Exploring the Oxy-Fuel Combustion in Spark-Ignition Engines for Future Clean Powerplants

In the last years, restrictions to mitigate emissions and improve the energy efficiency of internal combustion engines (ICE) have been updated in response to the adverse effects of global warming. In this framework, world economies seek to implement sustainable development plans, where the productive activity is efficiently kept while reducing environmental impact and prioritizing the decarbonization and the circular economy. An example is the Green Deal at the European Union, which has established objectives and policies related to climate and energy [1], targeting a reduction of around 40% of greenhouse gas emissions by 2030 and climate neutrality in 2050.

Considering that the electrification of automobiles and light transport vehicles is a smart partial solution, these options are not viable for certain hard-to-abate sectors regarding CO₂ emissions. Those are the heavy-duty road and maritime transport, representing respectively 25% and 13% of greenhouse gas emissions in Europe [2,3], and also some stationary applications. Thus, policies and research in these sectors are looking to develop propulsive systems with low or zero-emissions. Although alternatives such as fuel cells are being studied, the ICE might be difficult to replace in those sectors thanks to its competitive advantages in cost, ease of refueling, and autonomy. However, the traditional compression ignition engine is under discussion due to its CO₂ emissions, so the shift toward sustainable transport has promoted the use of gaseous fuels with lower carbon content, such as LPG or NG (liquid or compressed), usually in spark engines, as well as the development of new fuels with low environmental impact (e.g., biofuels and e-fuels).

On the other hand, other pollutant emissions are an important issue related to transport and air quality. Thus, the European Environment Agency reported last year [4] that 28% and 19% of the NO_x emissions are related to the road transport and maritime sector, respectively. In addition, CO emissions indicate that the road transport sector is the major responsible, with a contribution of 18% (in this case, the maritime sector represents less than 1% of the impact). Related to the generation of these emissions, CO is generated due to the incomplete combustion of hydrocarbons, which are the main components of the fuels used in ICE. Nitrogen oxides (NO_x) are generated due to the presence of nitrogen in the air and the high temperatures reached in the combustion chamber. Regarding the unburned hydrocarbons (uHC), they are generated as a consequence of an incomplete combustion process which can be originated by the flame extinction near the walls, the abnormal combustion like misfiring, or the short circuit process among others. Finally, soot is produced under rich local fuel-air conditions (⁠$λ<1$⁠) promoted by nonhomogeneous mixtures; thus, this pollutant is mainly produced in compression ignition engines or SI engines using a gasoline direct injection system. These soot emissions are usually neglected if the fuel is injected into the intake port with a port fuel injection (PFI) system [5].

In this context, oxy-fuel combustion (OFC) arises as a promising technology to counteract the aforementioned problems of the ICEs, thanks to enabling easy CO₂ capture and its potential for zero-NO_x emissions and reduced CO, uHC, and PM (in the latest case, mainly if a homogeneous mixture is used). Oxy-fuel combustion consists in a combustion process produced in a highly O₂-enriched ambient, where excessively high temperature is controlled using a dilution gas that can be the exhaust gases recirculated (EGR), mainly composed of CO₂ and H₂O vapor, instead of N₂. Among the benefits of this unconventional combustion mode, its potential for CO₂ capture and storage (CCS) must be highlighted, thanks to the fact that H₂O can be easily removed from the exhaust gases by condensation, and thus, the resultant gas is mainly composed of CO₂. This unconventional combustion mode has been studied in SI engines in the past years. First, Yu and Wu implemented experimental and simulation studies using a single-cylinder research engine with a PFI system to inject propane [6,7]. In their studies, they assessed the effect of injecting water near the top dead center (TDC) for different dry-EGR dilutions rates (i.e., only composed of CO₂). They found that this strategy has a positive influence over the efficiency, reaching values of 46%. Using this strategy, Li et al. [8] numerically studied and compared the effects of the water injection on a gasoline direct injection, PFI, and dual-injection gasoline SI engine also operating with dry EGR dilution where, in contrast with a conventional engine using air, they found an increase of the brake specific fuel consumption when oxy-fuel combustion is applied to the three fuel injection methods.

These studies and the others that can be found in the literature related to OFC in ICEs, both for compression ignition [9,10] and spark ignition [11,12], obviate one of the main technological drawbacks of oxy-fuel combustion ICE (OFC-ICE), that is the supply of O₂. It can be taken from pressurized tanks, an option that can compromise the economic and logistics viability of the concept, except when O₂ is available as a by-product. It is also considered that O₂ production can be carried out based on oxygen transport membranes (OTM), which typically operate between 700 °C and 1000 °C [13] and are currently in development [14]. Thus, based on numerical simulations, Arnau et al. [15,16] evaluated an oxy-fuel combustion engine with in situ oxygen production, coupling the propulsive system with a mixed ionic and electric conducting membrane. In their concept, the EGR that is recirculated into the cylinders is composed of CO₂ and H₂O (i.e., humid EGR), and they found that an efficiency diminish may be expected in contrast with a conventional SI engine.

Keeping in mind the current state of the art, literature shows some limitations. First, oxy-fuel combustion experimental studies have only been performed with dry EGR, but recent concept design [15] suggests that humid EGR should be considered to step this unconventional concept closer in terms of production viability. Second, there is a lack of systematic studies that assess dilution strategies with both EGR and $λ>1$⁠. Additionally, emissions and knocking issues are hardly studied in the OFC context. Finally, there is the already commented omission of oxygen generation requirements.

It is the author's opinion that this work covers, at least partially, these literature gaps. Thus, it has the novelty of studying in detail the OFC using humid EGR in combination with O₂ dilution, and considering emissions and knocking analysis in a SI engine. Moreover, in addition to thermo-mechanical engine limitations, the suitability for integrated O₂ generation is considered. To the best of the author's knowledge, no other oxy-fuel combustion study included it.

In this framework, this work pursues two main objectives: on the one hand, assessing the dilution conditions (λ and EGR) where oxy-fuel combustion can be used in a spark-ignition engine (SI), considering thermo-mechanical limitations in the chamber and knocking; on the other hand, evaluating the potential of this combustion concept in optimal conditions, from the indicated efficiency point of view, identifying its benefits and drawbacks, to finally discuss the viability of the concept taking into account the emissions and the knocking issue. These objectives have been reached following a methodology that combines experimental and theoretical tools, including zero-dimensional (0D) thermo-chemical combustion modeling, 0D-one-dimensional (1D) fluid dynamic modeling, computational fluid dynamics (CFD) calculations, and measurements in a single-cylinder research engine operating with both conventional and oxy-fuel combustion.

The paper is structured as follows: Sec. 2 presents the experimental setup and the simulation tools used along with their respective validation. Section 3 describes the methodology followed in this work to study the oxy-fuel combustion applied to ICEs. Section 4 deals with the analysis of the dilution strategies for this combustion concept taking into account the thermo-mechanical limitation of the engine. In this section, also an experimental comparison of oxy-fuel and conventional combustion is performed to evaluate the potential of the concept to finally deal with the analysis of the knocking. Finally, Sec. 5 summarizes the main conclusions of the work and remarks its main contributions.

Experimental tests were performed in a gasoline SI single-cylinder research engine equipped with a PFI system that was used to avoid additional uncertainties related to the charge inhomogeneities. Table 1 contains the main characteristics of the engine, and Table 2 shows relevant information about the combustion system.

The engine was installed in a test cell instrumented according to the scheme presented in Fig. 1. The original layout of the test bench was adapted for supplying O₂ and CO₂ from two pressurized tanks. The latter was specifically installed to provide the required oxidant dilution to keep engine integrity during the engine startup; once a steady operation was reached, the external supply of CO₂ was substituted by EGR. Moreover, the system was designed to be flexible to switch between conventional and oxy-fuel combustion operation. The valve located downstream the intake settling chamber, and the CO₂ and O₂ flow control valves are used for this purpose. For conventional combustion cases, an external compressor was used to provide the compressed air to reproduce boost conditions.

A knife-gate valve was located after the exhaust settling chamber in the exhaust line to control the exhaust back-pressure. A high-pressure EGR system provided the required levels of cooled EGR. To avoid H₂O condensation in the EGR and intake pipes (important to reduce the CO₂ measurement uncertainty and thus EGR rate determination), when the engine operates under OFC, a heat exchanger installed before the EGR settling chamber carefully controls the temperature over these lines. Oil and cooling circuits were independent of the engine, as depicted in the layout.

In-cylinder pressure was measured with a Kistler piezo-electric sensor. An additional piezoresistive pressure sensor was installed at the cylinder liner close to the bottom dead center for the pressure signal pegging. Moreover, to acquire the intake and exhaust instantaneous pressure, two piezoresistive sensors were mounted. All engine fluid temperatures were controlled and monitored during the experimental tests using K-type thermocouples. NO_x, CO, and uHC emissions were measured by a HORIBA MEXA-7100EGR gas analyzer. This system was also used to register CO₂ concentrations (to calculate the EGR rate), but in this case, a specific device was installed to dilute the gas sample to meet the specification limit of the gas analyzer when OFC is used, and N₂ is not present in the intake and exhaust flows. Complete details about the instrumentation and its accuracy are summarized in Table 3.

The global parameters related to the combustion process, i.e., the indicated mean effective pressure (IMEP), combustion phasing, maximum cylinder pressure, combustion misfiring, cycle-to-cycle variability (CoV_IMEP)_, and heat release rate (HRR), were obtained from the in-cylinder pressure signal using the thermodynamic combustion diagnosis tool CALMEC [17,18].

Auto-ignition delay (AID) and flame temperature for different fuel-oxidizer mixtures and thermodynamic conditions were estimated using zero-dimensional (0D) chemistry calculations. Data generated using a well-stirred reactor model, assuming constant pressure were used for predicting the knocking combustion tendency and maximum local temperatures. The ignition delay timing was defined as the time required for increasing 400 K from its initial temperature [19].

The chemical kinetic mechanism proposed by Liu et al. [20], based on a Primary Reference Fuel (PRF), was chosen for its good balance between accuracy and computational requirements. Benajes et al. [21] demonstrated the capabilities of this mechanism under realistic and representative engine conditions, validating its results against the experimental data reported by Fieweger et al. [22].

The laminar flame speed in a premixed oxidation reaction can be estimated by considering a freely propagating flame in a channel with a fixed cross-sectional area for a specified temperature, pressure, and mixture composition. Although this parameter is traditionally calculated by using empirical correlations [23], they systematically tend to under-predict the laminar flame speeds at realistic engine conditions [21]. Thus, a 1D laminar flame speed solver was used to get more accurate predictions under OFC conditions. Again, the chemical kinetic mechanism from Liu et al. [20] was selected due to its consistency with the experiments performed by Jerzembeck et al. [24] and Heimel et al. [25].

0D-1D tools show a good compromise between accuracy and computation time [26,27]. To characterize the premixed oxy-fuel combustion, the in-cylinder pressure and temperature, gross indicated efficiency (GIE calculated between −180 cad and 180 cad), and the exhaust temperature under different EGR dilutions were simulated with a 0D-1D tool to complement the experimental measurements. The engine and test bench layout were implemented in the GT-SUITE code. These simulations allowed reducing experimental test campaigns while evaluating different dilution strategies to implement in the engine and modeling full load conditions (in combination with CFD) for the knocking assessment. In addition, a proportional-integral-derivative controller was implemented to optimize the GIE by changing the start of combustion (SoC) while keeping the maximum in-cylinder pressure below 150 bar.

The model was initially calibrated using experimental data obtained in the single-cylinder engine operating under conventional SI conditions at 3000 rpm and for two different loads (4 and 11 bar of IMEP). In both cases, the combustion process was reproduced by imposing the heat release obtained with the combustion analysis tool [17,18] described. Results of the validation are summarized in a previous work [28], where the measured in-cylinder pressure and temperature (estimated by the combustion diagnosis tool) were compared against simulations, showing a nice agreement at 3000@4 and 3000@11 operating conditions (along this work an operating condition is described as $X@Y$⁠, where X is the engine speed and Y is the IMEP).

Computational fluid dynamics simulations were carried out using the converge v2.4 CFD software, a commercial code based on the finite volume method particularly developed for ICE applications. The model was built according to the real engine geometry, considering the combustion chamber and both intake/exhaust ports.

As shown in Fig. 2, a hexahedral grid strategy based on an orthogonal basis was used for meshing the complete computational domain with a base cell size of 4 mm. The mesh was refined up to 2 mm in the intake and exhaust and up to 1 mm in the cylinder. The cell resolution was increased (0.5 mm) near the cylinder walls, including the moving piston and valves. Additionally, an Adaptive Mesh Refinement algorithm was used to increase the grid resolution where spatial gradients of velocity and temperature surpass a determined limit. The algorithm considers a subgrid criteria of 1 m/s and 2.5 K to decrease the cell size up to a minimum of 0.125 mm. Finally, the grid resolution was further increased by reducing mesh size down to 0.0625 mm at the spark gap electrodes to capture the initial flame kernel development. Full details about the grid definition are given in Table 4.

Turbulence modeling was performed under the unsteady Reynolds-averaged Navier–Stokes framework. In particular, the Re-Normalization Group variant of the k-epsilon model (RNG k-ϵ model [29]), based on an eddy-viscosity-based two-equation turbulence model, was used. The gas-to-wall heat transfer was modeled by the approach proposed by Angelberger [30]. This combination has been widely used in ICE applications [31]. The simulations were performed using a second-order central difference scheme for spatial discretization and a first-order scheme for temporal discretization. A pressure implicit with splitting of operators algorithm modified by Issa [32] was considered for the pressure and velocity fields coupling. The ideal gas equation of state was selected for calculating the compressible flow properties. The detailed chemistry solver [33] was combined with a multizone approach for combustion modeling, considering 5 K temperature bins [34]. Previous studies [35] demonstrated the suitability of this approach for unsteady Reynolds-averaged Navier–Stokes-based gasoline combustion, even considering that it does not use an explicit turbulent combustion closure [36]. As in the 0D-1D chemistry simulations, the chemical kinetic mechanism proposed by Liu et al. [20] was used to mimic the thermo-chemical properties of the fuel. The spark kernel was modeled by a volumetric source located between the spark plug electrodes. An energy deposition of 40 mJ was spatially and uniformly distributed in a sphere of 0.5 mm along a L-type profile [37].

Experimental instantaneous pressures were set as inflow and outflow boundary conditions at the end of the intake and exhaust ports. The surface temperatures of all wall boundaries were predicted by a lumped model [38]. The model was validated with the same experimental data used in the validation of the 0D-1D fluid-dynamic model. A comparison between experiments and CFD results at 3000@11 was performed in Ref. [28], where a nice prediction of the in-cylinder pressure and HRR was achieved.

The methodology followed through this work can be found in the flow diagram depicted in Fig. 3. As a first step, an assessment of the dilution strategies is carried out, evaluating the adiabatic temperature (T_ad), laminar flame speed (S_L), and AID using 0D-1D chemical simulations. With this initial analysis, a theoretical operation range in terms of dilution will be obtained. Later, these limits will be validated with experimental measurements in the single-cylinder SI engine.

Afterward, an experimental comparison at medium speed and load (3000@11) between oxy-fuel and conventional combustion will be performed, exploring the differences in terms of in-cylinder pressure (P_cyl)_, temperature (T_cyl)_, heat release, adiabatic coefficient γ, GIE, emissions (NO_x, uHC, and CO), and the maximum amplitude of pressure oscillations (MAPO). Finally, knocking will be assessed, first experimentally, through the analysis of the in-cylinder pressure signal for different EGR ratios when the spark timing is advanced at 3000@11; later, since the knocking is more likely to appear at high load, a validated coupled method between 0D-1D and CFD simulations [28] is used to evaluate this phenomenon at high load (3000@25).

The results obtained along this work will be divided into three sections. The first one will discuss different dilution strategies that can be used in an OFC engine. Secondly, the experimental characterization of the OFC and its comparison with conventional combustion will be carried out. Finally, the knocking phenomena will be studied under OFC operation at medium and high load conditions.

Since the first objective was to analyze whether it is better to implement a dilution strategy based on O₂ or exhaust gases, an assessment of the impact of λ and EGR on the combustion process was performed to guide the selection. Thus, based on the 0D thermo-chemical simulations, Fig. 4 shows the flame temperature, the laminar flame speed, and the auto-ignition delay as a function of these parameters.

In terms of temperature, a straight line from λ = 1 and $EGR=65%$ (point 1) to a condition without EGR and λ = 5 (point 2) is shown. It is the limit of the region with flame temperatures below 3000 K; this limit was set from conventional combustion to keep the thermal integrity of the engine. In other words, all the dilutions rates combinations of λ and EGR that falls over the aforementioned line have the minimum dilutions to avoid exceeding the thermal limitations of the materials.

The second constraint considered from the combustion point of view will be the minimum laminar flame speed to have a combustion that is fast enough to be stable. A minimum value of 0.5 m/s was derived from a complete λ and EGR sweep in conventional combustion, identifying the longer combustion duration that ensures a stable progression of the flame. The straight line from λ = 1 and $EGR=75%$ (point 3) to $λ=6.5$ and $EGR=0%$ (point 4) limits the higher dilution that could be used without compromising the combustion stability.

Finally, a preliminary simplified knocking assessment was carried out. Due to the complexity of the phenomena involved in the knocking, the 0D chemical model was assumed not accurate enough to predict its real occurrence. However, predicting its probability relative to conventional combustion is a nice approach. Thus, the dashed line in the bottom graph delimits the region where AID leads to a similar probability of knocking as in conventional combustion. Even though it does not sets a real limit for knocking (that will be later evaluated with the more accurate CFD modeling), as a first approach, it sets a third boundary limit that should be avoided to keep knock propensity with respect to conventional combustion.

These three dilution constraints are useful to establish the operating region where the oxy-fuel combustion can take place in the SI engine keeping the combustion stability and ensuring the engine safety. However, to identify the benefits and drawbacks of each dilution strategy, also some technical issues must be taken into account to evaluate the concept's viability:

• In the case of selecting a dilution based on increasing the λ ratio, a higher O₂ mass rate will be needed. As commented, this is a limitation in OFC, whether O₂ is obtained from tanks or produced on-board. Thus, if an OTM was used for O₂ production, higher exhaust temperatures would be required to ensure energy availability in the membrane to increase the O₂ demand [15].

• In contrast to the λ strategy, EGR systems have been widely used for several decades, and different mature technical solutions are available. Thus, if higher exhaust gas recirculation is required to increase the dilution, it could be accomplished by any of the several ways described in the OFC patent of Desantes et al. [14]. Moreover, this approach is energetically more efficient since it does not require additional energy in the exhaust to guarantee the desired dilution.

• Another aspect to take into account is the exhaust gas composition. If the engine operates at $λ≈1$⁠, those gases will be composed mainly of CO₂ and H₂O vapor. This means that by a simple condensation of the H₂O, a CO₂-enriched stream will be obtained easily to be processed with any CCS technique.

Taking into account the predicted operation range according to the thermal and combustion limitations, and the technological viability discussed, it was decided that the best approach is diluting the in-cylinder mixture with EGR. Therefore, in the experimental work in the single-cylinder engine, this strategy was explored at 3000 rpm and $IMEP=11 bar$⁠. Initially, λ = 1 was measured to validate the results obtained by the 0D thermo-chemical simulations. Thus, in the zoom depicted at the top of Fig. 4, the striped rectangle (red zone) delimits the experimental range that was possible to measure. The experimental combustion stability threshold is also included in the zoom plot. On the one hand, there is a very good agreement between the maximum EGR rate that was possible to measure (73%) and the one theoretically predicted (75%). In this condition, the cyclic variability was quite high since the coefficient of variation of the indicated mean effective pressure (IMEP_cov) shown in Fig. 5 reaches values of about 8%. It was not possible to achieve a stable operation of the engine for higher EGR levels. On the other hand, 67% was the lowest rate able to be measured because this condition produced exhaust gas temperatures of around 1200 K, and this value jeopardized the engine specification recommended by the manufacturer and the pressure sensors located in that region. As will be discussed, EGR dilution reduction is neither interesting for the engine efficiency nor the emissions, so it was considered that no further validation was necessary, and the values predicted with the theoretical tool could be achieved if the engine thermal limitation was relaxed. In any case, reducing the EGR rate is not interesting for the OFC operation.

The experimental threshold states that if the engine operates at $λ=1.2$⁠, about 70% will be the maximum EGR that could be recirculated to the cylinder. Further increase of λ will require a reduction of EGR rate to guarantee combustion stability. As later discussed, values of λ slightly higher than one will also be explored at the test bench because of their interest in emission reduction. However, the main conclusion discussed regarding prioritizing EGR dilution will be respected, and only $λ=1.1$ will be analyzed.

Once the most suitable dilution range has been discussed, the next objective is to assess the differences between the oxy-fuel and the conventional combustion operation. To accomplish this purpose, the instantaneous in-cylinder pressure, temperature, heat release rate, and γ are analyzed to clarify the differences between the combustion modes. For the oxy-fuel combustion cases, four EGR ratios were measured. The spark timing was swept to reach the maximum GIE for each EGR level with a constant fuel mass. In all the cases, the intake temperature remains at a constant value of 80 °C to avoid water condensation in the EGR and intake lines. In the conventional combustion case, the same strategy was followed to set the spark timing, but in this case, no EGR was used. Operating conditions providing the best GIE are summarized in Table 5.

The following analysis can be made from the results obtained:

• In Fig. 6 (top), for the oxy-fuel cases, it can be seen that the pressure at the end of the compression phase increases with the EGR due to the higher mass trapped at the intake valve closing (IVC) when the dilution increases. Nevertheless, in contrast with the conventional combustion, a pressure reduction between 1 and 3 bar can be observed at the spark timing, before the combustion start, with EGR rates of 73% and 67%, respectively. Several accumulated effects produce this pressure reduction. On the one hand, the trapped mass at the inlet valve closing is lower in OFC because the dilution EGR mass (67–73%) is lower than the N₂ mass in the air (77%) for conventional combustion. Also, the specific gas constant R of the inlet flow is lower for OFC than for the air; thus, lower pressure would be required for the same inlet temperature to provide the same mass flow. These reasons lead to a lower pressure at the inlet valve closing. Later, during the compression stroke, there is an additional effect due to the change in the adiabatic coefficient, γ, evolution along the close cycle, as depicted in Fig. 7. There, it is shown that replacing N₂ with CO₂ and H₂O in the oxidizer will cause a diminution of about 8%. Considering that during the compression phase, the real polytropic process is affected by the lower γ, it will lead to lower pressure at the spark time.

• Regarding the pressure increase during the combustion process, several effects can be highlighted. On the one hand (later discussed), the combustion duration is longer when EGR increases but then spark timing is advanced to seek the best GIE. Combining these effects with the higher trapped mass for higher EGR rates leads to a non-monotonous peak pressure when EGR increases. On the other hand, the comparison of OFC with conventional combustion is straightforward and shows a clear trend to diminish the peak pressure due to the much lower pressure at the spark time and the global combustion slowdown in the EGR range considered. Even though faster combustion can be achieved with less diluted mixtures, the maximum pressure reached with oxy-fuel combustion will not exceed the one in conventional combustion.

• In terms of temperature, Fig. 6 (bottom) shows that despite being higher at the IVC in the oxy-fuel combustion cases, the values reached at the start of combustion are similar in both combustion modes (it was controlled to have comparable conditions). The reduced temperature change during compression can be explained by the γ evolution along the close cycle depicted in Fig. 7, as discussed for the pressure. Also, during the combustion process, a non-homogeneous trend is observed due to the same effects as in the pressure, that is, the change of trapped mass, combustion velocity, and spark timing. Moreover, the additional effect of the higher specific heat in OFC can be included in this case. As a result, the peak temperature can be higher or lower in OFC. It has been checked (not shown) that if the spark timing is kept, the same conclusion is obtained, and peak temperature is lower for OFC with high EGR rates and higher for low EGR rates.

• In any case, results indicate that the temperature reached at the exhaust valve opening will always be higher with oxy-fuel combustion, even when the peak temperature in OFC is lower than in the conventional combustion, as shown in the case with $EGR=73%$⁠, where peak temperature is almost 250 K lower. Again this can be justified by the smaller temperature variation during the expansion stroke due to the reduced γ obtained in the absence of N₂. This trend explains the temperatures measured at the exhaust with the thermocouples (see Table 6), which are also higher under oxy-fuel combustion operation.

• The HRR and heat release law (HRL) depicted in Fig. 8, and the values of CA90-CA10 shown in Table 6 indicate that the oxy-fuel combustion duration is longer than in the conventional case. Thus, the spark timing needs to be advanced to reach the maximum efficiency for each EGR rates, as detailed in Table 5. When EGR decreases from 73% to 69%, the CA90-CA10 duration is reduced from 28.5 to 21.2 cad, being in both cases above the 20.8 cad obtained in the conventional case. It can be stated that with λ = 1, an EGR rate of about 67–69% is required to have a combustion duration similar to the conventional case. When the EGR is below 70%, the maximum heat released exceeds the values obtained in the conventional case, reaching approximately 77 J/cad. However, for higher EGR dilution values, a less reactive mixture will produce peaks below the conventional case. It should be noted that the case with the lowest dilution value (⁠$EGR=67%$⁠) has not reached an optimal combustion centering, so it has been omitted in the plots.

Follow through the in-cylinder analysis, an emission comparison has been made in Fig. 9. In each graph the NO_x, uHC, and CO obtained at different EGR ratios for the oxy-fuel cases are shown and compared with the measurement in conventional combustion. The emissions are expressed in grams of pollutant per kilogram of fuel, more suitable than ppm for the comparison with the conventional engine because due to the N₂ absence, the concentration tends to distort the actual OFC-ICE emissions. It is worth mentioning that PM analysis was omitted since the PFI system was used to ensure homogeneous fuel-air mixture and stoichiometric or slightly poor mixtures will prevent the formation of this pollutant [5]. As can be seen, NO_x emissions got reduced more than 99% for λ = 1, thanks to the absence of N₂ coming from the air composition. Even though not being important (having in mind the dramatic reduction achieved in comparison with conventional combustion), a slight trend to diminish NO_x when EGR increases is observed. The different performance with EGR = 67% can be justified because its spark timing is not the optimal one. The NO_x produced (for the sake of rigor, it is not zero level) can be justified because of the small N₂ content of the fuel (0.2% of mass according to the chemical analysis) or the possible air entry due to lower pressure in some points of the engine (intake manifold, crankcase) with respect to ambient conditions. In order to assess the benefit of O₂ excess on efficiency, CO, and uHC, some points with $λ=1.1$ were measured following the same methodology as those with λ = 1 previously discussed. For $λ=1.1$⁠, EGR = 73% could not be measured because of low combustion stability. As expected, there is a small penalty on NO_x, but in any case, the values are more than 99% smaller in comparison with conventional conditions, and thus no specific after treatment system would be required for its reduction, giving room for oxidation catalyst.

It is interesting to highlight the apparently abnormal NO_x increase when λ becomes higher in Fig. 9. To explain this effect, the normalized heat release (HRL, from 0 to 1) and the in-cylinder temperature are depicted as a function of the crank angle in Fig. 10 for the OFC mode with a constant EGR ratio of 69% for both λ = 1 and $λ=1.1$⁠. HRL shows that about 80% of the combustion process takes place before 10 cad after TDC. During this combustion period, the in-cylinder temperature is higher when the O₂ concentration increases because of the combustion enhancement, and thus NO_x production is promoted under these conditions. Beyond 10 cad, the temperature with $λ=1.1$ start to become lower, and thus the maximum in-cylinder temperature (around 15 cad) and the temperature during the expansion stroke are lower for $λ=1.1$⁠, as expected. Nevertheless, during this late combustion, almost 90% of the fuel has already been burnt, and thus it has a low impact on NO_x generation.

Regarding the uHC, results showed a reduction of 81% when the engine operates with 73% of EGR at λ = 1; this reduction arises for the less diluted mixtures, thus for the case with 67% of EGR, the uHC were cut down almost 95%. If λ is slightly increased up to 1.1, reductions between 90% and 97% are achieved when the engine operates with 70% and 67% of EGR, respectively. These results show a clear potential of the OFC not only for NO_x reduction but also for uHC.

As for the CO emissions, experiments show a high sensitivity to λ, thus if the engine operates under oxy-fuel combustion at stoichiometric conditions, reductions of up to 35% could be achieved, but in the case of $EGR=70%$ an increase of 24% of the CO emissions with respect to conventional combustion is observed. In this case, the use of a slight O₂ excess is key to control CO. Thus, if $λ=1.1$ is used, CO emissions are reduced around 93–94.3% for the three EGR levels considered.

Finally, the experimental comparison between OFC and conventional combustion is performed in terms of engine efficiency GIE. As shown in Fig. 11 the oxy-fuel efficiency is expected to rise when both the EGR and λ increases, explained by the higher trapped mass at the IVC, which will lead to lower in-cylinder temperature, as previously discussed, and therefore a lower heat rejection to the walls. With λ = 1, when EGR arises from 69% to 73% (67% is omitted because spark timing was not optimized), the efficiency will increase almost 1% reaching 30.5%. A significant GIE increase of about 2% is achieved when $λ=1.1$ reaching more than 32%. This value is still about 7.7% points below the one obtained with a conventional combustion (39.7%). This is an issue to be improved, in spite of the advantage of OFC because the composition of the exhaust gases facilitates the application of CCS techniques to have zero tail pipe emissions of CO₂.

The last objective of this work is to evaluate the knocking propensity in a SI engine operating with oxy-fuel combustion. Therefore, as a starting point in Fig. 12, MAPO was calculated from experimental in-cylinder pressure as a function of spark timing for each EGR dilution rate at a medium load operating conditions (3000@11) and λ = 1. Results showed that the MAPO slightly increases when the spark is advanced toward the compression stroke. However, all the values remained under MAPO = 1, which is usually assumed to be a save threshold to avoid knocking [39].

For the same operating condition, it has been measured that when the engine operates under conventional combustion without EGR at an optimal spark timing (14 cad BTDC), knocking phenomena appear in some cycles since the average MAPO along the 250 cycles measured increases to a value of almost 5, exceeding the threshold mentioned above. It can also be observed that the MAPO in conventional conditions (represented by black stars) is much more sensitive to the spark timing than in oxy-fuel operation. Thus for spark timings earlier than 10 cad BTDC, conventional combustion shows a MAPO below 0.4, but then, it increases dramatically. On the contrary, OFC shows a smoother trend along the complete operating range measured.

Even though the cases with an 11 bar of IMEP had not shown signs of knocking with OFC, the effect should be studied at higher values since the appearance probability of this unwanted effect increases with the load. Taking into consideration that no experimental tests were carried out at a very high load due to test bench configuration limitations and engine specifications, simulations have been performed with the aforementioned 0D-1D-CFD method (consisting basically of the iterative coupling of 0D-1D fluid-dynamic modeling to set boundary conditions and three-dimensional-CFD calculation of the chamber phenomena during the closed cycle). Thus, in Fig. 13, the simulated HRR in OFC for stoichiometric conditions and with an EGR ratio of 70% at 3000 rpm and with 25 bar of IMEP is represented. The start of the spark was optimized with the 0D-1D fluid-dynamic tool to achieve the maximum GIE, leading to a SoC around 22.5 cad before the TDC. In these simulations, the HRR predicted by the model does not show any signs of end-gas knocking, which suggest that knocking is not expected in OFC using an EGR dilution strategy even at extreme high load operating conditions. This conclusion is interesting because considering that OFC leads to reduced peak pressures, there is room to increase the engine compression ratio to recover part of the GIE lost. Due to length limitations, this study is out of the scope of this work; however, some estimations with the 0D-1D fluid dynamic model have shown that about half of the indicated efficiency loss due to the OFC can be recovered by increasing the compression ratio, thus having a final penalty of about 4% points.

An experimental and simulated assessment has been carried out in a SI single-cylinder research engine operating under oxy-fuel combustion mode and conventional conditions. The results obtained in the test bench and with both 0D–1D thermo-chemical modeling and 0D-1D-CFD fluid-dynamic modeling allows inferring the following conclusions:

• Oxy-fuel combustion allows operating an SI engine with both O₂ and EGR dilution strategies in a wide operating range, even though EGR dilution seems more appropriate to achieve an integrated oxy-fuel combustion concept.

• The engine should operate near-stoichiometric conditions (λ = 1 or slightly higher than 1) with EGR rates between 67% and 73%. This will ensure the combustion stability, and fulfilling thermo-mechanical limits.

• Oxy-fuel combustion leads to lower peak pressure and higher exhaust temperatures than conventional combustion due to the changes in the gas properties and combustion development. The temperature increase can be interesting for on-board O₂ production with mixed ionic and electric conducting membrane.

• The main drawback of oxy-fuel combustion in SI engines is the gross indicated efficiency reduction of about 7% points. However, there is room to increase OFC efficiency with design strategies such as increasing the compression ratio or increasing the water content of the oxidizer gas.

• NO_x emissions are almost completely eliminated with oxy-fuel combustion in spark engines. Therefore, the tradeoff between NO_x and CO-THC is avoided, and combustion strategies can be oriented to increase oxidation rates. Also, paths the way for using thermally insulated combustion chambers and only oxidation catalyst for purification of CO₂ in further after treatment stages.

• A slight O₂ excess is key to control CO emissions, thus λ values between 1 and 1.1 seem a good compromise to maximize the emissions reduction (NO_x, CO, and uHC) and reducing the O₂ demand.

• Oxy-fuel combustion shows great potential to reduce knocking risk. This effect, in combination with the reduced peak pressure, confirms that there is some room to increase engine compression ratio and thus recovering part of the GIE lost with OFC.

As a final summary, it can be highlighted that the main novelties of the work are that it shows a comprehensive analysis of humid EGR and O₂ dilution in oxy-fuel combustion, including engine performance (indicated efficiency) and emissions (NO_x, CO, and uHC) under thermo-mechanical and knocking restrictions. Results show that using a slightly poor mixture near stoichiometric condition and EGR rates around 70%, NO_x are reduced more than 99%, and CO and uHC up to 90% with respect to the conventional ICE. The main oxy-fuel combustion issue, the efficiency reduction, can be partially compensated by increasing the compression ratio thanks to the lower knocking propensity, thus having an estimated final penalty of about 4% points. Thus, OFC-ICE arises as a smart transient alternative toward decarbonization, where the mature ICE technology can be used in combination with carbon capture and an oxidation catalyst.

The authors want to express their gratitude to CONVERGENT SCIENCE Inc. and Convergent Science GmbH for their kind support with the CFD calculations using the CONVERGE software. The authors would like to thank Ing. Gabriel Alcantarilla Ballesteros for his inestimable work during the experimental campaign.

• MCIN/AEI (Grant No. PID2021-123351OB-I00; Funder ID: 10.13039/501100011033).

• Generalitat Valenciana (Grant No. CIPROM/2021/061; Funder ID: 10.13039/501100003359).

• Santiago Grisolía (Grant No. GRISOLIAP/2021/108; Funder ID: 10.13039/501100003359).

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

AID =

auto ignition delay

AMR =

adaptive mesh refinement

ATDC =

after top dead center

BSFC =

brake specific fuel consumption

BTDC =

before top dead center

CAD =

crank angle degree

CCS =

carbon capture storage

CO =

carbon oxides

EGR =

exhaust gas recirculation

EVO =

exhaust valve opening

GDI =

gasoline direct injection

GHG =

green house gases

GIE =

gross indicated eficiency

HRL =

heat release law

HRR =

heat release rate

ICE =

internal combustion engine

IMEP =

indicated mean effective pressure

IVC =

intake valve closing

MAPO =

maximum amplitude pressure oscillation

MIEC =

mixed ionic-electric conducting

NO_x =

nitrogen oxides

OFC =

oxy-fuel combustion

OTM =

oxygen transport membrane

PFI =

port fuel injection

PM =

particulate matter

SI =

spark ignition

S_L =

laminar flame speed

ST =

spark timing

T_ad =

adiabatic flame temperature

TDC =

top dead center

uHC =

unburned hydrocarbons