Concepts for Hydrogen Internal Combustion Engines and Their Implications on the Exhaust Gas Aftertreatment System

: Hydrogen as carbon-free fuel is a very promising candidate for climate-neutral internal combustion engine operation. In comparison to other renewable fuels, hydrogen does obviously not produce CO 2 emissions. In this work, two concepts of hydrogen internal combustion engines (H 2 -ICEs) are investigated experimentally. One approach is the modiﬁcation of a state-of-the-art gasoline passenger car engine using hydrogen direct injection. It targets gasoline-like speciﬁc power output by mixture enrichment down to stoichiometric operation. Another approach is to use a heavy-duty diesel engine equipped with spark ignition and hydrogen port fuel injection. Here, a diesel-like indicated efﬁciency is targeted through constant lean-burn operation. The measurement results show that both approaches are applicable. For the gasoline engine-based concept, stoichiometric operation requires a three-way catalyst or a three-way NO X storage catalyst as the primary exhaust gas aftertreatment system. For the diesel engine-based concept, state-of-the-art selective catalytic reduction (SCR) catalysts can be used to reduce the NOx emissions, provided the engine calibration ensures sufﬁcient exhaust gas temperature levels. In conclusion, while H 2 -ICEs present new challenges for the development of the exhaust gas aftertreatment systems, they are capable to realize zero-impact tailpipe emission operation.


Introduction
Sustainability is the key driver for the transformation of powertrains for mobile and stationary solutions. It requires the reduction of both greenhouse gas emissions and pollutant emissions. In this regard, facing the mobility sector, internal combustion engines (ICE) need to compete with battery electric vehicles (BEVs) and fuel cell electric vehicles (FCEVs). Selected advantages of ICEs compared to BEVs and FCEVs are the robustness towards ambient conditions as well as fuel and air impurities, the low demand for rare earths and precious metals and the well-established development and production processes [1,2]. Air/hydrogen charges have ignition limits of 0.15 < λ < 10.5 which are much wider than those of conventional fuels such as gasoline or diesel, enabling both operation principles. Furthermore, operation on hydrogen enables ICEs to reduce fuel-based carbon dioxide (CO 2 ) emissions down to zero. Lubrication oil-based CO 2 emissions are expected to be on a negligibly low level [3]. The latest research on combustion engines in general targets zero-impact tailpipe emissions [4]. The idea of zero-impact combustion engines are negligible tailpipe emissions, e.g., a pollutant contribution of traffic below a clean rural background. Addressing this challenge will bring down the propulsion system discussion from a political level to an efficiency-based, use-case specific evaluation.
Hydrogen internal combustion engines (H 2 -ICE) have already been in development for some years [5][6][7][8]. In the early 2000s, BMW introduced their Hydrogen 7 as series This paper presents measurement results from two recent experimental test campaigns with series production engines. Both engines are modified but not explicitly optimized or redesigned for hydrogen operation, leaving room for future progress. One of the two engines is based on a compressed natural gas (CNG)-fueled passenger car engine, once originally developed as a gasoline engine. The other one is based on a series production diesel engine that was previously modified for CNG operation. Both are presented in Figure 1 and will be referred to as gasoline and diesel engine-based concepts in the following.

Materials and Methods
As explained above, an H2-ICE can be equipped with PFI or DI. In both cases, knocking combustion can occur at high loads, early centers of combustion and low air/fuel ratios [33,34]. At the same time, enleanment improves the efficiency. As a result, the H2-ICE features spark-ignition (SI), but enables quality control, which classifies it in between conventional gasoline and diesel concepts ( Figure 1). In this work, representative engines of both concepts are investigated experimentally. The applied test procedure is also presented in Figure 1. Thermodynamic work packages will be introduced in separate publications. The present paper focuses on the impact of thermodynamic measurement results on the exhaust aftertreatment system.
The gasoline engine-based concept features a three-cylinder passenger car engine with a total displacement of 1 L. The cylinder charge motion is tumble supported and the CNG injectors of the base engine are modified with an external lubrication system for H2 direct injection with 20 bar. To limit the occurrence of knocking combustion, the compression ratio was kept at the CNG engine reference of ε ≈ 10, which is at the typical level of gasoline engines. This enables stoichiometric operation of the engine, which is necessary to achieve higher specific loads. In addition, the engine is equipped with a high-pressure exhaust gas recirculation (EGR) system. No further optimization was conducted, especially no replacement of the boosting system or of the charge motion support as defined by the shape of the intake ports or the piston. This enables a concept study, but not the optimum performance that would be possible with a completely optimized and redesigned engine.
The heavy duty (HD) diesel engine-based concept uses a modified six-cylinder engine with a total displacement of 7.8 L. It was modified from diesel to CNG operation beforehand by the implementation of spark plugs and PFI. To realize high mean effective pressures in a lean-burn operation, the single-stage turbocharging system was replaced by a two-stage turbocharger arrangement. The compression ratio of ε = 13 of the diesel

Materials and Methods
As explained above, an H 2 -ICE can be equipped with PFI or DI. In both cases, knocking combustion can occur at high loads, early centers of combustion and low air/fuel ratios [33,34]. At the same time, enleanment improves the efficiency. As a result, the H 2 -ICE features spark-ignition (SI), but enables quality control, which classifies it in between conventional gasoline and diesel concepts ( Figure 1). In this work, representative engines of both concepts are investigated experimentally. The applied test procedure is also presented in Figure 1. Thermodynamic work packages will be introduced in separate publications. The present paper focuses on the impact of thermodynamic measurement results on the exhaust aftertreatment system.
The gasoline engine-based concept features a three-cylinder passenger car engine with a total displacement of 1 L. The cylinder charge motion is tumble supported and the CNG injectors of the base engine are modified with an external lubrication system for H 2 direct injection with 20 bar. To limit the occurrence of knocking combustion, the compression ratio was kept at the CNG engine reference of ε ≈ 10, which is at the typical level of gasoline engines. This enables stoichiometric operation of the engine, which is necessary to achieve higher specific loads. In addition, the engine is equipped with a high-pressure exhaust gas recirculation (EGR) system. No further optimization was conducted, especially no replacement of the boosting system or of the charge motion support as defined by the shape of the intake ports or the piston. This enables a concept study, but not the optimum performance that would be possible with a completely optimized and redesigned engine.
The heavy duty (HD) diesel engine-based concept uses a modified six-cylinder engine with a total displacement of 7.8 L. It was modified from diesel to CNG operation beforehand by the implementation of spark plugs and PFI. To realize high mean effective pressures in a lean-burn operation, the single-stage turbocharging system was replaced by a two-stage turbocharger arrangement. The compression ratio of ε = 13 of the diesel engine was retained. This leads to severe knocking at low lambda values, which prohibits Energies 2021, 14, 8166 4 of 13 stoichiometric operation [35]. At the same time, this compression ratio enables very lean operation up to λ = 8 and shows efficiencies that can compete with diesel engines.
Both engines were equipped with extensive emission analysis equipment. This includes an emissions analyzer from FEV containing a chemiluminescence detector (CLD), a paramagnetic detector (PMD), a nondispersive infrared sensor (NDIR) and flame ionization detector (FID) for measurements of NO X , O 2 , CO, CO 2 and HC emissions. Additionally equipped are a HSense from V&F for hydrogen measurements and a FTIR from MKS for H 2 O and multiple other emissions.

Gasoline Engine-Based Hydrogen Engine for Passenger Car Applications
The small three-cylinder passenger car hydrogen engine is intended to compete with the gasoline version in terms of power and maximum mean effective pressures. The hydrogen version was thus adapted for stoichiometric operation at full load. Here, the achievable mean effective pressure level is limited by the unchanged turbocharging system. In addition, the combustion chamber design is not optimized for mixture formation with hydrogen, causing mixture inhomogeneities. Despite the compression ratio of ε = 10, knocking occurs, which limits the usable range of the ignition timing at full load. To illustrate, where the engine is not operated as efficiency optimal to prevent from knocking, Figure 2 shows the point of 50% mass fraction burned (MFB50) in the engine map as indication for the center of combustion. The highest efficiency is achieved at MFB50 = 8 • CA aTDC as a compromise between the constant volume process and wall heat losses. However, the ignition and accordingly also the MFB50 is retarded beyond 16 • CA aTDC at loads above a brake mean effective pressure of BMEP = 16 bar. engine was retained. This leads to severe knocking at low lambda values, which prohibits stoichiometric operation [35]. At the same time, this compression ratio enables very lean operation up to λ = 8 and shows efficiencies that can compete with diesel engines. Both engines were equipped with extensive emission analysis equipment. This includes an emissions analyzer from FEV containing a chemiluminescence detector (CLD), a paramagnetic detector (PMD), a nondispersive infrared sensor (NDIR) and flame ionization detector (FID) for measurements of NOX, O2, CO, CO2 and HC emissions. Additionally equipped are a HSense from V&F for hydrogen measurements and a FTIR from MKS for H2O and multiple other emissions.

Gasoline Engine-Based Hydrogen Engine for Passenger Car Applications
The small three-cylinder passenger car hydrogen engine is intended to compete with the gasoline version in terms of power and maximum mean effective pressures. The hydrogen version was thus adapted for stoichiometric operation at full load. Here, the achievable mean effective pressure level is limited by the unchanged turbocharging system. In addition, the combustion chamber design is not optimized for mixture formation with hydrogen, causing mixture inhomogeneities. Despite the compression ratio of ε = 10, knocking occurs, which limits the usable range of the ignition timing at full load. To illustrate, where the engine is not operated as efficiency optimal to prevent from knocking, Figure 2 shows the point of 50% mass fraction burned (MFB50) in the engine map as indication for the center of combustion. The highest efficiency is achieved at MFB50 = 8 °CAaTDC as a compromise between the constant volume process and wall heat losses. However, the ignition and accordingly also the MFB50 is retarded beyond 16 °CAaTDC at loads above a brake mean effective pressure of BMEP = 16 bar. Engine operation strategy as result of previous thermodynamic investigations. The position of 50% fuel mass fraction burned (MFB50) shows an efficiency optimal operation for low and mediums loads, but a retarded combustion due to knocking at higher loads.
The operation strategy derived from thermodynamic investigations is presented in Figure 2. At full load, stoichiometric operation and a decreasing Miller timing is required to deliver the required cylinder charge. In the range of medium BMEPs, Miller timing and exhaust gas recirculation are favorable for de-throttling. In addition, the flame cooling effects of the EGR lead to a NOX raw emission reduction [36]. In order to achieve shows an efficiency optimal operation for low and mediums loads, but a retarded combustion due to knocking at higher loads.
The operation strategy derived from thermodynamic investigations is presented in Figure 2. At full load, stoichiometric operation and a decreasing Miller timing is required to deliver the required cylinder charge. In the range of medium BMEPs, Miller timing and exhaust gas recirculation are favorable for de-throttling. In addition, the flame cooling effects of the EGR lead to a NO X raw emission reduction [36]. In order to achieve maximum efficiency, the ignition angle is set for an MFB50 of approximately 8 • CA aTDC , wherever knocking allows. At low BMEPs, lean engine operation becomes more efficient than operation with EGR. The reason is a reduction in NO X formation as can be seen in Figure 3.
Energies 2021, 14, x FOR PEER REVIEW 5 of 13 maximum efficiency, the ignition angle is set for an MFB50 of approximately 8 °CAaTDC, wherever knocking allows. At low BMEPs, lean engine operation becomes more efficient than operation with EGR. The reason is a reduction in NOX formation as can be seen in Figure 3. Measurements at other engine speeds also account into the interpolation, but as fourth dimension they cannot be displayed.
The graphs in Figure 3 show the reasonable BMEP ranges as a function of the air/fuel ratio. It can be seen that the area of BMEP narrows at higher loads towards stoichiometric conditions. The maximum air/fuel ratio is limited by the amount of air introduced into the cylinders and decreases with increasing fuel mass required for increasing load. Another boundary condition is shown by the white dashed line. It marks the range up to which an efficiency optimized MFB50 is applicable. At higher loads, the ignition angle αIgn needs to be increased to prevent knocking.
In Figure 3a, the NOX emissions are shown as a function of the relative air/fuel ratio and load. It can be seen that, starting at stoichiometric conditions, a minor enleanment leads to increased NOX emissions, as known from gasoline engines. With more oxygen present in the exhaust gas, the NOX formation tendency increases. After reaching a maximum around λ = 1.3, the NOX emissions decrease significantly below the level of stoichiometric conditions. This is a consequence of the decreasing temperature, as shown in Figure 3c. As a result, lean operation is beneficial in terms of NOX emissions only at low loads. This is also the reason for splitting of the operation strategy into EGR for de-throttling at medium loads and enleanment at low loads ( Figure 2). Looking at the unburned fraction of hydrogen in Figure 3b, significant H2 emissions are detected around stoichiometric operation. These are attributed to the not optimized combustion chamber design. An optimized mixture formation and combustion process will significantly increase the indicated efficiency.
Additional important parameters for the exhaust gas aftertreatment system are temperature and mass flow of the exhaust gas. The exhaust gas temperature (Figure 3c Measurements at other engine speeds also account into the interpolation, but as fourth dimension they cannot be displayed.
The graphs in Figure 3 show the reasonable BMEP ranges as a function of the air/fuel ratio. It can be seen that the area of BMEP narrows at higher loads towards stoichiometric conditions. The maximum air/fuel ratio is limited by the amount of air introduced into the cylinders and decreases with increasing fuel mass required for increasing load. Another boundary condition is shown by the white dashed line. It marks the range up to which an efficiency optimized MFB50 is applicable. At higher loads, the ignition angle α Ign needs to be increased to prevent knocking.
In Figure 3a, the NO X emissions are shown as a function of the relative air/fuel ratio and load. It can be seen that, starting at stoichiometric conditions, a minor enleanment leads to increased NO X emissions, as known from gasoline engines. With more oxygen present in the exhaust gas, the NO X formation tendency increases. After reaching a maximum around λ = 1.3, the NO X emissions decrease significantly below the level of stoichiometric conditions. This is a consequence of the decreasing temperature, as shown in Figure 3c. As a result, lean operation is beneficial in terms of NO X emissions only at low loads. This is also the reason for splitting of the operation strategy into EGR for de-throttling at medium loads and enleanment at low loads ( Figure 2). Looking at the unburned fraction of hydrogen in Figure 3b, significant H 2 emissions are detected around stoichiometric operation. These are attributed to the not optimized combustion chamber design. An optimized mixture formation and combustion process will significantly increase the indicated efficiency.
Additional important parameters for the exhaust gas aftertreatment system are temperature and mass flow of the exhaust gas. The exhaust gas temperature (Figure 3c) highly depends on the air/fuel ratio and varies in a wide range from below T Exh = 300 • C to above T Exh = 600 • C. The isolines of the exhaust mass flow in Figure 3d are vertical to the temperature isolines. This means that several combinations of temperatures and mass flows are possible, which is an additional degree of freedom for the EATS operation strategy in comparison to gasoline engines. In this regard, lower mass flows resulting in lower space velocities and thus more reaction time are positive for the exhaust gas system at temperatures with slow reaction kinetics.

Diesel Engine-Based Hydrogen Engine for Heavy Duty Applications
A strongly different concept to the one shown in the chapter above is the diesel enginebased hydrogen engine. It combines a high compression ratio with hydrogen port fuel injection and a two-stage turbo charger arrangement. This leads to benefits in the operation efficiency. However, with this high compression ratio, stoichiometric operation is no longer possible. The knock limitation to lower lambda values ranges in between 1.5 ≤ λ ≤ 2. The main exhaust gas parameters in the engine map of an efficiency oriented calibration is depicted in Figure 4. temperature isolines. This means that several combinations of temperatures and mass flows are possible, which is an additional degree of freedom for the EATS operation strategy in comparison to gasoline engines. In this regard, lower mass flows resulting in lower space velocities and thus more reaction time are positive for the exhaust gas system at temperatures with slow reaction kinetics.

Diesel Engine-Based Hydrogen Engine for Heavy Duty Applications
A strongly different concept to the one shown in the chapter above is the diesel engine-based hydrogen engine. It combines a high compression ratio with hydrogen port fuel injection and a two-stage turbo charger arrangement. This leads to benefits in the operation efficiency. However, with this high compression ratio, stoichiometric operation is no longer possible. The knock limitation to lower lambda values ranges in between 1.5 ≤ λ ≤ 2. The main exhaust gas parameters in the engine map of an efficiency oriented calibration is depicted in Figure 4. In Figure 4a, the relative air/fuel ratio is plotted in the engine map as the function of engine speed and torque. The engine operates at lean conditions, close to the lean operation limit, where incomplete or instable combustion occurs. As noted for the gasoline engine-based concept in Figure 3, the lean operation limit decreases towards higher loads. The lean limit of stable operation under varying boundary conditions ranges from λ > 4 at low loads up to λ < 2 at the low-end torque. Moreover, also similar to the gasoline engine-based hydrogen engine in Figure 3, the exhaust temperatures in Figure 4c show a strong correlation to the air/fuel ratio. The lowest temperatures occur in ultra-lean operation at low load and even reach values below TExh < 200 °C. Even at high loads, the exhaust temperatures do not exceed TExh = 350 °C with this configuration.
As a result of this ultra-lean engine calibration, low cylinder temperatures and accordingly low exhaust gas temperatures apply (the latter depicted in Figure 4c). This results in low NOX emissions measured in the raw exhaust gas. A wide area of more than three quarters of the map shows emissions of ψNOx < 20 ppm (lower marked area in Figure  4b). At engine speeds of n > 1500 1/min, maximum NOX emissions at full load are still limited to ψNOx = 100 ppm (upper area). Only the area of low-end torque shows severe  In Figure 4a, the relative air/fuel ratio is plotted in the engine map as the function of engine speed and torque. The engine operates at lean conditions, close to the lean operation limit, where incomplete or instable combustion occurs. As noted for the gasoline engine-based concept in Figure 3, the lean operation limit decreases towards higher loads. The lean limit of stable operation under varying boundary conditions ranges from λ > 4 at low loads up to λ < 2 at the low-end torque. Moreover, also similar to the gasoline engine-based hydrogen engine in Figure 3, the exhaust temperatures in Figure 4c show a strong correlation to the air/fuel ratio. The lowest temperatures occur in ultra-lean operation at low load and even reach values below T Exh < 200 • C. Even at high loads, the exhaust temperatures do not exceed T Exh = 350 • C with this configuration.
As a result of this ultra-lean engine calibration, low cylinder temperatures and accordingly low exhaust gas temperatures apply (the latter depicted in Figure 4c). This results in low NO X emissions measured in the raw exhaust gas. A wide area of more than three quarters of the map shows emissions of ψ NOx < 20 ppm (lower marked area in Figure 4b).
At engine speeds of n > 1500 1/min, maximum NO X emissions at full load are still limited to ψ NOx = 100 ppm (upper area). Only the area of low-end torque shows severe NO X emissions as a result of the comparably rich air/fuel ratio. However, the engine control functions and the calibration of the respective parameters were not yet finalized and a significant reduction towards more reasonable NO X emissions can be expected. Again, the exhaust mass flow isolines (Figure 4d) are more or less vertical to the temperature isolines. This challenges the EATS to perform under all possible combinations.
With the findings from above and additional thermodynamic investigations, it was found that for λ > 2, only very small advantages were observed in terms of the indicated efficiency. In a second development step, another calibration approach was applied. This time, the target is not at the highest indicated efficiency, but at moderate exhaust temperatures. The recalibration of the engine control functions focuses on the low speed and load ranges, which poses the greatest challenges for cold start and the subsequent warm-up phase. The resulting new engine maps are presented in Figure 5. NOX emissions as a result of the comparably rich air/fuel ratio. However, the engine control functions and the calibration of the respective parameters were not yet finalized and a significant reduction towards more reasonable NOX emissions can be expected. Again, the exhaust mass flow isolines (Figure 4d) are more or less vertical to the temperature isolines. This challenges the EATS to perform under all possible combinations. With the findings from above and additional thermodynamic investigations, it was found that for λ > 2, only very small advantages were observed in terms of the indicated efficiency. In a second development step, another calibration approach was applied. This time, the target is not at the highest indicated efficiency, but at moderate exhaust temperatures. The recalibration of the engine control functions focuses on the low speed and load ranges, which poses the greatest challenges for cold start and the subsequent warm-up phase. The resulting new engine maps are presented in Figure 5. The new engine control strategy targets an exhaust gas temperature above 250 °C as it is needed for conventional SCR systems. Therefore, lambda is decreased to λ = 2.3 as shown in Figure 5a. By this measure, the exhaust temperature increases above 300 °C in the whole map except for a very small area. This shows the strong impact of the engine calibration on the exhaust gas characteristics.
As discussed before, NOX emissions are highly sensitive to changes in the relative air/fuel ratio. However, similar to the gasoline engine-based concept depicted in Figure  3a, at the relative air/fuel ratio of λ = 2.3, the NOX reduction due to a dilution induced exhaust gas temperature dominates over the effect of an oxygen availability-induced NOX increase. Thus, the positive effect of enleanment is achieved and the NOX emissions are significantly below those of stoichiometric operation. Figure 5b shows NOX emissions in the range of 30 ppm and below.
As an aside, it should be mentioned that the exhaust gas mass flow decreases slightly due to the reduced air mass flow with less enleanment (Figure 5d). Accordingly, in terms of absolute emissions (in g/kWh) and assuming a constant NOX concentration, the reduction in exhaust mass flow leads to a corresponding reduction in NOX emissions. In other The new engine control strategy targets an exhaust gas temperature above 250 • C as it is needed for conventional SCR systems. Therefore, lambda is decreased to λ = 2.3 as shown in Figure 5a. By this measure, the exhaust temperature increases above 300 • C in the whole map except for a very small area. This shows the strong impact of the engine calibration on the exhaust gas characteristics.
As discussed before, NO X emissions are highly sensitive to changes in the relative air/fuel ratio. However, similar to the gasoline engine-based concept depicted in Figure 3a, at the relative air/fuel ratio of λ = 2.3, the NO X reduction due to a dilution induced exhaust gas temperature dominates over the effect of an oxygen availability-induced NO X increase. Thus, the positive effect of enleanment is achieved and the NO X emissions are significantly below those of stoichiometric operation. Figure 5b shows NO X emissions in the range of 30 ppm and below.
As an aside, it should be mentioned that the exhaust gas mass flow decreases slightly due to the reduced air mass flow with less enleanment (Figure 5d). Accordingly, in terms of absolute emissions (in g/kWh) and assuming a constant NO X concentration, the reduction in exhaust mass flow leads to a corresponding reduction in NO X emissions. In other words, the mass flow reduction already contributes to a slight increase in NO X concentration without producing more NO X in absolute terms.

Discussion
The results of both measurement campaigns depict very different exhaust gas characteristics. The gasoline engine-based H 2 -ICE with a low compression ratio and small displacement causes rather high NO X raw emissions at higher loads, but also high exhaust gas temperatures. On the contrary, the diesel engine-based H 2 -ICE with a higher compression ratio and two-stage turbocharging emits NO X raw emissions on a very low level. This is achieved by further enleanment, which leads to lowered exhaust gas temperatures.
Considering suitable exhaust gas aftertreatment systems, the determining factor is the stoichiometric operation of the gasoline engine-based hydrogen engine at higher loads. Stoichiometric operation is necessary to achieve the desired specific loads of up to BMEP > 20 bar, where the air charge is limited by the series production boosting system of the gasoline engine. At stoichiometric operation, however, an optimization of the mixture formation towards nearly complete combustion leads to very low oxygen concentrations. Under these stoichiometric conditions, conventional DeNO X systems such as an SCR or NO X storage catalysts are not applicable. Instead, a three-way catalyst (TWC) is necessary for exhaust gas aftertreatment.
In contrast to gasoline engines, the exhaust gases contain no or only negligible carbonbased emissions. Thus, the TWC lacks CO or hydrocarbons as reducing agents that react with oxygen from the Ceroxide or convert stored NOx in the Barium oxide phase to nitrogen [37]. This role must be taken over by hydrogen. Appropriate references can be found in the literature [33,38]. However, hydrogen as reductant could lead to new TWC material formulations in the future and will remain an object of research [39].
When switching between stoichiometric and lean operation, another subject is the TWC performance under lean conditions. Krishnan Unni et al. [40] operated a H 2 -ICE equipped with a TWC at lean conditions and still observed catalytic effects. With the full and thus practically deactivated oxygen storage, the catalytic effects can be attributed to the platinum group metals (PGM) acting as an oxidation catalyst (OC). This enables the TWC to replace an oxidation catalyst (OC) and provide NO 2 to a downstream SRC system, which requires an NO 2 /NO X ratio of about 1/2 to operate efficiently, depending on the material [41]. Accordingly, the aftertreatment concept for lean conditions could include a TWC and a SCR system. At that temperature level, a NH 3 -SCR system is more sensible than an H 2 -SCR. The NH 3 -SCR might even operate as passive concept fed by secondary ammonia emissions of the TWC.
Another option for the exhaust gas aftertreatment is the three-way NO X storage catalyst (TWNSC). It is capable of NO X conversion under stoichiometric conditions but has a strongly increased NO X storage capacity [42,43]. With this, the TWNSC is also applicable for lean operation, where it stores the NO X . For the regeneration, cyclic rich operation is required. Alike the TWC, hydrogen as the only reducing agent leads to changed boundary conditions, posing similar material development challenges.
From this discussion, the EATS concepts for gasoline engine-based H 2 -ICEs, including stoichiometric operation, shown in Figure 6a, emerge. Due to the stoichiometric operation of the gasoline engine-based concept, the introduction of a TWC or TWNSC is mandatory. However, for lean operation it needs to be supported by an additional DeNO X system, which is here an SCR catalyst as discussed beforehand.
Furthermore, Figure 6b introduces exhaust aftertreatment systems for the diesel engine-based concept. Here, lean engine operation enables low NO X raw emissions. The aftertreatment concept is mainly depending on the exhaust gas temperature. Due to the favorable burn characteristics of hydrogen, a wide calibration range is possible [44]. Two scenarios based on the same diesel engine-based hydrogen engine are illustrated in Figures 4 and 5.
An engine calibration of the whole map at, e.g., λ = 2.3 enables exhaust gas temperatures high enough for SCR systems. The central aftertreatment component for lean engine concepts will be a selective catalytic reduction (SCR) catalyst, most likely combined with Furthermore, Figure 6b introduces exhaust aftertreatment systems for the diesel engine-based concept. Here, lean engine operation enables low NOX raw emissions. The aftertreatment concept is mainly depending on the exhaust gas temperature. Due to the favorable burn characteristics of hydrogen, a wide calibration range is possible [44]. Two scenarios based on the same diesel engine-based hydrogen engine are illustrated in Figures 4

and 5.
An engine calibration of the whole map at, e.g., λ = 2.3 enables exhaust gas temperatures high enough for SCR systems. The central aftertreatment component for lean engine concepts will be a selective catalytic reduction (SCR) catalyst, most likely combined with some kind of oxidation catalyst upstream. One option with oxidizing characteristics under lean conditions is a NOX storage catalyst (NSC).
The NSC could support the SCR to increase the overall conversion efficiency and is divided into two types: either lean NOX traps (LNTs), which require rich conditions for regeneration phases, or passive NOX adsorbers (PNAs), which release the NOX at higher temperatures. Especially PNAs can also cover slightly lower exhaust gas temperatures and extend the EATS operation range.
As an alternative solution to the λ = 2.3 calibration, at ultra-lean operation almost no NOX emissions are present and operation without exhaust gas aftertreatment is theoretically possible. However, transient operation scenarios lead to dynamic NOX emission peaks that are not compliant without any EATS. At the same time, heavy enleanment leads to significantly lower exhaust gas temperatures below the light-off temperatures of conventional DeNOX systems, causing the EATS to be inactive.
Especially during the heat-up phase after a cold start, the low exhaust gas temperatures are challenging. A solution could be the application of an H2-SCR system which has the highest efficiency at low temperatures. Installing it in first close-coupled position enables an early light-off even at extremely low exhaust gas temperatures. At higher temperatures, it could work as oxidation catalyst to form NO2 for the SCR, as discussed before with the TWC. However, the H2-SCR technology is still under development. As a result, all three EATS concepts in Figure 6b are most promising for the diesel engine-based H2-ICE. The final decision will then be depending on the engine application. The NSC could support the SCR to increase the overall conversion efficiency and is divided into two types: either lean NO X traps (LNTs), which require rich conditions for regeneration phases, or passive NO X adsorbers (PNAs), which release the NO X at higher temperatures. Especially PNAs can also cover slightly lower exhaust gas temperatures and extend the EATS operation range.
As an alternative solution to the λ = 2.3 calibration, at ultra-lean operation almost no NO X emissions are present and operation without exhaust gas aftertreatment is theoretically possible. However, transient operation scenarios lead to dynamic NO X emission peaks that are not compliant without any EATS. At the same time, heavy enleanment leads to significantly lower exhaust gas temperatures below the light-off temperatures of conventional DeNO X systems, causing the EATS to be inactive.
Especially during the heat-up phase after a cold start, the low exhaust gas temperatures are challenging. A solution could be the application of an H 2 -SCR system which has the highest efficiency at low temperatures. Installing it in first close-coupled position enables an early light-off even at extremely low exhaust gas temperatures. At higher temperatures, it could work as oxidation catalyst to form NO 2 for the SCR, as discussed before with the TWC. However, the H 2 -SCR technology is still under development. As a result, all three EATS concepts in Figure 6b are most promising for the diesel engine-based H 2 -ICE. The final decision will then be depending on the engine application.
In summary, the engine operation concept and the exhaust gas aftertreatment concept must be developed in parallel. A matching process is enabled by choosing the engine control strategy with respect to the requirements of the EATS. The NO X raw emissions and the exhaust gas temperature are the dominating parameters. Compared to gasoline or diesel engines, the presence of hydrogen in the raw exhaust gas, high water fractions relative to the air/fuel ratio, and the absence of carbon-based emissions lead to new boundary conditions. Accordingly, the concepts developed for switching (between stoichiometric and lean) operation and permanent lean operation must face the same challenges, although they follow two contradictory approaches. As a result for both strategies, the potential trade-off between low NO X raw emissions and high catalytic conversion efficiency requires compromises that are specifically designed for the target application. With regard to future emission legislations, the environmentally friendly zero-impact tailpipe emissions include ultra-low emissions of N 2 O and NH 3 . This requires very precise ammonia dosing strategies that also consider passive NH 3 formation provided by upstream catalysts (TWC, TWNSC or NSC). As an alternative, additional ammonia slip catalysts could be integrated. However, the oxidizing characteristics of ammonia slip catalysts (ASC) may increase the N 2 O emissions. For that reason, ASCs were not considered in the suggested concepts within Figure 6.
Finally, the previously discussed catalysts can also be combined on coated particulate filters, which could be considered due the potential of particulate formation resulting from combustion of lubrication oil [45]. The resulting ash accumulation is known to increase the filtration efficiency but may need further investigation due to the absence of fuel-induced soot particulates [46].
As a final remark, naming the two approaches gasoline engine-and diesel enginebased concepts is not intended to limit the applicable use range. These labels were chosen to describe their origin and some basic engine parameters such as compression ratio and displacement. Future solutions will be based on one of the concepts or something in between, considering the specific use-case and the performance priorities such as maximum efficiency, high specific load, low production costs or others. Additionally, both engines were not optimized for hydrogen operation with dedicated mixture formation or boosting strategies. This leaves further potential for future H 2 -ICEs.

Conclusions
The conversion of internal combustion engines shows that contrary approaches are feasible, e.g., with different compression ratios enabling different operating strategies. Depending on the chosen approach, the following implications on the exhaust gas aftertreatment systems (EATS) apply.
For the gasoline engine-based hydrogen combustion engine with lower compression ratio (here ε = 10) aiming at high specific power, the following conclusions are drawn:

1.
Lean operation is limited by the boosting system. Highest specific power is reached with stoichiometric combustion, which is accompanied by relatively high NO X emissions. Thus, the EATS has to include either a three-way catalyst (TWC) or a three-way NO X storage catalyst (TWNSC); 2.
At low loads, lean operation is still more efficient. Here, we see an SCR catalyst as the system of choice. Under these conditions, the upstream TWC/TWNSC can be used as oxidation catalyst to increase the NO 2 /NO X ratio. If hydrogen is available in the exhaust, the TWC/TWNSC will also produce ammonia, which reduces the required amount of urea injection; 3.
Switching between stoichiometric and lean operation will require dedicated engine operation strategies. Here, the avoidance of NO X , NH 3 and N 2 O emissions will most likely require high attention. Main conclusions for the diesel engine-based hydrogen combustion engine with higher compression ratio (here ε = 13) targeting high efficiency are the following: 4.
Ultra-lean operation enables ultra-low NO X raw emissions down to the limit of detectability. However, the heavy enleanment comes along with very low exhaust gas temperatures that are below the light-off temperatures of currently available catalysts. As a result, the exhaust gas aftertreatment system is not instantly available in the event of a transient operating point change; 5.
From the exhaust gas aftertreatment point of view, less lean operation is beneficial. An engine operation, e.g., at an air/fuel ratio of λ = 2.3 still causes low NO X emissions, but increases the exhaust temperature to ensure catalyst activity; 6.
Under these operation conditions, the SCR catalyst is the most promising EATS component and can be directly transferred from diesel applications. For operation areas, where the SCR system is not fully active, a combination with an upstream NO X storage catalyst or H 2 -SCR system increases the overall NO X reduction performance.
In summary, the exhaust aftertreatment systems have to focus on nitrogen related emissions: NO X as primary and NH 3 and N 2 O as secondary emission species. Engine operation with decreasing exhaust temperatures challenges the exhaust gas aftertreatment system. The trade-off between NO X raw emission reduction and catalyst conversion efficiency losses requires application-specific investigations to achieve the lowest tailpipe emissions.
Finally, the hydrogen combustion engine will be capable of achieving zero-impact tailpipe emissions. The according exhaust gas aftertreatment systems may be less complex than such for gasoline or diesel applications with similarly low tailpipe emission levels. However, for that, the engine and exhaust aftertreatment layout as well as engine and aftertreatment operation strategies have to be closely aligned. Additionally, further research on catalyst materials, operation strategies and degradation mechanisms is necessary.