The X-Ray FuelSpray FAQ

This FAQ is intended for scientists and engineers interested in the work we do. It attempts to answer general questions about our work,without getting into too many details of the measurements.

For a more detailed FAQ intended for our collaborators, please see the Collaborators FAQ.

  1. Contact Information
  2. Motivation
  3. Technical Overview
  4. Data Acquisition
  5. Fuel
  6. X-Ray Beam
  7. Can Similar Measurements be done using a Lab X-ray Source?

1. Contact Information

Current Members of the Group
Christopher F. Powell Research Scientist, Principal Investigator Energy Systems Division
Alan L. Kastengren Physicist
 Advanced Photon Source
Katarzyna Matusik
 Beamline Scientist
 Advanced Photon Source
Brandon Sforzo
 Postdoctoral Researcher
 Energy Systems Division


Past Members of the Group
Seong-Kyun Cheong, Georgia Tech University
Stephen Ciatti, Energy Systems Division, Argonne National Laboratory
Yong Yue, AVL, Plymouth MI
Zunping Liu, Advanced Photon Source, Argonne National Laboratory
Jinyuan Liu
Xin Liu, Mayo Clinic, Rochester MN
Ramesh Poola
Satyam Parlapalli
Thomas Riedel, Robert Bosch GmBH
Essam El-Hannouny, Argonne National Laboratory
Jian Gao
Seoksu Moon, AIST.
Xusheng Zhang
F. Zak Tilocco, Michigan State University
Nicholas Sovis
Andrew Swantek, ITW
Daniel Duke, Monash University



2. Motivation


The fuel injectors are one of the key components in the combustion process which produces pollutants in automotive engines.  Sprays from injectors determine the distribution of fuel in the combustion chamber, which has a direct impact on the emissions.  If the spray consists of excessively large droplets, for example, the engine may emit unburned hydrocarbons or soot.  If the droplets are too small, they may burn too quickly and produce high temperatures which result in the emissions of nitrogen oxides. Engine manufacturers have found that changes in the fuel injection systems (such as increased fuel pressure and smaller nozzle holes) can lead to significant decreases in pollution from gasoline and diesel engines. For this reason, the structure of the fuel spray has been an active area of research for the last 20 years or so. Most automakers and engine research labs have been using visible light imaging and, more recently, lasers to make measurements of sprays. However, light with a wavelength in the visible region has a very strong probability of scattering from the many droplets in a fuel spray.  The scattering prevents reliable, quantitative measurements, particularly in the region nearest the spray nozzle where the density is highest.  Since this is the region where the spray starts, it is the most crucial region for understanding the structure and formation of the spray.

In 1999 we began a project using x-rays to study fuel sprays. 

3. Technical Overview

X-rays have a much shorter wavelength than visible light, and thus have only a very small probability of scattering from fuel droplets.  This makes it possible to penetrate through the cloud of droplets and make measurements of the core region of the spray, even immediately adjacent to the nozzle. The x-ray radiograph technique is simple but very powerful. As the x-rays pass through the spray, a small fraction of the x-rays are absorbed by the fuel. The number of x-rays that are absorbed depends on the amount of fuel present.  Therefore, counting the number of x-rays before and after they pass through the spray allows us to accurately determine the amount of fuel in the spray.  This quantitative measurement of the mass of fuel in the spray cannot be made with visible light, because the change in the number of photons reaching the detector is due to the combination of both absorption and scattering, and it is very difficult to disentangle the two contributions. 

Because a suitable x-ray camera does not exist, we typically measure an "image" of the spray one pixel at a time.  We probe the spray with the x-ray beam at one particular position in the spray, fire the injector and make our intensity measurements, then move the injector so that the x-rays pass through the spray at another position.  Measurements from different positions are all acquired from different sprays.  These are stored by computer, and afterward the pixels are put together to form a complete image of the mass distribution of the spray.  From these measurements we can determine the ensemble-averaged fuel density throughout the lifetime of the spray.

4. Data Acquisition (this section is out-of-date)

Measuring the x-ray intensity is the job of our data acquisition system, and this measurement is complicated by several factors.  The first is that the time duration of the fuel sprays is short.  A typical spray only lasts a few milliseconds or less, and within this short time the spray changes in size, position, speed, and density. We need to make many measurements within this short period of time so that we can observe the spray throughout its entire lifetime.  This requires measuring the x-ray intensity with a very fast detector that gives a continuous readout of the x-ray intensity.  We are using an Avalanche Photodiode (APD), an x-ray detector which has a time response of a few nanoseconds.  This detector gives a voltage level output which is proportional to the x-ray intensity.  Our data recording hardware must record the output of this detector continuously over a duration of a few milliseconds.

The second complicating factor is that the x-ray beam is pulsed.  The APD responds by giving a voltage pulse for each x-ray pulse with a time duration of a few nanoseconds.  Our data acquisition system must have a sufficiently fast sampling rate so that we can accurately measure the height of this fast pulse.

Another complication for the data acquisition system is that the signal-noise-ratio of our measurement is too small to give us statistically significant results from a single measurement.  We have to average the measurements of multiple (10-100) sprays to get the statistical accuracy we need.  This fact leads to the requirement that the triggering of the data acquisition must be very repeatable.  Our data acquisition must be precisely synchronized (<1 ns) with the x-ray pulses and with the fuel spray, otherwise averaging the results would be meaningless.

For the above reasons, our primary requirements for a data acquisition system are the ability to perform analog-to-digital conversion at a rate of one point per ns or faster, and the ability to do this for durations of several milliseconds.  These requirements led us to the choice of a fast digitizing oscilloscope.  We considered other options (such as Flash-ADC and digitizer boards) but none offer the advantages of a scope, such as the very long record lengths and the ability to perform averaging onboard.  Onboard averaging is an advantage to us because data transfer from the digitizer to disk storage is typically the slowest step.  With onboard averaging we only need to perform this time-consuming operation once.

For my group, choosing among the available oscilloscopes depends on a single factor: triggering rate in averaging mode.  To build a complete mapping of the mass distribution within the spray we must measure thousands of pixels, and at each pixel we must average the results of tens of sprays.  Fuel injectors are designed to fire fifty times per second or more, but current oscilloscopes can aquire and average long record lengths at a rate of only a few times per second.

Our first oscilloscope for this application was a Tektronix TDS 784D, which we bought in 1999.  In averaging mode with a record length of 1M points, time setting of 1 ms full scale, this scope can trigger at a rate of about 0.9 Hz.  In 2001 we bought a Lecroy WavePro 940, which can trigger at about 2.2 Hz at the same settings.  We recently upgraded to a Yokogawa DL7200, which can trigger in averaging mode at a rate faster than 11 Hz.  Three years ago a complete mapping of a fuel spray required about 5 days of round-the-clock measurements.  This new oscilloscope should allow us to achieve the same results in about 6 hours.

Currently we are using a single detector (APD) and measuring images one pixel at a time.  We are investigating the use of segmented APD's, which would allow us to measure several pixels simultaneously.  For this application, we would use one oscilloscope channel per APD segment, possibly using several scopes in parallel.  This would allow us to increase the speed of our measurements even further.  Our goal is to make it practical for an injector manufacturer to supply us with several different fuel injectors, and within a few days return to them the x-ray measurements that allow them to choose the injector that performs best.   

5. Fuel (this section is out-of-date)

The fuel used in our experiments is a blend of calibration fluid and an additive that enhances the x-ray absorption.  The calibration fluid is obtained from Rock Valley Oil & Chemical Co. located in Rockford, Il, and is either Viscor 1487 for diesel spray experiments or Viscor 16BR for gasoline spray measurements.  Note that our very early experiments (Bosch1, Bosch2, Bosch3, Bosch4) used Amoco No 2 diesel fuel rather than Viscor calibration fluid.

We then blend this fuel with a fuel additive that contains a heavy metal to increase the x-ray absorption. The additive we currently use is obtained from Rhodia and is called "DPX9". We blend 9 parts Viscor with one part Rhodia DPX9.  This gives us a blended fuel with a cerium concentration of about 4% by weight. The blend has physical properties (density, viscosity, surface tension) that are within the normal range for diesel fuels.

In the past we used another additive from Rhodia called "Organo Compound 40". This additive was  more concentrated than DPX9 and allowed us to blend in lower concentrations, but Rhodia no longer makes the OC40.  The Rhodia additives are commercially available compounds that are usually used as fuel-borne catalysts.

We chose to use this additive because it is commercially available in a concentrated form. We need a final metal concentration in the blended fuel of about 4% to give us the contrast in our x-ray images. Other metal additives are available, but they are typically in concentrations of about 8% metal by weight, this would force us to use a 50/50 mixture of fuel and additive, and the properties would be very different from diesel fuel.


Typical Properties of the blend of Viscor1487/Rhodia OC40
Kinematic Viscosity
 2.30 cSt @ 40 C
Kinematic Viscosity 1.03 cSt @ 100 C
Specific Gravity
 0.89 g/ml
Interfacial Tension
 29.9 Dynes/cm
These values are typical, but the fuel used in each measurement may vary.  Each set of data will have slightly different fuel properties.

We have not yet measured other important properties of the fuel blend.  The compressibility (bulk modulus of elasticity) is not known for this mixture.  The compressibility of diesel #2 has been measured and can be found in elsewhere (e.g. ASAE Paper Number 026084).

6. X-Ray Beam

Beam Size

Defining the size of the x-ray beam is a somewhat complicated issue. The x-ray intensity is not uniform across the beam area, it peaks in the center and falls off toward the edges, so any definition of the beam area has some uncertainty.  The way we determine the beam area is by measuring the intensity distribution across the beam in two dimensions.   From these, we can determine the beam width and height, but these can only be defined in terms of FWHM, or FW 10% Max, or  whatever other criterion you choose to define the shape and size of the distribution.

The precise definition of the size of the beam is not particularly important for the quantitative analysis of our data since we report the results in terms of mass per unit area.

Beam Energy

The x-ray beam energy used for the measurements is a critical consideration.  In general, the x-ray absorption by the fuel increases as the x-ray energy decreases (longer wavelength).  However, this is true not just for absorption by the fuel, but also for absorption of anything else in the path of the beam, such as windows, air, and detectors.  Therefore the amount of absorption must be balanced by the x-rays ability to penetrate through your vessel windows and any pressurized gas inside. 

At a photon energy of about 6 keV, we get significant x-ray absorption by the fuel. However, 6 keV x-rays do not penetrate very well, and tend to be absorbed by windows, air, and pressurized gas in the spray chamber. If we move slightly higher in x-ray energy, we can get more flux from the x-ray source, better penetration through materials, and still have significant absorption by the fuel. Therefore, our radiography measurements of fuel sprays usually use x-rays with a kinetic energy from 6- 8 keV.


7. Can similar measurements be done using a laboratory x-ray source?

X-ray radiography measurements can be done using a lab source, but there are some significant tradeoffs.

First, in order to get quantitative information, one must be able to correlate the amount of absorption with the amount of fuel in the path of the beam. This is easy with a synchrotron source and a monochromater: there is no wavelength dependence to worry about and application of the Beer-Lambert law is straightforward. However, if you use a monochromater with a lab source, you have discarded most of your flux. If you use a polychromatic beam, in order to use the Beer-Lambert Law you must account for all the things that depend on the x-ray wavelength: detector reponse, fuel absorption, beam hardening, etc. This becomes very difficult. 

Second, the high flux of the Advanced Photon Source allows us to use a short integration time to "freeze" the motion of the spray. If your sample isn't moving or changing quickly with time, you could integrate for a longer time period to gather the flux that you need to make the measurement. Thus, it might be feasible to make measurements of a continuous spray with a lab source. However, fuel injection sprays for reciprocating engines are intermittent, and have significant dynamic features. You really need the synchrotron source to measure these well.