Experimental Analysis of Water Pressure and Temperature Influence on Atomization and Evolution of a Port Water Injection Spray

Port water injection (PWI) is considered one of the most promising technologies to actively control the increased knock tendency of modern gasoline direct injection (GDI) engines, which are rapidly evolving with the adoption of high compression ratios and increased brake mean effective pressure levels in the effort to improve their thermal efficiency. For PWI technology, appropriately matching the spray evolution and the intake system design along with obtaining a high spray atomization quality, are crucial tasks for promoting water evaporation so as to effectively cool down the air charge with moderate water consumption and lubricant dilution drawbacks. In the present paper, a detailed experimental analysis of a low-pressure water spray is presented, covering a lack of experimental data on automotive PWI systems. Phase doppler anemometry and fast-shutter spray imaging allowed us to investigate the influence exerted by the injection pressure level and by the water temperature on spray drop size and global shape, obtaining a complete database to be used for the optimization of PWI systems. The obtained results evidence how significant benefits in terms of atomization quality can be obtained by adopting injection pressure and water temperature levels compliant with standard low injection pressure technologies.


Introduction
The environmental impact, in terms of global greenhouse effect and pollutant emissions, of the automotive sector is considerable. Hence, severe limitations are in force and designed for, in particular, the adoption of challenging CO 2 emission targets for the next few years [1][2][3]. Consequently, the automotive industry is quickly adopting innovative technologies, globally aiming for a rapid increase in automotive powertrain efficiency. The leading method followed to improve powertrain efficiency is electrification, with the adoption of a progressively more significant energy storage capacity and electric propulsion system power. The final stage of this evolution is foreseen to be the battery electric vehicle, which ensures a zero CO 2 local emission operation. In this scenario, the internal combustion engine will still play a significant role for many years [4] due to the complexities and costs related to the actual implementation of the electrification path. Accordingly, along with electrification, a significant evolution of the gasoline engine to improve its efficiency is mandatory. The widespread adoption of gasoline direct injection (GDI) in spray-guided or pre-chamber configurations, turbocharging coupled with downsizing and downspeeding, higher compression ratios and application of the Miller cycle seem to be the most interesting innovation lines [5,6]. Unfortunately, many of the aforementioned technologies cause a drastic increase in the knocking tendency due to the increased charge temperature before and during combustion, thus restraining the potential benefits in terms of engine efficiency. As a matter of fact, particularly for high-performance engines, the knock tendency in high load conditions is currently controlled by reducing the spark advance and by enriching

Materials and Methods
The PWI injector (Bosch EV14) used for the present analysis was characterized in terms of the mean injected mass, global spray evolution, drop size and velocity. The main characteristics of the tested injector are reported in Table 1. The static flow rate was measured by the dINJ injection analyzer, a Zeuch method-type injection analyzer specifically designed to operate with low-pressure injection systems [8,22].
The injector under testing was fed with distilled water statically pressurized with nitrogen in the range from 5 bar,g to 11 bar,g (up to 15 bar,g only for the flow tests). Pressurized water was accumulated in a 100-cc reservoir, which was used as rail to directly feed the injector to have negligible pressure fluctuations during the injection event, according to the rules of JSAE2715. The rail structure was used as a fixture for an electric heater used to control water temperature. The feedback thermocouple was installed at the injector inlet in the same position as the Keller PAA M5 HB sensor (20 bar f.s., 50 kHz bandwidth, 1% accuracy) used to monitor the rail pressure. The injector current time history was acquired by a Pico TA189 probe (30 A, 100 kHz bandwidth) and averaged over 30 consecutive injection events. A schematic of the experimental setup is reported in Figure 1.
In each operating condition, the mean injected volume was measured by a precision balance (Radwag PS 1000/C/2, resolution of 1 mg, accuracy ±1.5 mg) during three repetitions of a 3000-shot sequence with a 10 Hz injection frequency.
In previous studies (e.g., [8,23]), the effect of the test vessel's pressure and temperature on low-pressure water sprays was investigated, concluding that the air temperature's variation from ambient to 50-60 • C (typical of boosted conditions with an intercooler) has negligible effects on the spray's evolution and size. On the other hand, boosted pressure levels have the similar effect of a corresponding injection pressure reduction, with the pressure differential across the injector being the main driving force affecting the spray's evolution. As a consequence, in this research, the test vessel's pressure and temperature were maintained at 1 bar,a and 25 • C for all the operating conditions.
In Table 2, the operating conditions used for the flow test are reported, evidencing the imposed water temperature at the injector's inlet.
The global spray evolution was investigated by a fast-shutter imaging technique, applied according to an ensemble averaging approach. The imaging apparatus was based on a pulsed Nd-Yag laser (Litron Nano L 200-20, 200 mJ/shot, shot duration 6 ns), which was used as a light source. The laser was synchronized with a fast-shutter, high-resolution CMOS camera (Dalsa Genie Nano M4020, resolution 3008 × 4112, 12-bit). According to the ensemble averaging approach, only one image per injection event was acquired at a given delay from the injection event's start (the TTL signal enabling the injector driver was used as a trigger). The statistical analysis of the spray's global development at an assigned Appl. Sci. 2021, 11, 5980 4 of 16 delay was carried out by repeating the image acquisition over a series of consecutive injection events (30 in the present work). The repetition of the aforementioned acquisition sequence at different delays from the injection start allowed for characterization of the complete spray development throughout the entire injection event. The acquired images at the different timings were analyzed off-line by means of a proprietary digital analysis procedure developed in the LabVIEW™ environment, obtaining the spray tip penetration curves and global cone angle according to the JSAE2715 prescriptions. More details about the image analysis procedure are reported in [24,25]. In previous studies (e.g., [8,23]), the effect of the test vessel's pressure and temperature on low-pressure water sprays was investigated, concluding that the air temperature's variation from ambient to 50-60 °C (typical of boosted conditions with an intercooler) has negligible effects on the spray's evolution and size. On the other hand, boosted pressure levels have the similar effect of a corresponding injection pressure reduction, with the pressure differential across the injector being the main driving force affecting the spray's evolution. As a consequence, in this research, the test vessel's pressure and temperature were maintained at 1 bar,a and 25 °C for all the operating conditions.
In Table 2, the operating conditions used for the flow test are reported, evidencing the imposed water temperature at the injector's inlet.   The test plan used for the global spray analysis by imaging was based on a fullfactorial analysis of injection pressure P inj levels of 5, 7 and 11 bar,g with water temperature levels of 20, 55, 90 and 110 • C. All spray imaging tests were carried out with an ET of 5 ms in order to ensure adequate steady flow operation for the injector and evidence the effects of both the water temperature and the injection pressure.
The effect of the water temperature and injection pressure on the drop size was evaluated using a phase doppler anemometer (Dantec Dynamics P80) along a measuring traverse composed of 18 stations. The traverse was positioned at Z = 50 mm downstream of the nozzle plate and aligned with the projection of two of the four nozzle holes on the examined plane to evidence the global spray symmetry. In Table 3, the main specifications for the PDA system used for the tests are reported. The drop size and velocity characteristics were investigated in the operating conditions reported in Table 4. In all PDA tests, an ET of 5 ms was applied, acquiring data in a 30-ms time window after the start of the ET to capture the entire spray evolution, including its complete tail. Prescriptions from JSAE2715 were followed during the tests in terms of spray boundary detection and minimum number of samples per measuring position. The data were collected during 3000 consecutive shots, operating the injector at 8 Hz in each examined station.

Flow Test Results
In Figure 2, the results obtained in terms of the mean injected quantity, parametrically varying both the injection pressure level from 5 to 15 bar,g and the water temperature at the injector inlet from 20 • C to 110 • C, are reported. On the left in the same figure, the results are reported for when P inj 7 = bar,g was assumed as the reference injection pressure level.

Flow Test Results
In Figure 2, the results obtained in terms of the mean injected quantity, parametrically varying both the injection pressure level from 5 to 15 bar,g and the water temperature at the injector inlet from 20 °C to 110 °C, are reported. On the left in the same figure, the results are reported for when Pinj 7 = bar,g was assumed as the reference injection pressure level. From the obtained results, the effect of the water temperature increasing on the injected mass is evident. For a given injection pressure level, the higher the fluid temperature, the lower the injected quantity. With the ET at 5 ms and the water temperature changing from 20 °C to 110 °C, the mass percentage difference ranged from 5.7 % at Pinj = 7 bar,g to 11.5 % at Pinj = 15 bar,g. This is likely to be ascribed mainly to the reduced magnetic force exerted by the solenoid on the injector needle, caused by the increased coil resistance. As is reported in Figure 3, for the reference injection pressure level, the injector current was significantly reduced in high water temperature conditions. Furthermore, the current time history slope-and presumably the needle rise-was slowed at high water temperatures, as is suggested by the delayed occurrence of the needle's fully raised position, evidenced by the sudden slope change in the current profile around 1.5 ms after the ET started.   From the obtained results, the effect of the water temperature increasing on the injected mass is evident. For a given injection pressure level, the higher the fluid temperature, the lower the injected quantity. With the ET at 5 ms and the water temperature changing from 20 • C to 110 • C, the mass percentage difference ranged from 5.7 % at P inj = 7 bar,g to 11.5% at P inj = 15 bar,g. This is likely to be ascribed mainly to the reduced magnetic force exerted by the solenoid on the injector needle, caused by the increased coil resistance. As is reported in Figure 3, for the reference injection pressure level, the injector current was significantly reduced in high water temperature conditions. Furthermore, the current time history slope-and presumably the needle rise-was slowed at high water temperatures, as is suggested by the delayed occurrence of the needle's fully raised position, evidenced by the sudden slope change in the current profile around 1.5 ms after the ET started.  The effect of the fluid's temperature on the injector dynamics was particularly evident for short injection events, during which the injector operated in the so-called "ballistic regime". In ballistic operation, the needle did not attain the fully raised position, and consequently, the flow could not reach a static rate condition. As is reported in Figure 4a, where the flow results are plotted against the injection pressure level (with ET = 2 ms and Pinj = 15 bar,g), the injector's opening phase was drastically altered by high water temperature levels. In these conditions, the progressively reduced available magnetic force caused the mean injected mass to be lower at Pinj = 11 bar,g (and equal for Tw = 20 °C). Only for ET values longer than 3 ms was the expected rising dependence of the injected mass on Pinj obtained, given the presumable attained steady flow conditions and the reduced significance of the needle's opening phase with respect to the entire injection process.  The effect of the fluid's temperature on the injector dynamics was particularly evident for short injection events, during which the injector operated in the so-called "ballistic regime". In ballistic operation, the needle did not attain the fully raised position, and consequently, the flow could not reach a static rate condition. As is reported in Figure 4a, where the flow results are plotted against the injection pressure level (with ET = 2 ms and P inj = 15 bar,g), the injector's opening phase was drastically altered by high water temperature levels. In these conditions, the progressively reduced available magnetic force caused the mean injected mass to be lower at P inj = 11 bar,g (and equal for T w = 20 • C). Only for ET values longer than 3 ms was the expected rising dependence of the injected mass on P inj obtained, given the presumable attained steady flow conditions and the reduced significance of the needle's opening phase with respect to the entire injection process.
where the flow results are plotted against the injection pressure level (with ET = 2 ms and Pinj = 15 bar,g), the injector's opening phase was drastically altered by high water temperature levels. In these conditions, the progressively reduced available magnetic force caused the mean injected mass to be lower at Pinj = 11 bar,g (and equal for Tw = 20 °C). Only for ET values longer than 3 ms was the expected rising dependence of the injected mass on Pinj obtained, given the presumable attained steady flow conditions and the reduced significance of the needle's opening phase with respect to the entire injection process. The obtained flow results seem to suggest that Pinj = 15 bar,g is not a feasible operating condition for this injector with a high water temperature and will no longer be investigated in terms of the spray evolution and drop size.

Spray Global Development
The imaging set-up described in Section 2 was used to investigate the effect of the injection pressure and water temperature on the global spray's evolution. For the sake of brevity, only a short sequence of the spray evolution in the reference condition (Pinj = 7 bar,g, 20 °C) is reported in Figure 5 along with pictures of the fully developed spray under high injection pressure and water temperature conditions ( Figure 6).  The obtained flow results seem to suggest that P inj = 15 bar,g is not a feasible operating condition for this injector with a high water temperature and will no longer be investigated in terms of the spray evolution and drop size.

Spray Global Development
The imaging set-up described in Section 2 was used to investigate the effect of the injection pressure and water temperature on the global spray's evolution. For the sake of brevity, only a short sequence of the spray evolution in the reference condition (P inj = 7 bar,g, 20 • C) is reported in Figure 5 along with pictures of the fully developed spray under high injection pressure and water temperature conditions ( Figure 6).
Appl. Sci. 2021, 11, x FOR PEER REVIEW 8 of 17 As can be observed in Figure 5, four individual conical flow structures emerged from the nozzle at the beginning of the injection process, initially composed of relatively large ligaments. The four structures merged at a distance between 5 and 10 mm from the injector nozzle. After a short transition zone, at a 10-mm distance downstream, the primary break-up process seemed to be completed. The spray structure hereafter appears to be quite uniform and composed of relatively small drops; only in the most advanced and central part of the spray was the presence of significantly large ligaments clearly perceivable, probably originating from the initial injection transient when the flow velocity was restrained by the needle.
When the operating conditions were changed by increasing the injection pressure from 7 to 11 bar,g (Figure 6a,b), the spray structure was evidently altered, with an increased penetration and cone angle. Furthermore, the spray structure appeared to be composed of smaller droplets, with a reduced presence of large ligaments in the spray's bulk. Only in the spray tip were some large blobs originating from the injector's opening transient still present. Conversely, the increase in water temperature (Figure 6c,d) did not seem to produce dramatic changes in the spray's overall structure with respect to the corresponding lowtemperature operating condition. This evidence seems to suggest that in the tested operating conditions, flash boiling was not triggered despite the high temperature level at the injector inlet, possibly due to the progressive cooling of the water inside the injector body. As reported in Figure 7, the injection pressure had a significant effect on both the spray tip penetration and global cone angle for all the examined water temperature levels. Globally, 5 ms after the ET's start, changing the injection pressure from Pinj = 5 bar,g to Pinj = 11 bar,g led the spray tip penetration to increase by 11% at Tw = 20 °C and by 12% at Tw = 110 °C. The spray tip penetration slightly increased from 5 bar,g to 7 bar,g, with this  As can be observed in Figure 5, four individual conical flow structures emerged from the nozzle at the beginning of the injection process, initially composed of relatively large ligaments. The four structures merged at a distance between 5 and 10 mm from the injector nozzle. After a short transition zone, at a 10-mm distance downstream, the primary break-up process seemed to be completed. The spray structure hereafter appears to be quite uniform and composed of relatively small drops; only in the most advanced and central part of the spray was the presence of significantly large ligaments clearly perceivable, probably originating from the initial injection transient when the flow velocity was restrained by the needle.
When the operating conditions were changed by increasing the injection pressure from 7 to 11 bar,g (Figure 6a,b), the spray structure was evidently altered, with an increased penetration and cone angle. Furthermore, the spray structure appeared to be composed of smaller droplets, with a reduced presence of large ligaments in the spray's bulk. Only in the spray tip were some large blobs originating from the injector's opening transient still present.
Conversely, the increase in water temperature (Figure 6c,d) did not seem to produce dramatic changes in the spray's overall structure with respect to the corresponding low-temperature operating condition. This evidence seems to suggest that in the tested operating conditions, flash boiling was not triggered despite the high temperature level at the injector inlet, possibly due to the progressive cooling of the water inside the injector body.
The complete set of results in terms of the spray tip penetration and spray cone angle are reported in Figures 7 and 8 for the 12 examined operating conditions, evidencing the effect on the spray's global development of the injection pressure and water temperature, respectively.
As reported in Figure 7, the injection pressure had a significant effect on both the spray tip penetration and global cone angle for all the examined water temperature levels. Globally, 5 ms after the ET's start, changing the injection pressure from P inj = 5 bar,g to P inj = 11 bar,g led the spray tip penetration to increase by 11% at T w = 20 • C and by 12% at T w = 110 • C. The spray tip penetration slightly increased from 5 bar,g to 7 bar,g, with this trend being more evident in the final part of the spray evolution and at a higher water temperature. When the injection pressure was raised to 11 bar,g, the final penetration increase was more evident, despite the relatively high injection pressure tending to slow the injector opening transient, consequently leading to the initial spray tip velocity being smaller. As a result, the spray penetration for P inj = 11 bar,g was smaller than for P inj = 7 bar,g up to 1.5 ms after the ET's start, in which time the initial penetration gap was recovered. The effect of the injection pressure on the spray cone angle was even more evident; for all the examined water temperature conditions, the spray diffusion angle progressively increased with higher injection pressure levels, obtaining almost parallel trends for this quantity. It is also interesting to observe how the monotonically decreasing trend for the spray cone angle changed its slope after the injector closure around 5.8 ms after the ET's start.
The effect of the water temperature at the injector inlet on the spray evolution is analyzed in Figure 8. As can be observed, only marginal effects were exerted by the water temperature on the spray's global structure for all the examined injection pressure levels. To be detailed, an unclear tendency was observed in terms of spray penetration for P inj = 5 bar,g, while the effect was negligible for higher injection pressure levels. In terms of the spray cone angle, the increase in water temperature seemed to decrease the spray cone angle up to 90 • C, while a further increase to 110 • C seemed to attenuate or revert the trend. Globally, a moderate effect of the water temperature on the spray evolution was observed.

Spray Drop Size and Velocity
The PDA raw data for the measuring station corresponding to the injector axis projection on a plane at 50 mm from the nozzle (coordinates of X = Y = 0; Z = 50 mm) are reported in Figure 9a,b for the reference operating condition P inj = 7 bar,g and Tw = 20 • C. In these plots, all the records relevant to 3000 consecutive injection events are reported (blue dots), along with the average values computed in 0.1 ms time bins (red dots). The ET's start was used as a time reference. As reported in Figure 7, the injection pressure had a significant effect on both the spray tip penetration and global cone angle for all the examined water temperature levels. Globally, 5 ms after the ET's start, changing the injection pressure from Pinj = 5 bar,g to Pinj = 11 bar,g led the spray tip penetration to increase by 11% at Tw = 20 °C and by 12% at Tw = 110 °C. The spray tip penetration slightly increased from 5 bar,g to 7 bar,g, with this bar,g up to 1.5 ms after the ET's start, in which time the initial penetration gap was recovered. The effect of the injection pressure on the spray cone angle was even more evident; for all the examined water temperature conditions, the spray diffusion angle progressively increased with higher injection pressure levels, obtaining almost parallel trends for this quantity. It is also interesting to observe how the monotonically decreasing trend for the spray cone angle changed its slope after the injector closure around 5.8 ms after the ET's start. The effect of the water temperature at the injector inlet on the spray evolution is analyzed in Figure 8. As can be observed, only marginal effects were exerted by the water temperature on the spray's global structure for all the examined injection pressure levels. To be detailed, an unclear tendency was observed in terms of spray penetration for Pinj = 5 bar,g, while the effect was negligible for higher injection pressure levels. In terms of the spray cone angle, the increase in water temperature seemed to decrease the spray cone angle up to 90 °C, while a further increase to 110 °C seemed to attenuate or revert the trend. Globally, a moderate effect of the water temperature on the spray evolution was observed. As can be seen, the drops' diameters and Z-velocity time histories evidence how the spray approached the considered measuring station about 4 ms after the ET's start. The main part of the spray structure flowed through the considered measuring station between 4 and 8 ms. During this time window, the drops' velocity ranged between approximately 5 and 22 m/s, while the observed drops' diameter range was predominantly between 15 and 150 µm. A significant number of drops with diameters up to 350 µm was also observed, possibly related to the defective primary break-up in the first part of the injection process, as was observed in the spray images. The presence of even such a restricted number of relatively large drops inherently had a non-marginal effect on the resulting Sauter mean diameter.
After 8 ms from the ET's start, the velocity data rapidly decreased to values around 2-3 m/s for the drops pertaining to the spray tail. This part of the spray's structure was composed of very small droplets (diameter values below 40 µm) featuring reduced momentum which continued flowing through the observed position for a long time.
When the injector's water temperature was raised to 110 • C, the drops' velocities and sizes time histories for the same measuring station, reported in Figure 9c The effect of the water temperature on the drops' sizes, which was already commented on for the X = Y = 0 and Z = 50 mm position, was substantially confirmed for the entire measuring traverse; a significant but not dramatic effect from 20 °C to 110 °C was observed, with size reductions ranging between 8 and 15 µm in terms of the SMD and between 4 and 9 µm in terms of the MD. Minor consequences of raising the water temperature were observed in terms of the drop count and mean velocity, which were proven to change only for the central positions of the spray structure. As was observed when commenting on the spray images in the same operating conditions in Section 3.2, the water temperature's rise at the injector inlet from 20 °C to 110 °C, assumed to be compatible with standard PFI technology (e.g., injector body, rail and pressure sensor), was not adequate to trigger a net flash boiling mechanism for the injection process. Consequently, the injection process resulted in a spray with a basically unaffected shape but appreciably improved atomization quality. Nevertheless, the drops' size improvement was not as significant as was presumably attainable in the case of completely developed flash boiling conditions [25]. In Figure 10, the results obtained with four different water temperature levels at the injector inlet are reported for the complete measuring travers at Z = 50 mm and crossing the entire spray structure. In these plots, both the mean and Sauter mean diameter values are included, along with the mean velocity and relative drop count. All these quantities were computed with data relevant to the entire 30-ms time-window to have a full portrait of the spray quality in the examined positions.
The effect of the water temperature on the drops' sizes, which was already commented on for the X = Y = 0 and Z = 50 mm position, was substantially confirmed for the entire measuring traverse; a significant but not dramatic effect from 20 • C to 110 • C was observed, with size reductions ranging between 8 and 15 µm in terms of the SMD and between 4 and 9 µm in terms of the MD. Minor consequences of raising the water temperature were observed in terms of the drop count and mean velocity, which were proven to change only for the central positions of the spray structure. As was observed when commenting on the spray images in the same operating conditions in Section 3.2, the water temperature's rise at the injector inlet from 20 • C to 110 • C, assumed to be compatible with standard PFI technology (e.g., injector body, rail and pressure sensor), was not adequate to trigger a net flash boiling mechanism for the injection process. Consequently, the injection process resulted in a spray with a basically unaffected shape but appreciably improved atomization quality. Nevertheless, the drops' size improvement was not as significant as was presumably attainable in the case of completely developed flash boiling conditions [25]. A similar analysis is presented in order to separately analyze the potential effects exerted by an increase of the injection pressure in terms of the spray evolution and atomization quality. In Figures 11 and 12, the results obtained in terms of raw data for the measuring position X = Y = 0 and Z = 50 mm and in terms of the mean for the entire examined traverse are reported.
The injection pressure's effect on the drops' sizes is evident from the comparison of Figure 11b,d, with the Pinj = 9 bar,g drop size consistently reduced for the spray bulk (from 4 to 8 ms) for time bin mean values close to 50 µm for a large part of the considered time window. Higher values were observed only for the very initial part of the injection process and around 7 ms when, presumably, the drops produced during the injector's closing phase approached the measuring station. Both the initial and final transients were characterized by a severe channel flow restriction, with consistent flow velocity reduction and drops breaking up the process penalization. For Pinj = 5 bar,g, relatively large drops were attained at the Z = 50 mm measuring station with an almost constant velocity during the bulk spray time window, suggesting a moderate drag effect by the surrounding air. Conversely, with the higher injection pressure, the mean drops' velocities showed initially lower values, possibly due to a consistent drag exerted on the finely atomized drops, with the mean velocity rising only later when, presumably, the core of the completely developed spray attained the measuring station. A similar analysis is presented in order to separately analyze the potential effects exerted by an increase of the injection pressure in terms of the spray evolution and atomization quality. In Figures 11 and 12, the results obtained in terms of raw data for the measuring position X = Y = 0 and Z = 50 mm and in terms of the mean for the entire examined traverse are reported.
The injection pressure's effect on the drops' sizes is evident from the comparison of Figure 11b,d, with the P inj = 9 bar,g drop size consistently reduced for the spray bulk (from 4 to 8 ms) for time bin mean values close to 50 µm for a large part of the considered time window. Higher values were observed only for the very initial part of the injection process and around 7 ms when, presumably, the drops produced during the injector's closing phase approached the measuring station. Both the initial and final transients were characterized by a severe channel flow restriction, with consistent flow velocity reduction and drops breaking up the process penalization. For P inj = 5 bar,g, relatively large drops were attained at the Z = 50 mm measuring station with an almost constant velocity during the bulk spray time window, suggesting a moderate drag effect by the surrounding air. Conversely, with the higher injection pressure, the mean drops' velocities showed initially lower values, possibly due to a consistent drag exerted on the finely atomized drops, with the mean velocity rising only later when, presumably, the core of the completely developed spray attained the measuring station.
The analysis in Figure 12 confirmed the significant effect exerted by the injection pressure on drops' characteristics. The pressure increase from 5 to 11 bar,g caused a reduction of the SMD in excess of 20 µm in several stations, with minor effects in the spray's periphery. Smaller and more uniform effects were observed in terms of the mean diameter. The drops' mean velocities consistently increased for all the measuring stations as a direct consequence of the injection pressure's increase. As was observed in Figure 11, the velocity increase was observed mainly in the second part of the 4-8 ms time windows, when the fully developed spray crossed the measuring station. It is also interesting observing how, with P inj = 5 bar,g, the drop count profile was clearly asymmetrical, suggesting incipient off-design operation for the used injector. Correspondingly, the drops' atomization quality in the low-count region was particularly poor. The analysis in Figure 12 confirmed the significant effect exerted by the injection pressure on drops' characteristics. The pressure increase from 5 to 11 bar,g caused a reduction of the SMD in excess of 20 µm in several stations, with minor effects in the spray's The analysis in Figure 12 confirmed the significant effect exerted by the injection pressure on drops' characteristics. The pressure increase from 5 to 11 bar,g caused a reduction of the SMD in excess of 20 µm in several stations, with minor effects in the spray's Finally, in Figure 13, the diameter probability density function (PDF) for the droplet population is reported for the examined ranges of the injection pressure and water tempera-ture. As can be observed, both the temperature and pressure variations caused a significant improvement of the atomization quality, with a consistent reduction in the PDF values for the intermediate diameter values (50-150 µm), while only marginal effects were obtained for the few very large drops produced during the initial part of the injection process. To be more detailed, it is interesting to point out how increasing the injection pressure (Figure 13b) from 5 to 11 bar,g also increased the PDF values in the range of 35-50 µm, confirming the modest atomization quality obtainable with low injection pressure values. serving how, with Pinj = 5 bar,g, the drop count profile was clearly asymmetrical, suggesting incipient off-design operation for the used injector. Correspondingly, the drops' atomization quality in the low-count region was particularly poor.
Finally, in Figure 13, the diameter probability density function (PDF) for the droplet population is reported for the examined ranges of the injection pressure and water temperature. As can be observed, both the temperature and pressure variations caused a significant improvement of the atomization quality, with a consistent reduction in the PDF values for the intermediate diameter values (50-150 µm), while only marginal effects were obtained for the few very large drops produced during the initial part of the injection process. To be more detailed, it is interesting to point out how increasing the injection pressure (Figure 13b) from 5 to 11 bar,g also increased the PDF values in the range of 35-50 µm, confirming the modest atomization quality obtainable with low injection pressure values.

Conclusions
Water injection is promising technology for the mitigation of the knocking tendency of high-power density spark ignition engines, simplifying the adoption of high compression ratio levels and reducing the necessity for mixture enriching and spark retarding. The final goal is fully exploiting the efficiency potential of highly downsized, turbocharged spark ignition engines that have been rapidly spreading on the market in recent years.
The mixture temperature's control potential is related to the water's high latent heat of vaporization, and hence, despite the specific technology used to implement water injection, either injected directly in the combustion chamber or in the inlet runner, the water spray evaporation rate is crucial. To be more detailed, rapid and complete water spray evaporation enables controlling the mixture's temperature without significant drawbacks such as excessive water consumption (water/fuel ratios in the range of 0.2-0.5 are quite common for full load operation) or potentially destructive lubricant oil dilution.
Port water injection (PWI) is currently considered the best compromise among system cost, complexity and efficacy. In the present paper, a deep analysis of the effects of both the water injection pressure and temperature on the resulting spray evolution and sizing characteristics was carried out separately. The ranges for both the pressure and

Conclusions
Water injection is promising technology for the mitigation of the knocking tendency of high-power density spark ignition engines, simplifying the adoption of high compression ratio levels and reducing the necessity for mixture enriching and spark retarding. The final goal is fully exploiting the efficiency potential of highly downsized, turbocharged spark ignition engines that have been rapidly spreading on the market in recent years.
The mixture temperature's control potential is related to the water's high latent heat of vaporization, and hence, despite the specific technology used to implement water injection, either injected directly in the combustion chamber or in the inlet runner, the water spray evaporation rate is crucial. To be more detailed, rapid and complete water spray evaporation enables controlling the mixture's temperature without significant drawbacks, such as excessive water consumption (water/fuel ratios in the range of 0.2-0.5 are quite common for full load operation) or potentially destructive lubricant oil dilution.
Port water injection (PWI) is currently considered the best compromise among system cost, complexity and efficacy. In the present paper, a deep analysis of the effects of both the water injection pressure and temperature on the resulting spray evolution and sizing characteristics was carried out separately. The ranges for both the pressure and temperature were chosen in order to be fully compliant with the low-pressure injection technology, with the water temperature and injection pressure below 110 • C and 11 bar,g, respectively. The obtained results can be summarized as follows:

•
The injected mass per shot was significantly affected by the injected water's temperature. Increasing the water temperature from 20 • C to 110 • C caused a decrease of about 7% for the injected mass, depending on the injection pressure level and injection duration. This evidence seems to suggest the possible necessity of redesigning the injector coil in the case of high-temperature operation.
• In terms of the global spray evolution, no dramatic changes were observed for the spray tip penetration or cone angle in the examined injection pressure and temperature ranges which, in the present analysis, did not trigger flash boiling phenomena. Complete flash boiling development, with injection in sub-atmospheric conditions presumably not relevant for water injection operation, would require water temperature levels possibly not compatible with the standard injection components' integrity or durability.

•
The spray atomization quality was sensibly affected by the water temperature's increase, with a benefit in terms of the Sauter mean diameter between 8 and 15 µm in the examined conditions. At the same time, the observed injection pressure effect was more significant, with the Sauter mean diameter reduction being approximately 20 µm and the injection pressure increasing from 5 bar,g to 11 bar,g.
The obtained results seem to confirm that there is still significant potential in the optimization of the water spray characteristics, mainly in terms of the atomization quality. To be more detailed, adequate water temperature control could assist the increase of the injection pressure in obtaining an adequate sizing level to speed up the evaporation process, contributing to water injection technology spreading.

Conflicts of Interest:
The authors declare no conflict of interest.

GDI
Gasoline direct injection ET Energizing time (ms) MD Mean drop diameter (µm) SMD Sauter mean drop diameter (µm) PDF Probability density function PFI Port fuel injection PWI Port water injection T w Water temperature ( • C) P inj Injection pressure, gauge (bar,g)