Figure 3 shows the pre-chamber and cylinder pressure traces at 12:1 compression ratio with fixed spark timing (−25°aTDC). H₂ injection into the pre-chamber starts at −75°aTDC and causes a slight rise in pre-chamber pressure, forcing some of the NH₃-air mixture, residuals from the previous cycle, and hydrogen into the main chamber (Figure 3B). After the end of injection, the piston continues to rise, compressing ammonia, hydrogen, and air into the pre-chamber. Then the pre-chamber mixture is ignited by the spark plug, causing a pressure rise in the pre-chamber, which is a function of the burn rate and pre-chamber geometry. The maximum difference between pre- and main-chamber pressures (ΔP _max) are shown by zoomed in section (A) of Figure 3, where largest difference in pressures occurs for the leanest intake equivalence ratio. These results indicate that the pre-chamber burn rate is greatest for the lean intake condition. The excess air in the intake charge seems to enhance the reactivity of the pre-chamber by decreasing the local ammonia fraction, and converging local equivalence ratios towards stoichiometry. In contrast, the stoichiometric intake condition introduces more ammonia into the pre-chamber, decreasing hydrogen fraction and creating a more fuel-rich equivalence ratio, which results in slower burning and a reduced ΔP _max. Turbulent jets are driven by the pressure difference and emerge into the main-chamber, igniting it and initiating flame propagation across the cylinder bore. Although ϕ = 0.6 in Figure 3 had the greatest ΔP _max, and thus the most potent and distributed turbulent jets for main-chamber ignition, the main-chamber pressure was much lower than the other cases since the flame speed in the main-chamber was compromised by the excess air, leading to incomplete combustion.

Figure 4 compares the turbulent jet strength across different conditions using the maximum difference in pre- and main-chamber pressures (ΔP _max) as a metric. Each subplot shows the results of a given compression ratio (12:1, 16:1, and 20:1) over a range of equivalence ratios at fixed spark timing (25°bTDC) and at MBT spark timing. The cases represented in Figure 3 are circled in Figure 4A. At a fixed spark timing, ΔP _max increases with decreasing equivalence ratio suggesting higher pre-chamber reactivity because of decreased ammonia fraction and closer-to-stoichiometric equivalence ratios. At MBT spark timing, ΔP _max generally increases as equivalence ratio is decreased from 1 to 0.8, but then remains somewhat consistent as equivalence ratio continues to decrease. At 12:1 (Figure 4A), lean equivalence ratios require a more advanced timing for optimal gIMEP, sacrificing pre-chamber reactivity for additional time to propagate the flame in the main chamber. Conversely, at 20:1 (Figure 4C), most cases require a retarded spark timing to optimally phase combustion, which occurs more rapidly at high compression ratio with the possibility of an auto-ignition event in the main-chamber. The 16:1 compression ratio (Figure 4B) represents a sort of middle ground where MBT spark timing is slightly retarded for ϕ > 0.7 and advanced for the most lean cases.

Figure 5 shows the net heat release rates (NHRRs) for the three compression ratios tested where columns of subplots from left to right correspond to 12:1, 16:1, and 20:1, respectively. The top row of subplots in Figure 5 show the cases with fixed spark timing (25°bTDC) and the bottom row shows the cases with MBT spark timing. The corresponding figure containing the pre- and main-chamber pressure traces can be found in the Supplementary Material. Within each plot, NHRRs are shown along with the crank angle of 50% total heat release (CA50, dashed vertical lines), illustrating the burn rate and combustion phasing, respectively. Generally, NHRR exhibits an initial peak soon after spark timing which constitutes the pre-chamber burn and the jet emergence into the main-chamber; thereafter, heat is released in the main-chamber as the flame consumes the remaining fuel-air mixture. For the 12:1 cases (left column, Figure 5), the NHRRs all appear parabolic, or with a slight plateau, indicating that the heat release is gradual and primarily driven by flame propagation. At increased compression ratios (middle and right columns, Figure 5) most conditions exhibit a dual-peak heat release, where combustion occurs initially as flame propagation, but then the rate of heat release increases drastically around TDC by a supposed auto-ignition event.

At 12:1 compression ratio, the MBT spark timing happens to be the same as the fixed spark timing (25°bTDC) for the ϕ = 1 and ϕ = 0.8 cases. The NHRR for the stoichiometric case begins gradually and develops into a plateau-like shape as the flame propagates across the cylinder. The ϕ = 0.8 case is ignited by stronger turbulent jets, as shown in Figures 3, 4, thus the NHRR initially rises quickly with a small peak around −18°aTDC indicating a faster, more energetic ignition compared to the stoichiometric case. Then the main chamber burns by flame propagation, characterized by the parabolic NHRR (Figure 5A). Interestingly, CA50 is more advanced for the ϕ = 0.8 case compared to ϕ = 1, suggesting that the stronger turbulent jets and more energetic ignition compensate for the slower flame speed. The ϕ = 0.6 case at the fixed spark timing of 25°bTDC (Figure 5A) shows a strong initial peak in NHRR, but heat release from the slowly propagating flame is poor. Advancing the spark timing to 35°bTDC yields MBT (Figure 5D) where the magnitude of the initial NHRR spike is reduced slightly, but there is sufficient time for combustion to be completed in the main chamber. As a result of advancing spark timing, the CA50 for the ϕ = 0.6 case is also most advanced compared to the other mixtures at MBT spark timing.

A dual-peak NHRR is apparent for several cases when the compression ratio is increased to 16:1 (middle column, Figure 5). At fixed spark timing (Figure 5B) the ϕ = 1 and ϕ = 0.8 cases show somewhat gradual NHRRs until ≈ 2–3°aTDC, at which point both cases show a rapid increase in NHRR. This pattern of heat release is comparable to a SACI operating mode (Robertson and Prucka, 2019) where this second peak in NHRR corresponds to auto-ignition of the unburned portion of the gas mixture. The ϕ = 0.6 case at fixed spark timing does not exhibit such an obvious second peak in the cycle-averaged NHRR, suggesting that the ignition delay is longer than the engine residence time, thus heat is released by flame propagation. Similar trends are noted for MBT spark timing at 16:1 compression ratio (Figure 5E). Spark timing is retarded slightly for the ϕ = 1 and ϕ = 0.8 cases which cause the secondary peak in NHRR to be delayed to ≈ 7–8°aTDC and less prominent in magnitude. On the other hand, spark timing was advanced slightly for the ϕ = 0.6 case; a strong initial spike in NHRR is noted immediately after spark from TJI and then a parabolic heat release rate suggests flame propagation mode without auto-ignition.

All of the 20:1 compression ratio cases (right column, Figure 5) show NHRRs with clear secondary peaks which correspond to auto-ignition events in the main chamber. The ϕ = 1 case at fixed spark timing (Figure 5C) shows a gradual increase in NHRR after spark until the first peak at around −12°aTDC. After a slight decrease in NHRR, it then increases rapidly to greater than 50 J/° by a strong auto-ignition event. Interestingly, the ϕ = 0.8 and ϕ = 0.6 cases seem to auto-ignite at about the same crank angle, but peak NHRRs occur later and of lower magnitude. For MBT (Figure 5F), spark timing is retarded for all equivalence ratios to delay auto-ignition and reduce the magnitude of the second NHRR peak. As a result, the local minimums between the first and second NHRR peaks occur around TDC and the auto-ignition heat releases occur almost entirely during the expansion stroke. Thus, overall combustion phasing occurs later for MBT spark timing as shown by CA50 which is ≈5°aTDC for all equivalence ratios. Interestingly, the local minimum in NHRR between the first and second peaks (around TDC) in Figure 5F is lower than at the 16:1 compression ratio (Figure 5E). Such a local minimum is non-existent in the 12:1 cases. It is probable that this decrease in NHRR is caused by a partial quenching of the propagating flame as the piston approaches TDC, which forms a narrow channel between the flat-top piston and the firedeck. The decrease in NHRR is more dramatic at increased compression ratios since the distance between the piston and the firedeck decreases. In fact, at 20:1 compression ratio there is only ≈5.4 mm between the two surfaces, which would lead to greater heat losses and quenching as the flame propagates.

Although these results suggest auto-ignition phenomena, ringing intensity remains low ( < 5 MW/m²), thus avoiding engine mechanical damage and excessive combustion noise. Auto-ignition does, however, add a layer of complexity for engine control where spark timing impacts pre-chamber reactivity (and by extension, the ignition energy deposited into the main chamber) as well as the auto-ignition timing in the main chamber. Furthermore, though out of the scope of the present work, the distribution of hydrogen in the main chamber, driven by the hydrogen direct-injection strategy into the pre-chamber, influences flame propagation speed and the likelihood of auto-ignition.

Key emissions from the experiments are shown in Figure 6. Unburned NH₃ emissions (Figure 6A) are lowest across all intake mixtures at the 12:1 compression ratio. This result is in agreement with previous studies (Westlye et al., 2013; Mounaïm-Rousselle et al., 2022) which attribute the increased NH₃ emissions to increased cylinder pressures at elevated compression ratios, forcing more of the unburned gas mixture into engine crevices. Additionally, a quenching mechanism through the narrow clearance volume at TDC, as previously discussed, is more significant at increased compression ratios and may contribute to NH₃ emissions. As for NH₃ trends with intake equivalence ratio, minimum NH₃ emissions are achieved at ϕ = 0.8 for the 12:1 compression ratio. This result contradicts the intuition that decreased equivalence ratio will have lower combustion temperatures, slower flame propagation, and will not burn as close to the engine walls; all of which point to increased ammonia emissions with decreasing equivalence ratio. On the other hand, decreased equivalence ratio means lower ammonia fraction in the gases trapped in the engine crevices and greater pre-chamber reactivity with improved ignition of the main chamber; both of which would decrease ammonia emissions. It appears that these competing factors are optimized at ϕ = 0.8 for the 12:1 case. For the increased compression ratios, NH₃ emissions are fairly consistent to slightly decreasing with leaner intake equivalence ratios which suggests that lower ammonia fraction in the crevices and improved pre-chamber reactivity are dominating factors for ammonia emissions at high compression ratio.

Figure 6B shows that minimum N₂O emissions occur around ϕ = 0.7 for 12:1 and 16:1 compression ratios, but that the trend is somewhat random for the 20:1 compression ratio. These trends at 12:1 and 16:1 compression ratio can be explained by previous work which investigated nitrogen-based emissions of ammonia combustion at engine-like conditions a 0-D chemical reactor (Northrop, 2023). A re-creation of the emissions contours in ϕ-T space proposed by Northrop is presented in Figure 7, which considers ammonia combustion at 20 bar for 1 ms residence time using the mechanism proposed by Stagni et al. (2020). The ammonia-air mixtures investigated in this study are pre-heated to 600–1200 K by compression prior to ignition yielding adiabatic flame temperatures of 1700–2150 K for ϕ = 0.5 and 2,300–2,700 for ϕ = 1, as shown in Figure 7. Therefore, aside from the most lean equivalence ratio where N₂O formation can be explained by cool flame temperatures, these experimental results imply that N₂O must form elsewhere in the combustion chamber besides the propagating flame front. One possible formation mechanism, which is supported by the analysis of NHRR in Sec. 3.2, is by flame quenching on the engine walls and heat transfer near the walls. As the flame propagates, heat losses likely form a temperature gradient in the flame front where temperatures decrease near the walls to form N₂O locally by low temperature combustion. Immediately next to the walls, flame quenching occurs where N₂O, an intermediate of the NH₃ combustion mechanism (Kobayashi et al., 2019), becomes frozen in the combustion products. The effect of local low-temperature combustion and quenching is represented by the green highlighted region in Figure 7. Another formation pathway may manifest during during the expansion stroke when the unburned gases (NH₃ and air) are released from the crevices to induce the thermal de-NO_x mechanism (the reduction of NO_x to N₂ and N₂O in the presence of NH₃). However, 0-D analysis suggests that this pathway is insignificant by comparison to low-temperature zones and quenching (Northrop, 2023).

Fuel bound nitrogen opens an additional pathway for NO_x formation, causing it to form with shorter residence time at lower temperatures compared to hydrocarbon fuels (Westlye et al., 2013; Ryu et al., 2014). Figure 6C shows that NO_x emissions are generally high and have an opposite trend to the previously discussed emissions, peaking at slightly lean equivalence ratios. Figure 7 shows that the 10,000 ppm contour of NO_x intersects most with the adiabatic flame temperature envelope at equivalence ratios of 0.7–0.9, which correspond to the highest NO_x emissions in the experiment shown in Figure 6C. This suggests that the propagating flame front, approximated by the adiabatic flame temperature, is likely where NO_x forms. NO_x emissions do not exceed 6,000 ppm in any of the experiments since low-temperature combustion regimes and quenching likely occur near the walls, resulting in an overall lower concentration of NO_x in the engine exhaust.

Given the high NO_x emissions, it is likely that a NH₃-fueled internal combustion engine will require a SCR catalyst to keep NO_x emissions within legislated limits. The unburned ammonia in the exhaust can be utilized as a NO_x reductant where a NO_x to NH₃ ratio of unity is ideal to fully mitigate both species. This ratio is shown in Figure 8A. In all test cases, NO_x could be fully mitigated since there is an excess of NH₃ available. However, ϕ = 0.8 at the 12:1 compression ratio stands out as it is closest to unity whereby NO_x could be completely mitigated along with ≈70% reduction in NH₃.

Sufficient heat is produced by the combustion process to achieve SCR light-off temperatures of 300°C–450°C, as evidenced by Figure 8B. The exhaust temperature is measured ≈10 cm downstream of the exhaust valve, assessing the potential for a close-coupled catalyst. Equivalence ratios above 0.6–0.7 are suitable at the presented engine conditions to heat the catalyst. These results suggest that a production engine with further optimization (e.g., centrally mounted pre-chamber, and fewer engine crevices compared to the CFR) could be tuned to produce the proper ratio of NO_x and NH₃ and mitigate both of these emissions passively over an SCR catalyst. Additional, optimization of the in-cylinder combustion process is needed to better mitigate engine out N₂O.