategy, which in turn must be in synergy central axe of the cylinder, where chemical reaction starts. In addition,
with ignition timing both in PFI and DI [8]. Thus, multiple DI strategies the air-fuel mixture working with early injection timing and 5-nozzle
can be differentiated before and after the spark occurs, while in single DI configuration is normally concentrated on the centre of the cylinder,
is divided in late and early injections. generating a stratified and more enriched mixture for a better combus-
In H2 combustion, λ ratio variations exert a significant influence on tion reaction [35].
combustion velocity and temperature. As reported by Luo et al., λ values Finally, a 6.71 mm diameter was implemented based on the number
surpassing 1.67 yield a maximum combustion temperature below 1600 of nozzles and the nozzle diameter of 0.61 mm, which was selected from
K [33]. This temperature is insufficient to furnish the activation energy well stablished studies on 5-nozzle configurations [36].
required for complete combustion, leading to a degradation of engine In this context, among various configurations, the 5-nozzle spray
efficiency [34]. Conversely, Eichlseder et al., maintain that combustion injector was selected for the superior ability of this configuration to
temperatures must be held below 1800 K to prevent the formation of promote a favourable mixture stratification, enhancing thermal effi-
harmful NOx emissions. This critical threshold is reached at λ lower than ciencies, without compromising NOx emissions working with λ close to
1.4 [18]. Therefore, the optimum λ must be found ranging between 1.4 2, and avoiding complexities of other injection systems, which will be
and 1.67. crucial to compare DI and PFI engine configurations [35].
Despite de existence of various works addressing the distinctions The dedicated design is oriented to spark ignition and HPDI strategy,
between DI and PFI engines, none quantitatively delve into the ad- working under low loads of fuel with air-fuel ratios which ranges from
vancements and outcomes of converting petrol-fuelled ICEs to run on 1.5 to 2 in the case of H2 simulations. Thus, the low loads with early
H2. Under this framework, the primary contributions of the present injection strategy and spark plug ignition in HPDI is the best option to
study are referred to the technological implementation and comparison obtain high brake power outputs while preventing high NOx formation,
of the performance between PFI-SI H2-ICEs and high performance H2- which reaches the theoretical peak at air-fuel ratios of 1.25 [8].
ICEs approaches, mostly based on DI-SI configurations. Moreover, other In this case, the early DI also offers an homogeneous mixture which
fuels such as CH4 and COG have been assessed in order to validate the helps to improve thermal efficiency and reduce NOx through a reduction
engine design in e-fuel applications, comparing them with H2. in combustion temperature, even in comparison with PFI. Moreover, the
This comparison was conducted through computational fluid dy- general behaviour related to the existent trade-off between fuel load and
namic (CFD) simulations of both engines and based on performance and NOx formation also applies to PFI systems. For this reason, early DI
emission parameters such as brake power, torque, brake mean effective strategy is selected [8].
pressure (BMEP), brake specific fuel consumption (BSFC), brake thermal Finally, the closing and opening valve profiles in both engine con-
efficiency (BTE), volumetric efficiency, chemical heat release (CHR) and figurations are the same and do not change with the variation in spark
specific nitrogen oxides emissions (sNOx). advance (SA) and start of injection (SOI) or end of injection (EOI). In this
context, the main values about angles and distances are collected in
Table 2.

927
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

Fig. 1. Geometrical DI engine design.

Consequently, a chemical reaction pathway must be imported. In this

Table 2
regard, identifying a mechanism that accurately represents the com-
Valve profiles main values.
bustion of each fuel is critical as all results hinge on combustion
Intake Valves Profiles modelling. Therefore, the most comprehensive mechanism, GRI-Mech
Opened (IVO, ◦) Closed (IVC, ◦) Max. Lift (cm) 3.0, is implemented. The GRI-Mech 3.0 effectively models H2 and
18.2 15.2 0.84 CH4, including all their possible reactions, and COG, defined as a
Exhaust Valve Profiles
mixture of several gases such as H2, CH4, nitrogen (N2), carbon dioxide
Opened (EVO, ◦) Closed (EVC, ◦) Max. Lift (cm)
19.4 16.4 0.86 (CO2) and carbon monoxide (CO) demonstrating the model complete-
ness. Thus, this reaction model developed by the University of Califor-
nia, in Berkeley, comprises 325 reactions and 53 species [37].
Thus, the valve profiles remain unchanged across both configura- In this sense, the GRI-Mech 3.0 optimizes the combustion reactions
tions to ensure consistent operation comparable to the Volkswagen Polo carried out within a temperature range among 1000 and 2500 K and air-
1.4L PFI. Therefore, this model has been implemented with minimal fuel ratios from 0.2 to 10. In addition, run simulations over this vali-
geometric changes, without altering the exhaust mixture evacuation dation ranges could be also reasonable, as elementary reactions en-
times or air intake timings. This approach allows for a focused assess- compasses all the required steps deemed crucial for describing the
ment of the resultant advantages and disadvantages derived solely from combustion and flame propagation, including the reduction or forma-
changing the injection system and the intake manifold shape. tion of NOx [37].
Conversely, the valve profiles are also maintained to get a suitable The definition of the reaction mechanism is essential to establish
synergy among the different parts of the engine avoiding physical col- operational parameters, a critical step in achieving optimal performance
lisions between valves and piston which have been configured as mov- in H2-ICE simulations. Accordingly, spark characteristics are defined to
ing parts. determine their crankshaft angle duration (CAD), using Eq. (2). This
calculation utilizes the nominal spark angle (720◦ ), spark time (ms), and
engine speed (N) (rpm).
2.2. CFD Simulation
( rev ) 1 min 360 CAD 1⋅10− 3 s
Duration = N ⋅ ⋅ ⋅ 1 ms⋅ (2)
The designed geometrical model is imported to ANSYS Workbench to min 60 s 1 rev 1 ms
be adapted on the computational aided design tool ANSYS SpaceClaim Furthermore, the spark advance, which characterizes the timing of
and meshed on ANSYS Meshing to prepare the geometry. Subsequently, spark ignition, is derived from the values previously gathered in the
simulations were carried out in the computational fluid dynamics (CFD) experimental bench and reported by our research group [18]. Therefore,
simulation tool ANSYS Forte. two key operational parameters demand prior definition in the
The combustion chemistry of H2, CH4 and COG is first introduced.

928
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

simulation plan, the λ ratio achieved from simulations ranging from 1 to Table 4
2 and the engine speed, spanning from 2000 to 5000 rpm. Walls boundary conditions: Wall temperature (K).
Afterwards, some other variables and parameters are defined within Boundary Condition Temperature (K) References
the simulation software, with particular emphasis on establishing
Piston 570 [11]
appropriate boundary conditions. The boundary system encompasses Head 550 [11]
inlet, outlet, and wall boundaries. Thus, beginning with the inlets, which Liner 550 –
includes the “Air Inlet”, the main parameters have been compiled in In Valve 400 [9]
Table 3 of previous experiments carried out by Ortiz-Imedio et al. [29]. Out Valve 550 [9]
Intake Manifold 313 [9]
In the case of the air inlet, it is important to highlight the pressure Exhaust Manifold 500 [9]
and temperature effects related to the engine speed. Regarding the
pressure, the higher the engine speed, the lower the pressure used to
feed the intake manifold, because when the speed increases, the piston 4-cylinder engine, assuming that each cylinder is geometrically
creates a greater vacuum inside of the cylinder, reducing the pressure in identical.
the manifold. In parallel with a decrease in intake temperature upon an Finally, to summarize all the steps to carry out the simulation and the
increase in engine speed, a corresponding drop in pressure occurs due to modelling test points, a general outline is depicted in Fig. 3.
elevated air flow velocity. This results in lower residence time of air In this sense, ANSYS Forte solves different equations such as fluid
within the intake manifold, consequently limiting the time available for continuity, momentum and energy conservation and gas-phase mixture
air to warm up. In this context, the outlet system consists of a single equation of state through different models to carry out several purposes.
boundary condition, called the “Exhaust” condition. Thus, the “Exhaust” The explained combustion model GRI-Mech 3.0 includes all the feasible
condition only requires an outlet pressure definition, which should be reactions with all the fuels assessed in this study, even the formation of
set to atmospheric pressure. NOx. In addition, the turbulent flow model used in this study was the
Finally, relating to the boundary conditions of the wall systems, these Reynolds-Averaged-Navier-Stokes (RANS) Re-Normalization Group
have a strong dependency on the wall temperature, being very closely (RNG) k-ε model.
related to the heat transfer phenomena. In this sense, Table 4 summa- Finally, regarding the spray model selected, the droplet collision
rizes the wall boundary conditions implemented in the simulations, model is based on the adaptive collision mesh model and the solid-cone
which were based on diverse literature sources. injection is initialized with the nozzle-flow model, which offers a
As can be observed, several wall boundaries are unaffected by description about nozzle discharge characteristics. This nozzle-flow
changes in the type of injection since they are not involved in the model describes the instantaneous flow conditions inside the nozzle.
modified parts or in combustion reaction, and hence, they have been Furthermore, unsteady gas jet model is employed in order to reduce the
defined with values obtained in PFI experimental bench by our research mesh-dependency of the Kevin-Helmholtz/Rayleigh-Taylor (KH-RT)
group, keeping both systems comparable [29]. breakup model [38].
On the other hand, the walls located in the combustion cylinder re-
gion such as piston, head and liner, must be modified in comparison with
2.3. Simulation plan
PFI, in order to get a more realistic result which have been changed to
the values reported in literature for DI engines [11]. Moreover, in the
Thereafter, the simulation plan requires meticulous definition of the
case of the liner, similar temperature as head is assumed.
injection velocity (vi ) and pulse width (PW), deemed as critical opera-
In this sense, ANSYS Forte assumes that the heat flux through the
tional variables governing the performance and demands of each
cylinder is mainly produced as consequence of gas-phase convection,
simulation. These parameters are mathematically represented in Eqs. (3)
fuel-film conduction and radiation [38]. However, the temperature
and (4) [39].
alongside the wall is undertake as a constant, without spatial variation,
being all of them configured with the Law of the Wall slip condition for ṁf
vi = (3)
the shear stress to accurately capture wall boundary layer’s effect [29, ρi ⋅ANozzle ⋅NNozzle ⋅100
38].
Furthermore, it is necessary to incorporate the defined profiles for N ⋅mfull
PW = (4)
the valve opening and closing movements, which are synchronized with ṁf
the piston strokes, SA and SOI timings.
At this point, the main feature for modelling this type of injection on Where ṁf is the mass fuel inlet flow (g/s), ρi is the density at atmospheric
ANSYS Forte compared with PFI simulations is the Forte spray model, conditions (g/cm3), ANozzle is the cross-section of the nozzle (cm2 ), NNozzle
which allows to reproduce the injection of fuel. Afterwards, the DI 5- is the number of nozzles, N is the engine speed (rpm), ṁf is the mass flow
()
nozzle configuration is selected to be implemented in ANSYS Forte of fuel injected per cycle gs and mfull is the total mass of fuel injected per
since it is a widely used layout and easy to reproduce. To achieve this,
cycle (g).
the dispensed jets from each injector nozzle must be modelled, being
Finally, the simulation plan contains the values of λ obtained from
indispensable to give the exit position and the polar and cylindrical
Eq. (1) for each engine speed (rpm), the SA (dBTDC), vi (m/s), PW
coordinates defining their trajectory, as can be seen in Fig. 2.
(CAD), ṁf (g/s), the injection pressure (PInj ) and the SOI and EOI, pre-
Additionally, simulation time is saved by assessing a single engine
cylinder. This simplifies the simulation and power extrapolation for the sented in Table 5, comparing analogous experiences between DI and PFI.
In this study the simulations were executed in a workstation which
uses a message passing interface (MPI) solver with two processors Intel®
Table 3 Xeon® Gold 6148 of 20 physical cores each one and 256 GB of RAM.
Air inlet critical boundary conditions: Pressure and Temperature [29]. Furthermore, to maximize the efficiency of the simulation workflow a
Python script to post-process the ANSYS Forte results has been
Air Inlet
developed.
Speed (rpm) Pressure (bar) Temperature (K)
2000 0.92 314
3000 0.91 311 2.4. Engine performance basis
4000 0.89 309
5000 0.88 307
Low carbon fuel performance was compared at optimum SA and

929
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

Fig. 2. Spray injection geometry at the simulation model.

Fig. 3. General simulation procedure.

Table 5
Simulation plan.
Direct Injection (DI)

Fuel λ RPM SA (dBTDC) DI PFI

PInj (bar); Vi (m/s) PW (CAD) ṁf (g/s) EOI (◦) PInj (bar); Vi (m/s) PW (CAD) ṁf (g/s) SOI (◦)

H2 1.5 2000 10 100; 494 15 5.83 630 3; 275 93 0.30 485

3000 10 22 5.83 600 150 0.30 490
4000 15 34 5.83 570 216 0.30 495
5000 15 43 5.83 540 278 0.30 500
H2 2 2000 25 12 5.83 630 76 0.30 485
3000 25 17 5.83 600 120 0.30 490
4000 25 26 5.83 570 179 0.30 495
5000 25 33 5.83 540 225 0.30 500
CH4 1.5 2000 45 100; 207 13 19.43 630 3; 115 69 1.0 485
3000 60 20 19.43 600 107 1.0 490
4000 60 31 19.43 570 164 1.0 495
5000 60 39 19.43 540 201 1.0 500
COG 1.5 2000 25 100; 234 16 13.60 630 3; 150 16 0.7 485
3000 25 25 13.60 600 25 0.7 490
4000 25 40 13.60 570 40 0.7 495
5000 25 49 13.60 540 49 0.7 500

wide-open throttle (WOT) conditions to reach the maximum brake tor- and thus, reduced torque. This phenomenon is caused by the lower H2
que (MBT) across all speeds and lambda ratios, using identical initial concentration in the air-fuel mixture.
values. In this context, higher values of λ leads to brake power reduction, Conversely, the higher air mass fosters an increase in volumetric

930
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

efficiency (ηe ), the key variable to assess the different performance of DI ( )

and PFI. This improvement in ηe stems from the elevated O2 proportion, P⋅ηR
Wc N 1
as consequence of the greater air intake during the admission stroke. For BTE = =( ) = (9)
mf ⋅LHV ṁf ⋅ηR ⋅LHV BSFC⋅LHV
this reason, ηe and the air mass inlet flow (ṁe ) are directly correlated, as N
depicted in Eq. (5) [39].
ṁNOx
2⋅ṁe ma sNOx = (10)
ηe = = (5) P
ρa.i ⋅Vd ⋅N ρa.i ⋅Vd
CHR = HR (TA ) − HP (TA ) (11)
Where ρa.i is the air density at atmospheric conditions (g/cm3), ṁe is the
air mass inlet flow (g/h) and ma is the real air mass entering into the Where ṁf is the mass flow of fuel injected per cycle (g/s), LHV is the
cylinder (g). lower heating value in mass base (J/g), ṁNOx is the mass flow of
The selected parameters to characterize the engine performance, generated NOx, HR is the reagent enthalpy (J) and HP the product
power (P), brake mean effective pressure (BMEP) and torque are enthalpy (J), both of them at ambient temperature (TA ) (K).
described by the following expressions presented in Eqs. (6), (7)and (8) Finally, regarding NOx formation, which is commonly found in H2
[39]. combustion, this process is primarily governed by temperature, ac-
Wc,i ⋅N cording to the extended Zeldovich reaction model [40]:
P= = 2⋅π⋅N⋅T⋅10− 3
(6)
ηR N2 + O ↔ NO + N
P⋅ηR Wc,i N + O2 ↔ NO + O
BMEP = = (7)
Vd ⋅N Vd
∮ N + OH ↔ NO + H
Wc,i = p⋅dV (8)
3. Results and discussion
Where ηR is the number of crank revolutions for each power stroke per
cylinder (ηR = 2), Wc,i is the work delivered per cycle (J/Cycle), Vd is the This section presents the major outcomes derived from the simula-
cylinder volume displaced by the piston in all the cylinders (dm3), p is tions that compare the performance of DI and PFI engines in terms of
the pressure obtained during compression stroke (bar) and N is the en- brake power, torque, BMEP, thermal and volumetric efficiencies, spe-
gine speed (rev/s). cific fuel consumption, heat release and sNOx emissions. In addition, the
The relationship between power and volumetric efficiency stems analysis also assesses the potential of the different low carbon fuels used
from power density. The displacement volume effect arises from the in both engines, particularly focusing on H2.
higher density of the air compared to H2. When H2 is injected into an
intake air stream, its lower density relative to air leads to occupy a
greater volume in a certain physical space, displacing some of the sur- 3.1. Hydrogen engines performance comparison: DI vs PFI
rounding air and hindering the entry into the combustion cylinder.
On the other hand, after adaptation to DI, the full air stream is 3.1.1. Engine performance
introduced, followed by DI of H2 when inlet valve is completely closed, First, the maximum pressure reached during combustion exhibits the
increasing the amount of air-fuel mixture per cm3 of cylinder volume for power delivered. In this case, the pressure curves for H2 at different
the same λ, promoting the volumetric efficiency. Nevertheless, H2 is engine speeds and λ ratios of both engines are shown in Fig. 4.
injected at higher pressure (100 bar) than the injection pressure in PFI In H2 simulations, peak pressure was measured at engine speeds from
(3 bar), requiring prior fuel energization. The high injection pressure of 2000 rpm to 5000 rpm and λ ratios of 1.5 and 2. According to the results
100 is selected as high-speed flow injectors in DI generally requires this obtained, higher λ drives a peak pressure reduction as a consequence of
level of pressure to reach the instantaneous flow with H2 [30,42]. lower enriched air-fuel mixtures. Conversely, peak cylinder pressure
However, a primary drawback of this approach is a less stratified increases with engine speed (rpm). In this respect, it is important to
mixture, which could negatively impact combustion efficiency, brake- highlight that both engines have been reproduced using the same spark
specific fuel consumption, and NOx emissions. advance (SA) on each simulation experience, varying in the same pro-
In this way, torque represents the rotational force generated by the portion based on λ factor.
engine, produced after combustion during the expansion stroke. The Thus, in the case of PFI, working at λ = 1.5 and 5000 rpm results in a
release of chemical heat leads to torque generation, which is subse- peak pressure of 65.4 bar, while DI obtains 79.4 bar. Consequently, peak
quently converted into power. Additionally, BMEP measures engine pressure fuelling H2 at optimum λ (λ = 1.5) is among 8.7–21.4% higher
efficiency, defined as the mean effective pressure exerted over the piston using DI than PFI at different engine speeds.
and developed during combustion. As a result, both variables are crucial Furthermore, this figure illustrates the disparity in SA between
to evaluate engine performance, serving as a reference of the extent to simulations conducted at λ = 1.5 and λ = 2 in both engines. In this
which the various forces exerted as a consequence of the combustion context, simulations at λ = 2 exhibit an advancement of 15 crank angle
reaction in an ICE are utilized. degrees (CAD) compared to λ = 1.5 simulations. Consequently, when
Furthermore, the engine performance, especially in H2-ICEs, apart operating under leaner conditions (λ = 2), the combustion reaction
from the widely known parameters of power, torque and BMEP, must be initiates earlier, and the pressure peak is reached close to the top dead
also assessed in terms of BTE and brake specific fuel consumption centre (TDC), resulting in a smoother slope on the curve. On the other
(BSFC), related to peak temperatures reached during combustion stroke hand, conducting simulations at the optimal air-fuel ratio (λ = 1.5)
and the specific nitrogen oxides emission levels (sNOx), most of them portrays a steeper slope, from 710 CAD at 2000–3000 rpm and 705 CAD
directly or inversely related with brake power and chemical heat release at 4000–5000 rpm. This abrupt increase means that the combustion
(CHR), as it was defined in Eqs. (9), (10) and (11) [39]. reaction occurs in less time than under leaner conditions as consequence
of the higher flame temperature generated, which promotes the laminar
flame speed (SL ), leading to higher pressure peaks. Therefore, a more
efficient combustion reaction with a faster release of energy and higher
combustion temperatures is achieved.

931
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

Fig. 4. H2 -ICE pressure analysis during combustion at different air-fuel ratios a) λ = 1.5 and b) λ = 2.

It is worth noting that there is a direct relationship between pressure taking place.
and power which also affects to torque and BMEP by means of Wc,i . The Accordingly, an increase in engine speed conducts to greater intake
higher peak pressure in DI is enabled by the enhancement in volumetric and expulsion of the air-fuel mixture, resulting in combustion reactions
efficiency as consequence of a better distributed and homogeneous that generate more thermal energy in each cycle. As a result, higher
mixture inside the cylinder, which increases the power density. How- engine speeds yield more power. Furthermore, this leads to an enhanced
ever, the increase in pressure leads to a simultaneous increase in tem- volumetric efficiency using DI configuration, which ranges from 73.6%
perature, promoting the formation of thermal NOx. to 93.1% and a brake power that ranges from 4.2 kW to 12.8 kW within
In this context, Fig. 5 presents the key engine performance metrics, that engine speed range for just one cylinder analysed. Consequently,
encompassing brake power depicted as lines, volumetric efficiency considering the 4-cylinder engine configuration, an approximately 11
represented as columns, torque and BMEP, which are influenced by kW more powerful engine is obtained adapting the configuration from
engine speed (rpm) and λ ratio. PFI to DI, assuming a total increase of 14.8 horse power (HP).
In this case, simulations conducted at optimal air-fuel ratio (Fig. 5a, Conversely, simulations performed under leaner conditions (Fig. 5b,
λ = 1.5) reveal a notable discrepancy in volumetric efficiency between λ = 2), yield lower performance as less air is introduced per cycle,
PFI and DI, ranging from 18.3% to 30.6% higher in the case of DI. This thereby reducing the available energy during combustion. However, in
difference persists across the entire engine speed range, correlating with case of DI, compared to optimal air-fuel ratio (Fig. 5a, λ = 1.5), volu-
a corresponding power increase between 27.3% and 39.5%. The metric efficiency remains unaffected since the amount of air introduced
observed difference among type of engines arises as a result of the keeps constant while fuel mass is reduced. On the other hand, in PFI
displacement volume effect. In PFI configuration, the process impedes configuration, under leaner conditions, as consequence of the lower
air intake due to the lower density of H2, requiring a larger volume per amount of fuel injected, a higher amount of air is introduced inside of
mass unit. Consequently, a reduced quantity of air-fuel mixture is the cylinder enhancing volumetric efficiency, although the lower fuel
introduced to attain the desired λ ratio, and hence, control over the concentration in mixture causes a power output detriment. Conse-
dosage is more complex. This leads to a decline in power density and quently, brake power decreases when operating at λ = 2 compared to λ
simultaneously a decrease in the released energy while combustion is = 1.5. Specifically, DI delivers among 35.6%–41.4% less power from

932
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

Fig. 5. H2 engine performance comparison at different air-fuel ratios a) λ = 1.5 and b) λ = 2.

2000 to 5000 rpm, while the PFI engine experiences a reduction from combustion temperature reached across the entire range of engine
19.3% to 25.4%. speeds.
Thus, DI configuration achieves closer results, in terms of brake In this context, the elevation in engine speed involves an increased
power, to the PFI configuration operating at λ = 2, although there is still chemical heat release (CHR) per cycle due to the growth in the air-fuel
a gap between them ranging from 12.8% to 26.5% in favour of DI. mixture introduced. Moreover, brake specific fuel consumption (BSFC)
Finally, torque and BMEP are plotted against various engine speeds rises despite the power gain as consequence of the higher injection
and air-fuel ratios, operating at λ = 1.5 (Fig. 5a) and λ = 2 (Fig. 5b). flowrate (ṁf ). Therefore, as engine power increases, BTE diminishes as a
Under optimal conditions (Fig. 5a, λ = 1.5), a comparable pattern is result of the indirect correlation with BSFC.
noted in both variables, mirroring the trends observed in the brake In the obtained results, simulation runs at the optimal air-fuel ratio
power results because both are directly related, with torque and BMEP (λ = 1.5), H2 attains elevated CHR levels using a DI engine in compar-
in DI being 27.3%–39.5% greater than PFI. In contrast, under lean ison with PFI configuration. This is attributed to the enhanced capacity
conditions (Fig. 5b, λ = 2), torque and BMEP experienced reductions of DI to introduce a greater quantity of air-fuel mixture, culminating in
ranging from 34.9% to 40.7% compared to optimal λ. However, DI higher CHR values, effectively transforming more chemical energy into
configuration results still exhibit a superiority of 12.8%–26.5% over PFI. mechanical energy. Consequently, the utilization of DI instead of PFI
leads to percentage increase in CHR ranging from 19.0% to 35.5%,
3.1.2. Thermal efficiency and emissions which results in an additional CHR amount of 883.3 J/cycle when
Once engine performance parameters are assessed, as consequence of considering the entire 4-cylinder engine.
the higher performance achieved at the optimal air-fuel ratio (λ = 1.5), The rise in brake power requires a widening pulse width (PW) as a
this value is predetermined for analysing thermal efficiency and emis- consequence of the corresponding increase in the amount of fuel mass
sions behaviours in both engine configurations. injected. Thus, BSFC values elevate as engine speed increases. Never-
As illustrated in Fig. 6, the selected variables for describing thermal theless, the lower power output of PFI engines negatively impacts BSFC,
efficiencies are the chemical heat release (CHR), brake thermal effi- despite their lower fuel consumption per cycle compared to DI engines.
ciency (BTE) and brake specific fuel consumption (BSFC), which is In this context, DI results in a percentual reduction of BSFC compared to
represented by the blue line. Fig. 6 also depicts, sNOx and maximum PFI, ranging from 1.4% to 4.8%.

933
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

combustion, converting more energy released into useful work instead

of remanent heat [41].
Moreover, the reduction of WHT losses leads to better mixing, and
hence, the local peak temperatures in the cylinder are reduced [43].
Therefore, an increase in CHR using the DI configuration compared to
PFI suggests an increase in useful power, but not a temperature increase,
as WHT losses are reduced.
Thus, the maximum temperature reached in DI is reduced between
25.5K and 73.9K, together with a decrease in sNOx ranging from 10.4%
to 36.0% compared to PFI. In addition, DI system reaches the minimum
levels of sNOx at 3000 rpm (14.06 g/kWh), as consequence of the higher
brake power augmentation in comparison with 2000 rpm.
In the PFI configuration, the air excess burns a larger quantity of H2,
which possesses a higher LHV, further elevating the mixture tempera-
ture at high engine speeds and subsequently promoting the formation of
thermal NOx. For instance, a peak temperature rise of 266.1 K is
observed from 3000 rpm to 4000 rpm.
Consequently, it can be concluded that the H2-ICE DI exhibits su-
perior power output and lower emission levels compared to PFI, under
the studied conditions.

3.2. Low carbon fuels performance comparison: DI vs PFI

3.2.1. Engine performance

Ansys® Forte® software enables the assessment of different fuel
types, using the same variables and parameters. This facilitates the
comprehensive analysis of the performance, including previous H2 re-
sults, and those obtained in CH4 and COG simulations at (λ = 1.5). Fig. 7
depicts the comparison of the performance of the three fuels at the H2
optimal air-fuel ratio (λ = 1.5) within the entire range of engine speeds
(2000–5000 rpm). Thus, brake power, volumetric efficiency, torque and
BMEP are assessed for both injection systems.
In PFI configuration (Fig. 7a), the maximum power output achieved
is 9.5 kW when utilizing COG at 5000 rpm. The superior brake power
observed in COG is attributed to the higher SL compared to CH4, leading
to higher efficiencies since a more complete combustion is achieved.
Conversely, COG has a greater density than H2 and a substantial amount
of energy is delivered during combustion owing to higher volumetric
efficiency. In this sense, compared to H2, brake power increases from
0.9% (+0.1 kW/cyl) to 12.7% (+0.6 kW/cyl), and compared to CH4
brake power spans from 10.0% (+0.3 kW/cyl) to 30.2% (2.86 kW/cyl)
in favour of COG within the entire range of engine speeds.
In addition, the obtained volumetric efficiency is among 7.5%–
10.8% greater running COG in PFI engines than H2. The enhanced
volumetric efficiency stems from the higher volumetric density of COG,
which is in mass % primarily composed by CH4. This increased density
Fig. 6. Heat released, thermal efficiency and NOx emission comparison in mitigates the effect of volume displacement at the inlet, leading to gains
H2 engines. in volumetric efficiency. Furthermore, COG volumetric efficiency re-
mains lower than CH4 ranging from 4.7% to 6.7%. The volumetric ef-
Conversely, BTE and BSFC display strong interdependency, with ficiency of COG reaches the maximum at 4000 rpm, resulting in the
brake power significantly influencing them. In this sense, BTE di- highest power difference between fuels. Therefore, PFI system finds a
minishes with the increase in BSFC, stemming from the heightened fuel performance limitation on 4000 rpm, reaching almost the same peak
requirement to generate an equivalent amount of energy, resulting in pressure as seen in the latter simulation at 5000 rpm while injecting
greater losses. Subsequently, upon analysing the entire range of engine more amount of fuel and reducing the overall efficiency. This limitation
speeds (rpm), the utilization of DI configuration contributes to the is overcome by changing the injection system to DI, which means that DI
increment of BTE between 0.5% and 1.7% compared to PFI. Specifically, engine could even work at higher speeds resulting in a performance
H2 attains a peak BTE of 37.6% and a minimum BSFC of 79.7 g/kWh. improvement.
Finally, NOx emissions exhibit a direct proportional relationship Regarding DI configuration (Fig. 7b), H2 achieves a maximum power
with combustion temperature, which increase with the engine speed as output of 12.8 kW at 5000 rpm. Moreover, H2 obtains the highest brake
consequence of the greater amount of heat released during combustion. power using the DI configuration, even generating a power output
In the DI system, air is introduced before hydrogen injection, 21.9% higher than COG and 78.9% higher than CH4 both fuelled in DI
reducing the tendency to pre-ignition and also lowering the local in- engines. Thus, CH4 is greatly disadvantaged while working over the λSt ,
cylinder temperature, as consequence of H2 expansion derived from where CH4 optimal combustion conditions are normally achieved.
high-pressure direct injection [27]. This leads to the formation of a more This power enhancement in H2 combustion simulations is obtained
homogeneous air-fuel mixture, fostering more efficient and uniform as consequence of the growth of volumetric efficiency. In this context,
H2 exhibits the greatest sensitivity to engine changes in volumetric

934
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

Fig. 7. Low carbon fuels engine performances comparison with different engine configurations a) PFI and b) DI.

efficiency among the analysed fuels. However, in every case the volu- and PFI configurations. CHR, wall heat transfer (WHT) and maximum
metric efficiency is improved, and hence, the brake power. In accor- combustion temperature are represented by bars, and, sNOx are repre-
dance, H2 increases volumetric efficiency by 30.6% in the maximum, sented by trend lines. Finally, the comparison between BTE and BSFC is
COG by 14.8% and CH4 by 9.4%, since H2 has the lowest density and also illustrated.
CH4 the highest one. Regarding the PFI configuration (Fig. 8a), it is observed that the
Finally, torque and BMEP follows the same trends as brake power assessed fuels deliver comparable CHR. However, within this configu-
and volumetric efficiency. Accordingly, the maximum torque in the PFI ration, COG achieves the lowest wall heat transfer (WHT) values at
configuration is achieved when employing COG as fuel and operating at optimal air-fuel ratio (λ = 1.5). Consequently, H2 experiences higher
4000 rpm due to the higher brake power reached at this engine speed, wall losses than COG, from 2.9 to 20.3 J/cycle, while CH4 incurs losses
generating 19.2 N m and achieving a maximum BMEP of 6.9 bar. In 23.6–70.6 J/cycle above COG levels. This discrepancy in WHT con-
contrast, the DI configuration reaches the maximum torque and BMEP tributes to COG and H2 showing significantly higher brake power than
using H2 at 5000 rpm, resulting in 24.4 N m and 8.82 bar. This means a CH4, because more heat is converted into useful work, even though CHR
noteworthy enhancement in conversion efficiency of heat released into is similar and lower combustion temperatures are achieved. In the case
useful work, as evidenced by the increase in BMEP values. This of CH4, this effect occurs due to the negative work generated as conse-
improvement indicates a more favourable response in a high-load due to quence of reaching the pressure peak before the TDC, while piston keeps
the increased torque, in addition to the greater brake power delivered. compressing and reducing the effective work.
Thus, H2 is the best fuel compared at the designed DI engine in terms In light of the above, the elevated combustion temperatures during
of engine performance with regard to the final results and improvement. CH4 combustion, foster the formation of thermal NOx, resulting in the
Conversely, it is important to highlight that COG is obtained after val- highest observed sNOx levels. In contrast, the mitigating effect of CH4
orisation of an industrial residual gas stream, although it has interme- when mixed with H2 leads to a reduction in sNOx emissions during COG
diate properties between H2 and CH4 regarding density, LHV, and SL . combustion, showcasing reductions ranging from 31.9% to 111.6%
compared to H2. In this sense, CH4 results in higher combustion tem-
3.2.2. Emissions and thermal efficiency peratures but also in greater WHT, as consequence of the lower BTE.
As illustrated in Fig. 8, the performance and the efficiency are Regarding efficiency, within the PFI configuration, H2 is emerging as
assessed together with the emissions associated with each fuel used in DI the fuel with the minimum energy consumption for power generation,

935
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

Fig. 8. Performance, efficiency and sNOx emissions comparison in low carbon fuels with different engine configurations a) PFI and b) DI.

also manifests a high thermal efficiency similar to the COG. Further- addition, when compared to PFI, the WHT for H2 is reduced from 2.0%
more, relating to BSFC, H2 exhibits values between 65.5%-66.5% and to 10.9%, COG spans from 10.2% to 19.1% and for CH4 from 0.9% to
62.5%–69.9% lower values than COG and CH4, respectively. 28.7%. This reduction in thermal losses is attributed to the improved
In DI simulations (Fig. 8b) working under identical conditions (λ = combustion reaction facilitated by DI strategies, mitigating the effect of
1.5), since H2 obtains the highest power output, it also releases the incomplete reaction zones that release inefficient heat.
greatest amount of chemical heat, exceeding the CHR obtained for the In the case of sNOx, emission levels increase considerably at high
COG at the same λ between 16.4% and 28.8% and showing the higher speeds, reaching in CH4 simulations a value of 78.3 g/kWh that is almost
chemical combustion potential of H2 compared to other fuels. twice the sNOx emissions of DI H2 at 5000 rpm.
Moreover, when simulating within this injection system, COG pre- In the case of H2 simulations, the maximum emissions value is 29.6
sents intermediate results between H2 and also CH4 for DI. This results in g/kWh, representing a reduction of 62.1% in sNOx compared to CH4.
an improvement of CHR by 9.4% at maximum engine speed after However, this value is 55.7% higher than the emissions generated using
enhancing the volumetric efficiency of the air-fuel mixture through the COG. Consequently, the inclusion of a mixture of CH4 and H2, repre-
transition from PFI to DI configuration. sented in COG simulations exhibits NOx-reducing properties, achieving
However, the combustion of H2 results in greater energy losses the lowest sNOx levels of the three, even at maximum engine speed,
through the walls, that exceed of those of COG by 4.2–25.7 J/cycle, and measuring 14.9 g/kWh at 5000 rpm. Additionally, COG reduces the
CH4 also experiences losses from 8.7 to 65.3 J/cycle above COG. In maximum combustion temperature in the range of 219.0 K–305.0 K

936
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

compared to H2. ⁃ BTE is very similar in COG and H2 simulations in both injection
Therefore, DI systems contributes to reduce sNOx emission levels in systems, while CH4 achieves the lowest values, losing big amounts of
all scenarios. In summary, DI reduces sNOx emissions 7.9 g/kWh for H2 thermal energy during combustion in excess of air.
at 3000 rpm, 4.2 g/kWh for COG at 4000 rpm and 30.3 g/kWh for CH4 at
3000 rpm, compared to PFI. These findings underscore the necessity of further research on DI H2-
Finally, regarding BTE and BSFC, H2 achieves the best values, similar ICEs, considering the evident improvements in engine performance and
in DI and PFI configurations. However, H2 reduces BSFC in the range of the high levels of efficiency achieved. DI H2-ICEs demonstrate compet-
62.4%–71.8% compared to CH4 and 65.8%–67.2% compared to COG. In itiveness not only with dedicated H2 engines but also with other low
terms of BTE, both H2 and COG achieve comparable results. However, carbon fuels and even gasoline engines in terms of power output and
H2 consistently outperforms CH4 by a margin of 3.7%–11.3%. This su- emissions generation. Moreover, DI technology offers ease of imple-
periority stems from the exceptional calorific power per unit mass of H2 mentation compared to other technologies and the ability to employ
and remarkable thermal efficiencies, which facilitate a more efficient synthetic fuels or even fuels recovered from waste streams such as COG.
energy potential conversion. Therefore, despite the need for high These advantages position DI engines as frontrunner in the decarbon-
compression of H2 to achieve autonomy, H2 remains the preferred ization of road transport.
choice of fuel, specially achieved in DI engines.
CRediT authorship contribution statement
4. Conclusions
Fabián Musy: Data curation, Formal analysis, Investigation, Meth-
In this work a comparative simulation study of the engine perfor- odology, Software, Writing – original draft. Rafael Ortiz: Conceptuali-
mance in DI and PFI has been carried out, both adapted to be fuelled by zation, Investigation, Methodology, Software, Supervision. Inmaculada
different low carbon fuels H2, CH4 and COG. The simulation model has Ortiz: Funding acquisition, Investigation, Supervision. Alfredo Ortiz:
been developed based on a naturally aspirated Volkswagen Polo 1.4 L SI Conceptualization, Data curation, Formal analysis, Funding acquisition,
engine. Thus, simulations were focused on two main parameters with Investigation, Methodology, Supervision, Writing – original draft,
influence on engine performance, air-fuel ratio (λ = 1-2) and engine Writing – review & editing.
speed (2000–5000 rpm).
Regarding H2 simulations, the main conclusions drawn are. Declaration of competing interest

⁃ A CFD model was developed for DI H2-ICE, showing a performance The authors declare that they have no known competing financial
improvement. Simulations reveal that working under optimal air- interests or personal relationships that could have appeared to influence
fuel ratio conditions (λ = 1.5), the DI H2-ICE model increases a the work reported in this paper.
brake power by 39.5% and the volumetric efficiency by 30.6%
compared to the PFI system. These findings corroborate the detri-
Acknowledgement
mental effect of volume displacement on engine performance, a
characteristic limitation of PFI gas fuel combustion systems. On the
This research was supported by the European Project HYLANTIC
other hand, torque and BMEP increases in the same proportion.
EAPA_204/2016, which is co-financed by the European Regional
⁃ Concerning sNOx emissions, the DI offers a prominent NOx-
Development Fund (ERDF) in the framework of the INTERREG
reduction potential, emitting a maximum of 29.6 g/kWh, which
ATLANTIC program. The Spanish Ministry of Science and Innovation is
suppose a 36% reduction compared to PFI. This reduction is achieved
also supporting this investigation through the projects PID2021-
through the lower value of the maximum temperature reached and
123120OB-I00, TED2021-129951B–C21, and “Complementary Plan for
the higher power delivered which is inversely proportional to sNOx.
Energy and Renewable Hydrogen” additionally supported with funding
⁃ Using DI configuration, H2 attains a peak BTE of 37.6% and a min-
from European Union NextGenerationEU (PRTR-C17. I1) and by
imum BSFC of 79.7 g/kWh, performing better than PFI. In addition,
Comunidad Autónoma de Cantabria.
the utilization of DI instead of PFI leads to a 35.5% more energy
released, which results in an additional CHR amount of 883.3 J/cycle
References
when considering the entire 4-cylinder engine.
⁃ Leaner H2 mixtures (λ = 2) experience lower combustion pressures [1] European Commission. The European Green Deal 2019;53(9):24 [Online]
and temperatures, leading to reduced performance. Specifically, DI Available: EUR-Lex - 52019DC0640 - EN - EUR-Lex ([Link]).
delivers among 35.6%–41.4% less power. [2] Sens M, Danzer C, von Essen C, Brauer M, Wascheck R, Seebode J, Kratzsch M.
Hydrogen Powertrains in competition to fossil fuel based internal combustion
engines and battery electric powertrains. Environ Sci Eng 2021 [Online] Available:
On the other hand, comparing the performances obtained by Muster einer ersten Seite ([Link]).
different fuels. [3] Karagöz Y, Balcı Ö, Köten H. Investigation of hydrogen usage on combustion
characteristics and emissions of a spark ignition engine. Int J Hydrogen Energy
2019;44(27):14243–56. [Link]
⁃ H2-DI simulations running at the optimum air-fuel ratio (λ = 1.5) [4] Al-Baghdadi M. An Overview of Hydrogen as an Alternative Fuel Experimental and
show the best engine performance, generating a power output 21.9% numerical investigation of the dynamic characteristics of laminated composite
plate hybrid with steel View project PEM Fuel Cells R&D View project. 2020
and 78.9% higher than COG and CH4 respectively. In these cases,
[Online] Available: An Overview of Hydrogen as an Alternative Fuel |
simulations present the same differences in torque and BMEP. Encyclopedia MDPI.
⁃ COG exhibits the lower sNOx levels since the mixture H2–CH4 greatly [5] Verhelst S, Sierens R, Verstraeten S. A critical review of experimental research on
hydrogen fueled SI engines. SAE Int. J. Engines 2006;115:264–74. [Link]
reduces NOx while keeping good performances. However, H2-DI
10.4271/2006-01-0430.
configuration still reduces the sNOx levels by 62.1% compared to [6] Karim GA. Hydrogen as a spark ignition engine fuel. Int J Hydrogen Energy 2003;
pure CH4 at λ = 1.5. 28:569–77. [Link]
⁃ Regarding BSFC, the H2-DI combination at λ = 1.5 achieves the most [7] Dimitriou P, Tsujimura T. A review of hydrogen as a compression ignition engine
fuel. Int J Hydrogen Energy 2017;42(38):24470–86. [Link]
favourable performance. Thus, H2 is the more efficient option ijhydene.2017.07.232.
resulting in a reduction of 71.8% in BSFC compared to CH4 and [8] Stępień Z. A comprehensive overview of hydrogen-fueled internal combustion
67.2% compared to COG at the same operating conditions, due to the engines: achievements and future challenges. Energies 2021;14(20):6504. https://
[Link]/10.3390/en14206504.
higher LHV per mass unit. [9] Ortiz-Imedio R, Ortiz A, Urroz JC, Diéguez PM, Gorri D, Gandía LM, Ortiz I.
Comparative performance of coke oven gas, hydrogen and methane in a spark

937
F. Musy et al. International Journal of Hydrogen Energy 137 (2025) 925–938

ignition engine. Int J Hydrogen Energy 2021;46(33):17572–86. [Link] [26] Nemati A, Fathi V, Barzegar R, Khalilarya S. Numerical investigation of the effect
10.1016/[Link].2019.12.165. of injection timing under various equivalence ratios on energy and exergy terms in
[10] Boretti A. Hydrogen internal combustion engines to 2030. Int J Hydrogen Energy a direct injection SI hydrogen fueled engine. Int J Hydrogen Energy 2013;38(2):
2020;45(43):23692–703. [Link] 1189–99. [Link]
[11] Yang Z, Zhang F, Wang L, Wang K, Zhang D. Effects of injection mode on the [27] Zanforlin S, Frigo S. Implementation of a novel hydrogen direct-injection concept
mixture formation and combustion performance of the hydrogen internal in single and multi-cylinder engines: CFD, experimental and engine powertrain
combustion engine. Energy 2018;147:715–28. [Link] design studies. Int J Powertrains 2014;3(1):129–50. [Link]
energy.2018.01.068. IJPT.2014.059411.
[12] Kosmadakis GM, Rakopoulos CD, Demuynck J, De Paepe M, Verhelst S. CFD [28] Maccarley CA. Study of factors influencing thermally induced backfiring in
modeling and experimental study of combustion and nitric oxide emissions in hydrogen fueled engines, and methods for backfire control. Environ Eng Sci 1981;2
hydrogen-fueled spark-ignition engine operating in a very wide range of EGR rates. (16):1449–53. [Link]
Int J Hydrogen Energy 2012;37(14):10917–34. [Link] [29] Ortiz-Imedio R, Ortiz A, Ortiz I. Comprehensive analysis of the combustion of low
ijhydene.2012.04.067. carbon fuels (hydrogen, methane and coke oven gas) in a spark ignition engine
[13] Duan J, Liu F, Yang Z, Sun B, Chen W, Wang L. Study on the NOx emissions through CFD modelling. Energy Convers Manag 2022;251:114918. [Link]
mechanism of an HICE under high load. Int J Hydrogen Energy 2017;42(34): org/10.1016/[Link].2021.114918.
22027–35. [Link] [30] Verhelst S, Wallner T. Hydrogen-Fueled internal combustion engines. Prog Energy
[14] Diéguez PM, Urroz JC, Sáinz D, Machin J, Arana M, Gandía LM. Characterization of Combust Sci 2009;35:490–527. [Link]
combustion anomalies in a hydrogen-fueled 1.4 L commercial spark-ignition [31] Lee S, Kim G, Bae C. Effect of mixture formation mode on the combustion and
engine by means of in-cylinder pressure, block-engine vibration, and acoustic emission characteristics in a hydrogen direct-injection engine under different load
measurements. Energy Convers Manag 2018;172:67–80. [Link] conditions. Appl Therm Eng 2022;209:118276. [Link]
[Link].2018.06.115. applthermaleng.2022.118276.
[15] Verhelst S, Maesschalck P, Rombaut N, Sierens R. Increasing the power output of [32] Olavson LG, Baker NR, Lynch FE, Mejia LC. Hydrogen fuel for underground mining
hydrogen internal combustion engines by means of supercharging and exhaust gas machinery. SAE Int. J. Engines 1984;93:109–24. [Link]
recirculation. Int J Hydrogen Energy 2009;34(10):4406–12. [Link] [33] Luo Q, Hu JB, Sun BG, Liu FS, Wang X, Li C, Bao LZ. Experimental investigation of
10.1016/[Link].2009.03.037. combustion characteristics and NOx emission of a turbocharged hydrogen internal
[16] Faizal M, Chuah LS, Lee C, Hameed A, Lee J, Shankar M. Review of hydrogen fuel combustion engine. Int J Hydrogen Energy 2019;44:5573–84. [Link]
for internal combustion engines: towards Carbon-Free Combustion. J. Mech. Eng. 10.1016/[Link].2018.08.184.
Res. Dev 2019;42(3):36–46. [Link] [34] Barış O, Güler İ, Yaşgül A. The effect of different charging concepts on hydrogen
[17] Tang X, Kabat DM, Natkin RJ, Stockhausen WF, Heffel J. Ford P2000 hydrogen fuelled internal combustion engines. Fuel 2023;343:127983. [Link]
engine dynamometer development. SAE Int. J. Engines 2002. [Link] 10.1016/[Link].2023.127983.
10.4271/2002-01-0242. [35] Wallner T, Matthias NS, Scarcelli R. Influence of injection strategy in a high-
[18] Eichlseder H, Wallner T, Freymann R, Ringler J. The potential of hydrogen internal efficiency hydrogen direct injection engine. SAE Int. J. Engines 2011;5(1):
combustion engines in a future mobility scenario. SAE Int. J. Engines 2003. 289–300. [Link]
[Link] [36] Wallner T, Nande AM, Naber JD. Study of basic injection configurations using a
[19] Shinde BJ, Karunamurthy K. Recent progress in hydrogen fuelled internal direct-injection hydrogen research engine. Proceedings of the 2nd International
combustion engine (H2ICE) - a comprehensive outlook. Mater Today Proc 2022;51: Conference on Manufacturing Engineering, Quality and Production Systems 2010.
1568–79. [Link] [Link]
[20] Matthias NS, Wallner T, Scarcelli R. A hydrogen direct injection engine concept [37] Smith GP, Golden DM, Frenklach M, Moriarty NW, Eiteneer B, Goldenberg M,
that exceeds U.S. DOE light-duty efficiency targets. SAE Int. J. Engines 2012;5(3): Bowman CT, Hanson RK, Song S, Gardiner WC Jr., Lissianski VV, Qin Z. GRI-MECH
838–49. [Link] Mechanism 3.0. [Internet]. University of Berkeley, California. Available from:
[21] Mohammadi A, Shioji M, Nakai Y, Ishikura W, Tabo E. Performance and [Link]
combustion characteristics of a direct injection SI hydrogen engine. Int J Hydrogen [38] R1 ANSYS Forte theory manual. ANSYS; 2021.
Energy 2007;32(2):296–304. [Link] [39] Heywood JB. Internal combustion engine fundamentals. McGraw Hill Series in
[22] Oikawa M, Ogasawara Y, Kondo Y, Sekine K, Takagi Y, Sato Y. Optimization of Mechanical Engineering 1988.
hydrogen jet configuration by single hole nozzle and high-speed laser [40] Stan LC, Mitu DE. Simplified mechanism used to estimate the NOx emission of
shadowgraphy in high pressure direct injection hydrogen engines. SAE Int. J. Diesel engine. SAE International 2009;2(1):1221–30. [Link]
Engines 2011. [Link] 2009-01-1418.
[23] Yip HL, Srna A, Chun Yin Yuan A, Kook S, Taylor RA, Heng Yeoh G, Medwell PR, [41] Sahoo S, Srivastava DK. Numerical analysis of performance, combustion, and
Nian Chan Q. A review of hydrogen direct injection for internal combustion emission characteristics of PFI gasoline, PFI CNG, and DI CNG engine. Energy
engines: towards carbon-free combustion. Appl Sci 2019;9(22):4842. [Link] 2023;278:127749. [Link]
org/10.3390/app9224842. [42] Welch A, Mumford D, Munshi S, Holbery J, Boyer B, Younkins M, Jung H.
[24] Nakagawa K, Yamane K, Ohira T. Potential of large output power, high thermal Challenges in developing hydrogen direct injection technology for internal
efficiency, near-zero NOx emission, supercharged, lean-burn, hydrogen-fuelled, combustion engines. SAE International 2008. [Link]
direct injection engines. Energy Proc 2012;29:455–62. [Link] 2379.
egypro.2012.09.053. [43] Gomes Antunes JM, Mikalsen R, Roskilly AP. An experimental study of a direct
[25] Klepatz K, Rottengruber H, Zeilinga S, Koch D, Prümm W. Loss analysis of a direct- injection compression ignition hydrogen engine. Int J Hydrogen Energy 2009;34:
injection hydrogen combustion engine. SAE Int. J. Engines 2018. [Link] 6516–22. [Link]
10.4271/2018-01-1686.

938