Experiments in Fluids

, Volume 40, Issue 6, pp 836–846

Ballistic imaging of the near field in a diesel spray


    • Department of Combustion PhysicsLund Institute of Technology
  • Megan Paciaroni
    • Department of Combustion PhysicsLund Institute of Technology
  • Tyler Hall
    • Division of EngineeringColorado School of Mines
  • Terry Parker
    • Division of EngineeringColorado School of Mines
Research Article

DOI: 10.1007/s00348-006-0122-0

Cite this article as:
Linne, M., Paciaroni, M., Hall, T. et al. Exp Fluids (2006) 40: 836. doi:10.1007/s00348-006-0122-0


We have developed an optical technique called ballistic imaging to view breakup of the near-field of an atomizing spray. In this paper, we describe the successful use of a time-gated ballistic imaging instrument to obtain single-shot images of core region breakup in a transient, single hole atomizing diesel fuel spray issuing into one atmosphere. We present a sequence of images taken at the nozzle for various times after start of injection, and a sequence taken at various positions downstream of the nozzle exit at a fixed time. These images contain signatures of periodic behavior, voids, and entrainment processes.


Core region

The near field spray zone that is characterized by large (>100 μm) fluid structures near the axis that generate primary droplets in the process of breakup

Primary breakup

The destruction of the core region (among some researchers, “primary breakup” implies a specific breakup mechanism, but we make no implication regarding mechanisms)

Primary droplets

Droplets that have clearly originated from the core


Weber number of the liquid, We≡ (ρU2L)/σ


Ohnesorge number, \(Oh\equiv \mu_\ell/\sqrt{\rho_\ell \sigma_\ell L}\)


Liquid density (kg/m3)


Gas density (kg/m3)


Characteristic velocity (m/s)


Characteristic length (usually nozzle exit diameter) (m)


Surface tension of the liquid (N/m)


Viscosity of the liquid (kg/ms)

1 Introduction

Diesel engines are of interest because in practice they offer significantly higher cycle efficiency than do spark-ignited engines, and they have very low CO emissions. Unfortunately, the emission of soot and NOx from a traditional diesel engine is hard to control since the process of fuel–air mixture formation has a significant influence on flame progression. The majority of the heat release occurs in a nonpremixed flame front that surrounds a multi-phase fuel jet that is vigorously entraining the surrounding gas. For classical diesel engines, there is a difficult tradeoff between soot and NOx formation (Heywood 1988); engine or combustion modifications that decrease one of these will typically increase the other. As described by Flynn et al. (1999); NO is formed at the hot nonpremixed flame front and in the high temperature combustion products, while soot is most likely formed in the fuel-rich core of the spray. Fuel–air mixture preparation is thus critical for engine performance. New injector technologies are emerging, but the understanding of primary spray breakup (definitions of terms such as these are provided in the Nomenclature section of this paper) remains incomplete.

Primary breakup in the near field of atomizing sprays has not been studied in detail because a dense fog of droplets obscures the interior of the near field. Until recently, theories about primary breakup of the jet have thus been based upon secondary observations such as the behavior of droplets on the spray periphery. Uncertainty about the liquid core (if one exists, exactly what it consists of and how it might behave) is most pronounced in transient (e.g. diesel) fuel sprays. Smallwood and Gülder (2000), in fact, have postulated that there actually is no liquid core to the spray; cavitation within the jet itself destroys it very near the jet exit.

The current level of experimental understanding is typically represented in physical models by factors such as liquid breakup length, spray cone angle, and droplet size and number density at the spray exterior. These are semi-empirically correlated using the Weber (We) and Ohnesorge (Oh) numbers (defined in the Nomenclature section), together with the liquid to gas density ratio (ρg). Current approaches to spray sub-models used in three-dimensional combustion codes typically include a physical model that imposes very large drops (or “blobs”) with specific momentum as the liquid fuel inlet condition. The blobs then break up into finer droplets and vaporize using accepted droplet breakup and vaporization models (Reitz 1987; Reitz and Rutland 1995). These physical models for primary breakup contain parameters that are typically adjusted to match experimental results. This is the most efficient and appropriate way to include such a complex two-phase process in a much larger engine code. The code relies, however, upon experimental observations from existing spray hardware to infer a description of the near field of the spray because experimental results were used to set the constants in the model. To further improve design capabilities will require a more generalized and portable physical model of the primary spray breakup process, and this will require more detailed investigation of the near field of the spray.

Before going further, it is important to note that our results do not support the classic idea of a completely intact “liquid core” that sheds mass through the stripping of ligaments at the edge; ligaments which then break to form primary droplets. The images presented below indicate a more complex combination of processes leading to the onset of more vigorous breakup within several nozzle diameters. In this paper, therefore, the name “core region” will be applied to the near field spray zone that is characterized by large (∼100 μm − ∼ 1 mm) fluid structures near the axis that generate primary droplets in the process of breakup. This statement does not imply that no liquid core exists. The images do indicate liquid structures that are much larger than the droplets produced by the spray, and in many cases the structures maintain continuity within the field of view. The flowfield is not, however, a fully continuous, totally unbroken stream as implied by the words “liquid core”.

2 Methods for imaging the core region

Sprays have been studied for a significant period of time with both direct imaging and indirect characterization methods (Lefebvre 1989; Faeth 1996; Bachalo 2000). Here, we confine our review of existing methods to those that have been applied to primary breakup of the entire length of the core region of an atomizing spray. In the interest of brevity, related techniques such as classical high-speed shadowgraphy of the overall spray are not discussed.

A research group at Argonne National Laboratory (Powell et al. 2001; MacPhee et al. 2002; Cai et al. 2003; Renzi et al. 2002) has successfully used X-ray absorption techniques to locate phase transitions and provide two-dimensional images of high-pressure fuel spray structure. The fuel spray is illuminated with an X-ray beam generated by a monochromatic synchrotron, in a line-of-sight configuration. Fuel mass locations are determined by the level of X-ray beam attenuation, which is detected by a fast-framing two-dimensional X-ray pixel array detector (PAD) (Renzi et al. 2002). This technique has revealed axially asymmetric flows in the spray, but the small scale structures of this core region are not revealed by the X-ray images published so far. Insufficient X-ray absorption of fuel requires the use of additives, while low signal-to-noise (SNR) levels require averaging over several injection cycles. Furthermore, the PAD detector is sensitive and easily damaged by the high-energy source.

Ballistic imaging, the technique applied here, is a form of optical shadowgraphy that can acquire images through turbid materials that are opaque to simpler imaging techniques. A single ballistic image of the very near field of a spray (a water jet in a LOX injector) was provided by the ballistic imaging group at CUNY (Galland et al. 1995). The spatial resolution of their image was approximately 1 mm. This same group used image-processing techniques to slightly improve that figure (Xiang et al. 1997).

The ballistic imaging instrument used here (described in detail by Paciaroni and Linne (2004) and by Paciaroni (2004)) adapts the time-gated geometry originally developed for medical imaging by the group at CUNY, but it provides significantly better spatial resolution while maintaining high temporal resolution (one laser pulse, no averaging is required). Because this is an optical technique, the instrument does not require a synchrotron. It thus provides high fidelity images in a geometry that can be used by spray researchers in their own laboratories.

3 Time-gated ballistic imaging

When light passes through a highly turbid medium, some of the photons actually pass straight through without scattering, exiting the medium within the same solid angle that they entered (see Fig. 1a). These relatively few photons are termed “ballistic”. Because they travel the shortest path, they also exit first (see Fig. 1b). A somewhat larger group of photons is called the “snake” photon group, because they are scattered just once or twice. They exit the medium in the same direction as the input light but with a somewhat larger solid angle than the ballistic photons. Because they travel a bit further, they exit just after the ballistic photons. Light exiting the medium that has scattered multiply (“diffuse photons”) has a larger photon number density, but it also is scattered into a very large solid angle and it exits last.
Fig. 1

Schematic of ballistic, snake and diffuse photons

Due to their undisturbed path, ballistic photons retain an undistorted shadowgram image of structures that may be embedded within the turbid material. The ballistic photons can thus provide diffraction-limited imaging of these structures. Unfortunately, in most highly scattering and/or absorbing environments, the number of transmitted ballistic photons is often insufficient to provide the necessary SNR to form an image. In such a case, the snake photons retain slightly distorted information and can be used in imaging, together with the ballistic photons, with little degradation of resolution. In contrast, diffuse photons retain no memory of the structure within the material. If allowed to participate in the formation of an image, the various paths these multiply scattered photons take through the material will cause each image point they form to appear as if it came from an entirely different part of the object; this will degrade contrast and therefore resolution. Unfortunately, diffuse photons are the most numerous when light is transmitted through highly scattering media. The problem of obtaining a high-resolution image through highly scattering materials is thus a matter of separating and eliminating the diffuse light from the ballistic and snake light. This can be done using discrimination methods that make use of the properties that are retained by the ballistic and snake light, but are lost in multiple scattering events. These ballistic and snake photon signatures include: (1) the direction taken by transmitted light, (2) preservation of input polarization, (3) preservation of input coherence, and (4) exit time. Each of these properties can help to segregate diffuse photons from the imaging photons. In the work reported here, image segregation is achieved via spatial filtering (to select the light exiting at narrow scattering angles) and selection of the input polarization, together with time gating. The time gate consists of a very fast shutter [an optical Kerr effect (OKE) gate (Wang et al. 1991)] which selects just the leading edge of the image pulse that contains ballistic and snake photons.

The time-gated ballistic imaging instrument used here is shown in Fig. 2. A 1-kHz repetition rate Spectra-Physics Spitfire Ti:Sapphire regenerative amplifier, seeded with a Spectra-Physics Tsunami Ti:Sapphire mode-locked laser, generates 150 fs, 1 mJ pulses centered in wavelength at ∼800 nm. The linearly polarized output beam is split into OKE gating and imaging beams; 30% of the optical power is used as the imaging beam while the remaining power creates a switching beam used to create the OKE time gate.
Fig. 2

Schematic of the ballistic imaging system used in this work

The polarization state of the imaging beam is first linearized with a polarizer, because the OKE gate relies upon polarization switching, and then the polarization is rotated 45° using a polarization rotator. This optimizes switching efficiency for the polarization orientation of the switching beam, as discussed by Ho and Alfano (1979). Next the imaging beam is time delayed using an adjustable length delay arm that allows one to control the delay between the arrival of the switching and imaging pulses at the OKE gate, for optimum time gating. The imaging beam then passes through an optics train consisting of a telescope that controls the imaging beam size at the object, a system to relay the beam through the OKE switch, and a combined spatial filter telescope for imaging onto a display screen. This optical system was been designed and optimized using OSLO, a commercial ray-trace code. By careful choice of available optics, we have ensured that the optical train itself is diffraction limited; there are no spurious aberrations or distortions introduced by the imaging optics themselves.

The OKE gate works in the following manner. When there is no switching pulse present, no image is transferred to the display screen. This is because the OKE gate uses crossed calcite polarizers. The first polarizer in the OKE gate (second polarizer used in the imaging beam) is oriented to pass the polarization orientation of the imaging beam. The second OKE polarizer is oriented normal to the first, blocking an unperturbed imaging beam. The measured extinction ratio of the polarizers is >105; without a switching pulse present there is <10−5 transmission of the imaging beam through the second polarizer. Following the first polarizer, the imaging beam is focused into the Kerr active liquid (CS2 in this case) and then up-collimated past the cell. At the arrival of a switching pulse, the intense electric field of that pulse causes the CS2 dipoles to align along the polarization vector of the switching beam, creating temporary birefringence in the liquid. This birefringence rotates the polarization of the imaging beam, allowing most of it (∼70–75% transmission while open) to pass through the second polarizer. This OKE induced birefringence is limited in time by either the duration of the laser pulse or the molecular response time of the Kerr medium, whichever is longer. In our case, the incident laser pulse is much shorter in duration than the molecular relaxation time of ∼2 ps for CS2; a gate time of 1.8 ps has been confirmed by direct measurement.

The relative timing between an individual imaging pulse and the OKE gate is depicted in Fig. 3. It is important to point out that the image is a schematic; the exact time response of various scattering media to a short pulse remains a subject of research. The shape of the time response curve representing the OKE shutter is based upon a direct measurement that is consistent with a model for the gate (Paciaroni 2004). The shape of the image pulse is based upon a streak camera study conducted by Yoo and Alfano (1990). The shape of this pulse depends upon several parameters such as the collection angle of the imaging system and the relative amount of light lost via absorption. The exact shape for a diesel spray studied by the instrument described here is not currently known. The temporal overlap between the OKE gate and the imaging pulse is set by the time delay retro-reflector system shown in Fig. 2. In practice, this adjustment is done to optimize the image contrast, often starting with a resolution test chart located at the object location and then by some adjustment while sprays are studied if this appears to be necessary. The overlap between the two curves shown in Fig. 3 represents our expectations; it has not been observed directly and is thus somewhat speculative.
Fig. 3

Schematic of relative timing between the OKE gate and the imaging pulse

Past the OKE gate, the image is relayed to a display screen and the image is captured by a Roper Scientific Cascade 650 CCD camera equipped with on-chip multiplication gain. As described in more detail elsewhere (Paciaroni 2004; Paciaroni and Linne 2004), system development work has demonstrated that we can routinely achieve a single-frame spatial resolution between 40 and 50 μm in scattering environments characteristic of diesel sprays (see Sect. 2 for further discussion on spatial resolution). Among the sprays we have studied [a steady turbulent water jet (Paciaroni et al. 2006), a steady water jet in gaseous cross flow (Linne et al. 2005), and the diesel spray described here], atomizing diesel fuel sprays are the most highly attenuating.

4 Diesel spray facility

The spray studied here was produced by a Sturman (Sturman Industries, Woodland Park, CO, USA) diesel fuel injector, which has the capability for multiple injections in a single diesel combustion event. The results presented in this paper were acquired using an on-axis, single hole in the injector nozzle (155 μm diameter with a length/diameter ratio of about 6). The seat-hole type nozzle was manufactured by Bosch and the on axis orifice was produced by EDM (electrode discharge machining). Both commercial diesel fuel and dodecane were evaluated, and the jet issued into still air at ambient conditions (0.8 atm and approximately 300 K).

The overall fuel system is depicted in Fig. 4. The Sturman injector requires two fluid systems: a low-pressure delivery system for the fuel itself, together with a high-pressure hydraulic system that provides the work (or energy) to produce an ultra-high-pressure injection event. Separation of these two systems allows a range of fuels to be used with the injector without compromising hydraulic performance.
Fig. 4

Schematic of fuel delivery system

Fuel was delivered to the injector using a gas-charged piston accumulator. Fuel delivery pressure was quite low (approximately 350 kPa gage) and the piston style accumulator, along with careful fuel handling practices, insured that fuel did not contain an excess of soluble gas due to storage at pressure. The Sturman control mechanism relies upon high-speed solenoid valves that open and close the pathway from a high-pressure hydraulic supply (from 3.4 MPa to 24 MPa) to the large surface area piston of a pressure amplifier. The pressure amplification produced is 7.12–1; needle lift and subsequent reseating within the nozzle are controlled by the hydraulics of the pressure-amplified fuel acting on the relevant surface areas of the needle within the nozzle. Hydraulic oil was delivered to the amplifier at pressures up to 24 MPa. High instantaneous flow rates were required to avoid amplified-pressure loss from a fall in hydraulic pressure, and for laboratory purposes the injector was used in a single shot mode. Thus, a low volume hydraulic pump was used to fill a 9.5 L bladder accumulator which acted as a constant pressure source. Delivery from accumulator to injector used a relatively large diameter (18 mm id) hydraulic line and hydraulic pressure was controlled by a combination of pump outlet pressure and initial fill pressure in the accumulator. For these experiments the hydraulic pressure was varied from 7 to over 20 MPa. This produced fuel injection pressures in the range 50–140 MPa. The sprays presented in this document were all generated with 100 MPa injection pressure.

Synchronization between the transient spray and the imaging system was achieved with a user-controlled delay between the image pulse arrival at the spray and the injection start, with a jitter of 20 μs. Injection start was monitored with a diode laser beam positioned immediately below the nozzle tip; using attenuation of this beam, the timing of image capture relative to spray initiation was quantified with an uncertainty of 2 μs.

5 Experimental results and discussion

5.1 Spray data

The images presented in this paper were all generated under the same injector operating conditions: diesel fuel, 100 MPa injection pressure, a single injection event lasting ∼4 ms overall, injection into still air at 0.8 atm (at the elevation of Golden, CO) and 300 K. No dodecane images are presented here because no serious structural difference was observed between the two fuels. The spray repeatedly displayed three temporal regimes; a spray development period was followed by what was a nearly steady spray lasting for approximately 3 ms, and this steady portion was then followed by a very weak, end-of-injection stream.

Fuel flow rate measurements were made following the Bosch rate of flow meter method (Bosch 1966). For this measurement, a PCB model 111a23 pressure transducer (400 kHz resonant frequency) was placed 5 mm from the injector tip. The outlet tube was approximately 4 mm inner diameter and 5.5 m overall length. Pressure was maintained in the outlet tube at 3.5 MPa. The data were acquired at a 1 MHz sample rate, but they have been filtered to remove frequencies above 10 kHz (see Fig. 5). Note that the ∼3 kHz frequency that is apparent in the results (as well as high frequency content that was removed from the signal) was repeatable when using the single hole nozzle (the nozzle discussed here). Data acquired for a six hole nozzle with the same instrumentation and hardware had none of these frequencies present.
Fig. 5

Rate of injection versus time for the cases presented here; diesel fuel, 100 MPa injection pressure, 10 kHz filtered

These data indicate a velocity during the steady injection period of 370 m/s. In addition, viscosity and surface tension for diesel fuel were taken to be ν=3.5×10−6 m2/s and σ=2.25× 10−2 N/m (McCormick et al. 1997). The measured density was ρ=802 kg/m3, giving a density ratio of ρg=855. These data generate a Weber number equal to We=7.6×105, Ohnesorge number Oh=5.3×10−2 and Reynolds number Re=1.6×104.

5.2 General comments regarding the ballistic images

An example ballistic image of the spray core is shown in Fig. 6. The height of the displayed area depicts roughly 3.5 mm in the object plane. In the image, one can see dark areas representing the fluid phase and light areas representing the gas phase. The top of the image is the location of the nozzle. The jet thus issued from the top and one can see the core region breaking up as the liquid flows downwards. Laser speckle and spurious features that are smaller than the resolution limit of the system (caused by diffraction) can be seen in the gas-phase portion of the image. These should not be interpreted as small droplets. For every case included in this article there was a dense fog of droplets outside the core region, but they are excluded from a ballistic image owing to the optical interaction.
Fig. 6

Typical ballistic image, the spray tip is at the top of the image. This image was acquired during the steady spray period

These images are simply low light level shadowgrams that used extra optics to segregate certain classes of photons for the construction of the image. Hence, like any shadowgram, they are two-dimensional representations of three-dimensional phenomena. In classical shadowgrams of simpler liquid flows, one can find both light regions and dark regions (caused by destructive interference) depicting internal structure. In contrast, the single-shot data presented here are taken with a 1.8 ps shutter, at the limit of the camera’s capacity to image. As a result, one can see no internal structures in most of the current ballistic images. As it passes through the core region the beam is attenuated by both absorption and scattering, although absorption is not dominant in this fuel at this wavelength (Dombrovsky et al. 2003). So many photons are lost by droplet scattering, absorption, and refractive scattering at highly wrinkled interfaces that the camera does not detect what photons remain. The charge count-per-pixel in the dark regions of ballistic images is on the order of 1,000 and in the light regions it is on the order of 60,000. The fact that the dark areas in the ballistic images are at the detection limit is confirmed by comparison to completely darkened images from the same camera. An identical dark grainy texture is evident, at the same signal level. One can not simply increase camera gain to image the interior, however, as the light level that falls outside the liquid structures will saturate the camera and bleed over into other pixels.

A void filled with gas or vapor at the edge of the liquid core would allow some light to pass, and in fact one can clearly detect voids at the edge of the jet in Fig. 6, even when other internal structures are not imaged. These voids allow enough photons to pass so that the camera registers their appearance.

Next, spatial resolution for this system, together with a comparison to other techniques, requires some discussion. Paciaroni and Linne (2004) present a detailed measurement of spatial resolution at various levels of attenuation for the instrument used here. The resolution figure of merit they presented was the full width at half maximum (FWHM) of the measured point spread function (PSF) for the entire optical system, including scatterers (see e.g. Hecht (1998) for a detailed discussion of the PSF). The PSF is a continuous function, meaning that features smaller than the quoted FWHM do not suddenly become invisible. They are certainly detectable, but with reduced image contrast (i.e. with more blur). As features grow successively smaller, image contrast can become eroded to the point where one can comfortably label the feature as “not visible”, but it is a continuous progression from what could be called clearly visible to what could be called invisible. To quote the FWHM of the PSF is a convenient and recognized way to represent much more complex image resolution data, but one should not confuse this specification with much cruder visual estimates of resolution.

These points can be observed in Fig. 6. Note the scale bar of 100 μm. The quoted FWHM of the PSF is almost half that distance. Smaller features can be discerned in the images, but they are indeed a bit more blurred. Blur is manifested by a speckled texture produced by bit noise in the camera. This explains the “fuzzy” appearance of the small features in the images.

Paciaroni and Linne (2004) showed that low levels of attenuation produce better resolution (smaller features are observable) than do high levels of attenuation. This is because high spatial frequency components of the image (those that contribute to fine scale structures in the image) are more easily lost via multiple scattering, and as attenuation increases, more high frequency components are lost. Paciaroni and Linne have shown that for the system used here, the FHWM of the PSF for weakly attenuating sprays is on the order of 20–30 μm, while for highly attenuating sprays it is 40–50 μm.

Within the diesel spray community, the FWHM/PSF value 40–50 μm is sometimes judged unacceptably large. Often, the number 10 μm is quoted as the necessary level of resolution. If primary droplets are indeed 10 μm in diameter then this would cause a problem. An imaging system capable of that resolution limit would not be able to see through the dense fog of ∼10 μm droplets, the droplets would obscure the path. One can easily see this problem with more classical high resolution shadowgrams of diesel sprays in which the core region is always totally obscured. If the goal is to image the core region, then 10 μm is probably too small. Whether or not ballistic imaging can achieve resolution in the range 10 μm<FWHM/PSF<40 μm in a diesel spray is an ongoing subject of investigation.

The only other technique that has succeeded in imaging liquid in the entire core region is the X-ray attenuation technique described above. They have applied the technique to a number of sprays, but they have not reported values for spatial resolution. For transient diesel sprays, the X-ray imaging detector has given poor spatial resolution. It is clear by comparison that the ballistic images presented here have much better performance in this regard. No X-ray images depict voids as does ballistic imaging, for example. Hollow cone sprays were studied by the same group using the PAD detector in four azimuthal positions (Cai et al. 2003). Model-based tomography was used to extract three-dimensional images as a function of time from the basic data. While the spatial resolution in the images appears to be improved when compared to earlier X-ray work, no measurement of spatial resolution is quoted. They do claim, however, to have imaged “sub-millimeter” features, but that is clearly not within 10’s of microns. A crude visual comparison indicates that ballistic imaging offers better resolution in this case as well, although it would be inadvisable to apply ballistic imaging to a hollow cone spray where the liquid sheet is clearly accessible even to normal shadowgraphy.

One must keep in mind that a diesel spray is as highly attenuating as human tissue. Single-shot ballistic imaging at such high optical density is not found elsewhere. Most medical imaging researchers either acquire multiple images and average or they raster scan a single point, detect with a photomultiplier tube, and average. Neither of these approaches is appropriate for a transient event like a diesel spray. Despite some limitations, the technique described here enables imaging of the fluid/gas interface with spatial resolution that is significantly better than other techniques for the core region.

Referring back to Fig. 6, the boundary between the liquid and gas is not as distinct in the ballistic images presented here as it is in a steady jet in cross flow (Linne et al. 2005), for example. This is the case because the jet begins to disintegrate almost immediately. This fact is perhaps the most important finding of this work; that after just two or three nozzle diameters the core region is not an uninterrupted liquid stream according to the classical notion of a fully intact liquid mass that strips ligaments and droplets into a distinct gas phase. Here, the core region appears to consist of masses of liquid interspersed with voids. This dispersed nature of the images is not an artifact of the diagnostic technique described herein. In prior work [see e.g. reference (Linne et al. 2005)] ballistic imaging has visualized a fully intact liquid core when it did exist.

In Fig. 6 and in the other ballistic images presented below, one can see evidence of periodic structures in the core region. One can also see faint evidence for small ligaments, but one can not detect formation of primary droplets as we did previously in a turbulent water jet (Paciaroni et al. 2006). This is likely because the primary droplets formed by this spray are smaller than the limit of spatial resolution for this setup. As mentioned already, one can also detect the appearance of voids and what appears to be air entrainment in some of the images. These observations are discussed in more detail below.

5.3 Observations of transient behavior at the nozzle

The development of the core region over the entire injection event was observed just at the spray tip. Representative images are contained in Figs. 7, 8 and 9. Figure 7 shows the core region during the initial development period. The very early liquid stream develops a roll-up as it encounters the air, and in some cases this roll-up is penetrated by the oncoming liquid stream as it builds momentum (not shown in Fig. 7). Past 10 μs the core tip has left the image area but continues to penetrate the air. At 94 μs, the jet begins to establish the patterns observed during the steady spray. The core region spread angle is characteristic of the steady-spray period, and periodic structures can be observed. Distinct voids also appear intermittently at the edge of the region of high liquid fraction. Finally, by 980 μs voids appear regularly and the steady spray period has commenced.
Fig. 7

Ballistic images of the spray developing over early times; a 2 μs, b 10 μs, c 94 μs, d 980 μs (times are all given after the start of injection)

Fig. 8

Ballistic images during the steady spray; a 1,990 μs and b 2,010 μs after start of injection, followed by images at the end of injection; c 20 μs and d 1 μs before end of injection

Fig. 9

Shot-to-shot variation in the spray. Images were acquired at the nozzle tip; a 1,950 s after start of injection, b 1990 s after start of injection, and c 2,010 s after start of injection

The steady spray period (Fig. 8a, b.) has a much more regular structure. The core region (at this location) has almost the same width from frame to frame and the structure is characterized by the appearance of uniformly-sized voids along the edges of the image. Note that the observable periodic structures are more uniform than what can be found during the development period of the spray.

Finally, the end of injection (Fig. 8c, d) is characterized by a very thin and weak stream that is filled with voids. One can see through the liquid as one would with a more normal shadowgram of a liquid stream, implying that the dense cloud of small droplets has been significantly diminished.

Table 1 contains more details extracted manually from the entire set of 40 images acquired during the set of transient experiments at the nozzle. Note that the main periodic wavelengths presented in the table are just slightly larger than the nozzle exit diameter. As the spray developed, it appeared to generate several of periodic wavelengths, but these may have been caused by coherent addition of smaller structures with shifted phases. In fact, it was not possible to extract recognizable periodic wavelengths from some images even though the liquid/gas interface was obviously corrugated.
Table 1

Data on transient sprays at the nozzle

Time period (μs)

Core width top (μm)

Core width bottom (μm)

Periodic lengths (μm)

Average number of voids

Average void size (μm)



up to 600






















Data for 980 μs are not presented here because an insufficient number of images were acquired at that delay time

Three images taken during the steady spray time period are depicted in Fig. 9, as a way to present variations in the spray from one shot to another. It was not possible to acquire these images at exactly the same delay after start of injection because there is some timing jitter in the spray trigger system, as described above. These images were acquired within the steady period, however, where the system is fairly reproducible.

5.4 Observations of the core region as a function of position during the steady spray

The axial development of the core region with position was observed at about 1.5 ms after the start of the spray (see Fig. 10), by moving the spray 3.5 mm for each frame shown in the figure. Each individual image can almost be connected into one long, thin core region image (although the images were acquired during sequential injection events). Figure 10a is representative of the flow at the nozzle within the steady spray period. It is characterized by periodic structures and voids. Figure 10b (3.5 mm below Fig. 10a) shows more of this behavior at the top but the lower portion suggests the appearance of ligaments, some of which appear to roll up. This observation is not entirely conclusive in this regard, but thinner extended regions do seem to appear. Figure 10c in the sequence appears to contain further evidence of ligaments that roll up in a regular way and in fact they appear to entrain air during the process (see the labels in the figure). Note that Fig. 10c. also shows some narrowing of the core. Most of the images at this position indicate similar behavior. One can detect weaker examples of this phenomenon at the bottom of the top frame (Fig. 10a, see also Fig. 8a, b) and at the bottom of Fig. 10b. The position at Fig. 10g shows evidence of an even more distributed system of liquid mass with large variation from shot to shot. Finally, by the vertical positions of Fig. 10g, the core region appears to have spread horizontally beyond the boundary of the imaging system.
Fig. 10

Ballistic images as a function of position at a fixed delay during the steady period. Approximate relative distances are a at nozzle, b 3.5 mm past nozzle, c 7.0 mm past nozzle, d 10.5 mm past nozzle, e 14.0 mm past nozzle, f 17.5 mm past nozzle, g 21.0 mm past nozzle

The images for this sequence were also interrogated and the relevant data are presented in Table 2. Overall, the average core region width grows with distance as the liquid masses become more distributed. The initial half angle is largest at the nozzle exit and then it relaxes with distance. Measured angles based upon a spray periphery that includes the surrounding droplet fog are typically larger (on the order of 5° half angle for an injector like this), but one would not expect core region angles to match an overall spray angle that includes the droplet field. The small value of the angle at 7.0 mm is reproduced over six image sets; the image presented in Fig. 10c. is characteristic of the entire set. The data also seem to indicate that that the periodic wavelengths double with distance, but again this may indicate subharmonics formed by coherent addition. The number of voids decreases with distance and the average void size increases. Again, an onset of new behavior is suggested by the decrease in the number of voids and the increase in the distribution of liquid mass at 7.0 mm.
Table 2

Data on the spray during the steady period as a function of position

Distance from nozzle (mm)

Average core width (μm)

Average half angle (°)

Average periodic length (μm)

Average no. of voids

Average void size (μm)

























Data past 10.5 mm are not presented here because it became too difficult to define edges and lengths, and to discriminate a void from a roll-up

These are initial data. More detailed experiments and image evaluation are required in order to confirm these observations and comparisons.

6 Conclusions

We have proven the ability of OKE-gated ballistic imaging to acquire two-dimensional, single-shot images of the core region of a transient atomizing diesel spray injected into ambient air. Primary droplets are not observed because they are smaller than the resolution limit of the instrument. Voids are observed, as are spatially periodic structures.


Support for this work is provided by a grant from the Army Research Office via ARO Project Number DAAD19-02-1-0221. The equipment used and partial student support were funded by an NSF Major Research Instrumentation Grant number CTS-9711889. The authors wish to thank Lambda Research for free use of the OSLO software through the University Gratis program.

Copyright information

© Springer-Verlag 2006