X-ray Tomography of High Pressure Fuel Spray by Polycapillary Optics

Fuel Spray, X-ray, Tomography, Polycapillary


The study of transient high pressure fuel sprays by X-ray based techniques is worldwide diffused. Synchrotron radiation is successfully exploited for this aim because of its high intensity and pulsed nature. However top-table application are unusual.

This work reports the structure of a gasoline spray from an automotive GDI injection system obtained by X-ray Tomography desktop experiments using an 8 keV Cu Kα X-ray. Polycapillary semilens shaped the divergent X-ray beam into quasi-parallel one allowing to focus the radiation on the investigated spray region. High contrast focus images were collected by a CCD detector for X-radiation.

A 6-hole GDI injector has been coupled to the high pressure pump by a specially designed rotating device able to work up to 25MPa. X-ray absorption measurements have been performed with angular steps Δθ = 1° at the injection pressure of 12.0 MPa. The sinogram reconstruction of the jets by slices permitted to get information about the inner structure of the fuel spray downstream the nozzle tip, where conventional optical techniques are inhibited. A 3D spatial distribution of the fuel emerging from the injector has been obtained. The data have been used to perform spray density measurements. The results concerning the absorption profile along the fuel jets axis and the cross section distribution at different distances from the nozzle have been reported.


X-ray Computed Tomography (CT) is applied in a wide range of industrial applications (i.e. for quality control), where non-destructive techniques are required. It allows to get detailed information about the inner structure of the investigated sample, revealing position and morphology of each inhomogeneity. Recently, attempts to perform X-ray CT investigations on sprays have been performed.

Even if sprays are used in a lot of industrial fields their characterization is quite complicated because a large number of phenomena occurs such as atomization, mixing, coalescence, evaporation. Investigations on the spray structure are generally performed by applying conventional optical techniques in a wavelength ranging from the near ultraviolet (UV) to the infrared (IR) region (~180-900nm) with use of laser-based non-interfering tools.

However, strong limitations occur for high-pressure high-speed sprays, such as in direct injection internal combustion engines, due to the high fuel density reached close to the nozzle and in the core. In these zones both absorption effects and multiple scattering of the incident light violate the approach restricting it to low-dense surrounding areas of the spray.

On the other hand air/fuel mixing control plays a fundamental role for combustion efficiency and exhaust emission reduction 5-9. It has been estimated that Gasoline Direct Injection (GDI) could lead to an improvement of fuel consumption up to 25% if compared with Port Fuel Injection (PFI) system 10. Unfortunately, the reduced mixing time left to fuel and air inside the combustion chamber could induce Air Fuel (AF) ratio locally too rich or too poor and wall impingement phenomena with consequent efficiency loss and increase of particulate matter and unburned hydrocarbon production. Hence, an accurate control of fuel spray development is crucial to completely take advantage of the direct injection in the spark ignition engine.

Moreover, conventional optical techniques give useful information about liquid phase structures and liquid break-up, but they aren’t suitable for quantitative measurement of local fuel densities. This shortcoming can be easily overcome by means X-ray absorption measurements because x-radiation can penetrate liquid fuel structures and provide spatially resolved information along the line of sight.

First studies concerning X-ray attenuation by dense sprays date at the 80’s. In 1984 Gomi and Hasegawa demonstrated that it is possible to evaluate the mass distribution of liquid phase in a water/gaseous nitrogen spray by X-ray absorption method.

Recently, techniques based on X-radiation have been applied to investigate high-density regions of fuel sprays. X-rays penetrate the dense part of fuel spray because of its weak interaction with the hydrocarbon chain due to their low atomic number 13. X-ray radiography and tomography have been used to investigate the core of gasoline and diesel spray by reconstructing the three-dimensional (3D) structure in near nozzle region.

Generally, pulsed high-brilliant, sources like synchrotron radiation are used providing monochromatic beams, and pulsed time-structure and high time resolution. The main limit to utilization of X-ray tube source is the high energy loss when the radiation is converged to obtain a parallel beam.

This paper reports the results of a table-top experiment using a microfocus X-ray source for three-dimensional tomography of high pressure fuel sprays delivered by a 6-hole GDI injector. A Cu Kα X-ray source at 8.048 keV coupled with a polycapillary delivered high flux radiation on the region of interest. A CCD detector for X-radiation collected the emerging signal.

Polycapillary semilenses allowed to shape divergent X-ray beams emerging from the tube into quasi-parallel ones and to acquire high contrast images with a radiation intensity increase of about 1-3 orders of magnitude. The injector was inserted in a rotating device actuated by an electronically controlled stepping motor with the 0.1° angular resolution. The acquisition was carried out on 360° angular trip at the injection pressure of 12.0 MPa in the zone immediately downstream the nozzle. Digital image post-processing allowed to obtain 3D reconstruction of the spray structure.

Absorption profiles were obtained both on the cross and longitudinal section of a single jet.


X-Ray tomography experimental set-up has been realized to detect the characteristics of dense spray coming from a GDI 6 hole injector. Fig. 1 shows a sketch of experimental apparatus. An air-cooled X-ray tube (Oxford Apogee 5000) with a cupper target was used in this study. The accelerating voltage was adjustable from 5 to 50 kV, and the maximum current was 1 mA. The operating conditions for the X-ray tube were fixed as an accelerating voltage of 30 kV and a current of 1.0 mA. The beam stability was greater than 98% for 8 hours.

Fig. 1 X-ray µCT experimental setup for low concentration samples in fast processes and high pressures conditions

A polycapillary semi-lens was used to converge radiation into a quasi-parallel beam reducing the blurring effects to the minimum. Tested in the Xlab-Frascati, a beam profile with a diameter of 4 mm and a residual divergence of 1.4 mrad at the exit of the optics has been measured. The gain factor is 3 order magnitude greater than a pinhole placed at the same distance close to the sample; moreover the lens total efficiency is about 60%, at selected energies (KαCu is approximately at 8.048 keV).

A Photonic Science CCD camera (FDI 1∶1.61) was used,the sensitive area was 14×10 mm² and the pixel size was 10.4×10.4 um2, with 12 bit image digitalization. The spatial resolution was found to be 16.67 um/pixel, the acquisition time was 3ms. The CCD detector was synchronized via external trigger with the injection event.

The injection apparatus consists of a pneumatic pump activated by a pressured gas(injection pressure 2.5-25 MPa), a GDI 6-hole injector (0.193 mm hole diameter) and a programmable electronic control unit for pulses managing and control. Commercial gasoline was used (p=740 kg/m³). It was doped with fuel additive containing Cerium(4% in volume) to enhance the low fuel absorption. The fuel has been injected at 12.0 MPa pressure and 4 Hz frequency in an optically-accessible Plexiglas chamber at atmospheric back pressure and ambient temperature. The injection energizing time has been 4 ms. The start of acquisition has been shifted 500 μs after the start of injection in order to get signal when fuel injection rate is stable.

A homemade high pressure rotating device for injector motion was designed and successfully tested at pressures up to 20 MPa. The system is composed of a fixed part linked to the high pressure pump and a rotating one linked to the injector. It allows a control of the trip on a 360° rotation with a 0.1° precision step. The rotation of the system is induced by a high-torque stepper motor controlled in direction, total angle and angular step. A homemade LabView code allowed controlling the synchronization among injection system, CCD detector and stepper motor. A series of 100 images per 1°was collected for a 360° rotation.

2.1 X-ray Tomography

Tomography is a reconstruction methodology of sample internal characteristics, without physical sectioning. The mathematical basis is the Radon Transformation and its inversion, which can reconstruct a sample starting from an N-dimensional object through N-1 dimensional information: in this paper the whole 3D structure has been reconstructed from a set of 2D X-ray absorption images.

The projections set must be obtained around the sample from different directions. It means that the experimental setup must be able to ensure a relative rotation between the sample and the source-detector system. The projections, containing sample integral information, are analyzed and manipulated to create sinograms. Cross section reconstruction of sample (slice) is obtained by applying the inverse Radon transform or similar algorithm based on “filtered back-projection algorithm” to the sinograms.

The tomography slices have been obtained by means Octopus (inCT Co.) code, a dedicated software working on the projection frames. Finally, high resolution 3D image has been got by the Amira rendering software.


The low X-ray absorption by hydrocarbon chain as gasoline is fundamental to investigate the inner structure of sprays. On the other hand the weak signal induces a low signal to noise ratio. In order to improve up to acceptable levels, the acquired images have been processed by means of a homemade Matlab code.

A contrast enhancing function has been applied to exploit the whole bit depth of CCD Camera. Then, a stop-band filter working on a limited range of frequencies has allowed a strong reduction of signal noise(Notch filter)23. A circular mask has limited the region of interest to the sample rotation zone in order to reduce further noise effect when the sinograms are calculated. Finally, for each angular step, the2D absorption data have been obtained as the pixel by pixel ratio of spray acquisition(transmitted X-ray intensity profile I) versus background (incident X-ray beam intensity I0). The absorption images are processed by means of theOctopus software in order to obtain the reconstruction.

A 3D map of the absorption values can be obtained by merging the slices into a 3D matrix. The absorption is linked to the sample local density p by the well known Lambert Beer law.

where µl is the linear absorption coefficient and l is the crossed spray length. The above equation can also be written as:

where µM is the mass absorption coefficient. Considering the single cross section, M represents the fuel mass m related to the spray cross section area A.

A dedicated Matlab code has been developed to evaluate the absorption profile of the jet longitudinal and cross section.

The slices obtained by tomography processing are collected in a 3D matrix. The software locates the symmetry planes of each jet by using both jet axis and cone angle as input variables. In this paper the density has not been evaluated due to difficulties to precisely evaluate the path length l and the absorption coefficient µl and they will be shown in next work.

Here, the absorption maps are reported as I0/I.


Figure 2, on the left, shows the image of a spray collected by Mie-scattering using visible light source.

Fig. 2 On the left: Mie-scattering image of fuel spray at 120 MPa injection pressure and 300 μs after SOI, the red square correspond to the X-ray absorption ROI. On the right: X-ray absorption image

Gasoline was delivered at 12 MPa injection pressure, 0.1 MPa back pressure and ambient temperature; the image was acquired at 300 μs after the SOI. The white dot line highlights the injector nozzle. It is flat with a conical tip where the hole are placed.

Gasoline was delivered at 12 MPa injection pressure, 0.1 MPa back pressure and ambient temperature; the image was acquired at 300 μs after the SOI. The white dot line highlights the injector nozzle. It is flat with a conical tip where the hole are placed.

The red square represents the region of interest (ROI) investigated by mean X-ray tomography. Figure 2, on the right side reports the image of the X-ray extinction from the six-hole GDI spray at 12.0 MPa injection pressure and 4.0 ms duration. The image refer to the I0/I ratio. The X-radiography of the spray has been carried out with a rotation of the injector of 360° around its axis, step 1°, with image acquisition, and then processed off-line. Each picture is a sum of 100 images, 12 bit resolution. The Start of acquisition by detector was long 3 ms and it was shifted of 500 μs in order to avoid acquisition at the opening and closure of injector nozzle. The top part reveals the tip of the nozzle marked by a white dot line.

Finally, traces of three jets coming from the injector are shown while the remaining jets are hidden in the back. Acquisition of background images have been carried out per each measurement angle in order to reduce the additional noise produced by the deposition of the fuel on the Kapton windows, the residual fog and the injector precession mode during the rotation. Intensity stretching, Notch filtering and contrast enhancing have been applied to process the images. It depicts a clearer area, confines the spray and defines the contours of the jets. This procedure reduces strongly the noise also if it could be further improved increasing the number of acquisition per angular step and/or the angular resolution (to 0.5 or 0.1°).

Looking at figure 2, one of the jets seems to absorb a higher radiation than the others. Firstly, it should be considered that some jets emerging from the nozzle are overlapped, hence greater absorption signal is justified. Furthermore the Gaussian distribution of the beam should be taken in account; even if the images refer to background/signal ratio a loss of signal on the boundary is unavoidable.

The nozzle hole geometry is designed to fill the whole combustion chamber reducing the impact versus walls. For this reason the jets don’t have any symmetry. Particularly, one of the jet has just a little inclination respect to nozzle axis (jet 4 in figure 3).

Fig. 3 Sketch of spray from GDI nozzle and jets position referred to the nozzle axis

Though the rotation axis shows a slight eccentricity, the jet 4 is confined always in the central region of beam. Therefore it has been selected as target for absorption analysis and a mask has been applied to the projection to minimize the interference with signals emerging from the other jets.

The tomography reconstruction process has been explained in the previous paragraph: images are converted into sinograms and then sinograms into horizontal projections (slices). In the case of parallel beam geometry, realized thanks to polycapillary optics, every sinogram produces one reconstructed slice, while, for divergent/convergent beam, more sinograms are needed per slice. The minimum number of angular steps necessary to obtain an acceptable signal to noise ratio is related to the size of the region of investigation. Larger region of investigation requires higher number of view angles.

According previous studies the minimum number of view angles necessary to obtain an acceptable signal to noise ratio is given by the following formula:

Where Nv is number of viewing angles and Np is the number of pixels orthogonal to rotation axis. Therefore a circular mask has been applied to the projections in order to reduce the size of the region of investigation and to perform absorption analysis.

This procedure further limits the ROI to a circle sizing about 4.5 mm in diameter. On the other hand, an accurate investigation of the spray properties can be performed in the zone immediately close to the nozzle orifice, typically hostile for visible light optical diagnostic.

Measurements have been carried out by the Matlab approach described in the previous paragraph.

Figure 4 shows the 2D axial section distribution of the absorption signal for the jet 4 at different distances from the nozzle hole.

Fig. 4 2D Absorption distribution at different distances from nozzle

The values refer to I0/I ratios. They were obtained by a retrieving process from the 3D reconstruction. As expected all the jet cross sections appear circular and the typical Gaussian-shape distribution emerges. The spray is more dense in the core where droplets have larger diameter and disperses on the boundary due to stronger interaction of the air with fuel. Radial jet symmetry can be deduced in good agreement with literature. Interestingly, a slight asymmetry can be observed with the centre placed in the zone closer to the other jets, suggesting an effect of the single jets interaction on the mass distribution. The section radius slightly increases up to the 1500 µm from the nozzle; at 2000 µm section size is the same though the absorption peak is lower. Looking at the effect of distance, the peak of absorption is detected at 1000 µm from nozzle. As reported in equation of Lambert Beer law, the absorption is a function of local density and of cross section radius. Considering that cross section radius hasn’t reached the maximum value at 1000 µm, the increased intensity is due to an increase of the local density in the core of the jet. a rapid grown of values is detected.

Figure 5 reports the X-ray absorption distribution, I0/I in the jet longitudinal plane. X axis belong to a cross section of the jet, y axis is directed in the verse of distance from nozzle. At the nozzle hole exit, absorption immediately increases due to the combined effect of the increasing values of both cross section radius and density up to a peak, at about 1200 µm from the orifice. The fuel is delivered at high pressure, and its density is elevated. In this region the air-fuel interaction is significant just on the spray edge. Later, the signal decreases due to the combined effect of growing spray section, of the loss of momentum flux and of the enhanced air-fuel mixing. Moving from the centre toward the boundaries, a little intensity decrease can be noted at middle radial zone and it becomes more marked at the extreme jet periphery.

Fig. 5 Absorption distribution along jet 4 axis

On the left side of the absorption map shown in figure 5 it is evident an increase of signal due to the presence of another jet.

As mentioned above, the projections have been masked to obtain a more precise reconstruction of the jet 4, however part of the close jet goes in the ROI due to the rotation axis eccentricity.


X-ray tomography has been applied to investigate the inner structure of a gasoline spray delivered by a 6-hole Gasoline Direct Injection system. The experimental set-up is based on a Cu Kα X-ray tube coupled with a polycapillary halflens, that allowed to obtain a high intensity quasi parallel beam (lens total efficiency ~60%).

The technique has provided a detailed reconstruction of the spray structure in the region close to the nozzle where conventional optical techniques are generally prevented.

The 3D data analysis has allowed to get absorption profiles useful to study the fuel mass distribution. As expected, the mass distribution presented a like-lognormal trend along the jet axis and a Gaussian one along the cross section.

Influence of the interaction between the neighbor jets on the mass distribution has been analyzed.