You are on page 1of 18

Home

Search

Collections

Journals

About

Contact us

My IOPscience

An energy-dispersive technique to measure x-ray coherent scattering form factors of


amorphous materials

This content has been downloaded from IOPscience. Please scroll down to see the full text.
2010 Phys. Med. Biol. 55 855
(http://iopscience.iop.org/0031-9155/55/3/020)
View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 185.22.251.245
This content was downloaded on 09/06/2015 at 09:00

Please note that terms and conditions apply.

IOP PUBLISHING

PHYSICS IN MEDICINE AND BIOLOGY

Phys. Med. Biol. 55 (2010) 855871

doi:10.1088/0031-9155/55/3/020

An energy-dispersive technique to measure x-ray


coherent scattering form factors of amorphous
materials
B W King1 and P C Johns1,2
1 Ottawa Medical Physics Institute and Department of Physics, Carleton University,
1125 Colonel By Drive, Ottawa, Ontario, K1S 5B6, Canada
2 Department of Radiology, University of Ottawa, Canada

E-mail: Brian.King@newcastle.edu.au

Received 1 September 2009, in final form 22 November 2009


Published 14 January 2010
Online at stacks.iop.org/PMB/55/855
Abstract
The material-dependent x-ray scattering properties of amorphous substances
such as tissues and phantom materials used in imaging are determined by
their scattering form factors, measured as a function of the momentum transfer
argument, x. Incoherent scattering form factors, Finc , are calculable for all
values of x while coherent scattering form factors, Fcoh , cannot be calculated
except at large x because of their dependence on long-range order. As a
result, measuring Fcoh is very important to the developing field of x-ray scatter
imaging. Previous measurements of Fcoh , based on crystallographic techniques,
have shown significant variability, as these techniques are not optimal for
amorphous materials. We have developed an energy-dispersive technique that
uses a polychromatic x-ray beam and an energy-sensitive detector. We show
that Fcoh can be measured directly, with no scaling parameters, by computing
the ratio of two spectra: the first, measured at a given scattering angle and
the second, the direct transmission spectrum with no scattering. Experiments
have been constructed on this principle and used to measure Fcoh for water
and polyethylene to explore the reliability of the technique. A 121 kVp x-ray
spectrum and seven different scattering angles between 1.67 and 15.09 were
used, resulting in a measurable range of x between 0.5 and 9.5 nm1 . These
are the first measurements of Fcoh made without the need for a scaling factor.
Resolution in x varies between 10% for small scattering angles and 2% for
large scattering angles. Accuracy in Fcoh is shown to be strongly dependent on
the precision of the experimental geometry and varies between 5% and 15%.
Comparison with previous published measurements for water shows values of
the average absolute relative difference between 8% and 14%.

0031-9155/10/030855+17$30.00

2010 Institute of Physics and Engineering in Medicine

Printed in the UK

855

856

B W King and P C Johns

1. Introduction
X-ray scatter imaging is a new approach to x-ray instrumentation for medicine. Several groups
have been investigating ways to make use of scattered x-radiation to extend the usefulness
of x-ray imaging beyond conventional primary beam imaging (Brateman et al 1984, Harding
et al 1987, Arendtsz and Hussein 1995, Leclair and Johns 1998, 1999, Poletti et al 2002,
Davidson et al 2005, Leclair et al 2006, Van Uytven et al 2007, Griffiths et al 2007, Elshemey
and Elsharkawy 2009).
There are two types of x-ray scattering in diagnostic radiology: coherent scattering,
which is the basis of x-ray diffraction, and incoherent scattering, which is the Compton
effect with correction for electron binding. Attempts have been made to use both coherent
(Harding et al 1987, Davidson et al 2005, Leclair et al 2006) and incoherent (Brateman
et al 1984, Arendtsz and Hussein 1995, Van Uytven et al 2007) scattering for different
imaging tasks, but coherent scattering has several advantages over incoherent. First, it is very
strongly forward directed, making it relatively easy to collect the scattered photons. Second,
because coherently-scattered radiation interferes to give x-ray diffraction patterns, it is strongly
dependent on the structural properties of the material. Since there is significant variation
between the structural properties of different tissues, greater contrast is available using coherent
scattering.
Previous work in our group has analysed the improvement in image quality possible
through using scattered x rays (Leclair and Johns 1998). As an example, it was shown that to
distinguish between a 1 mm target of gray matter versus water, the signal-to-noise ratio can be
multiplied by about 14 by using the coherently-scattered photons as compared to the primary
(Leclair and Johns 1999). This analysis was based on the available data for the scattering
properties of the tissues in question. Specifically, the property of interest is the differential
scattering cross section per electron, de /d, which can be written in terms of the coherent
(Fcoh ) and incoherent (Finc ) scattering form factors:

de
de 0  2
Fcoh (x) + FKN (E, )Finc (x) .
(1)
=
d
d
Here, FKN is the KleinNishina factor and de 0 /d is the classical Thomson cross section for
scattering from a free electron. Both of these are readily calculable and independent of target
material. The incoherent scattering cross section is material dependent but can be calculated
for any given material based on its composition using the independent atom model (IAM). We
use the tabulated atomic values from Hubbell and verb (1979).
The coherent scattering form factor is conveniently tabulated as a function of the
momentum transfer argument x which combines the photon energy E and the scattering
angle :
 
 

E
1
sin
=
,
(2)
x = sin

2
hc
2
where is the photon wavelength, h is Plancks constant and c is the speed of light.
The standard method for measuring scattering form factors is to use a crystallographic
diffractometer. Previous work in our lab has shown that crystallographic techniques have
deficiencies when measuring scattering from non-crystalline materials (Johns and Wismayer
2004). We are investigating alternative methods of measuring coherent scattering form factors.
Previously, we have reported on a method using an image plate and polychromatic x-ray beam
that measures the form factor with low resolution in x (King and Johns 2008). Here, we detail
an energy-dispersive method that gives x resolution still slightly worse than diffractometer
measurements but with improved accuracy when measuring amorphous materials (King and

An energy-dispersive technique to measure x-ray coherent scattering form factors

857

Transmission

BC

Lt

I J
Y

Scatter

Lst

Ltd

Figure 1. Experiment schematic showing apertures for form factor measurements in transmission
(upper) and scatter (lower) configurations. All equipment except the x-ray tube (A), ion chamber
(B) and spectrometer (J) with its aperture (I) are mounted on an optical bench for precise positioning
and alignment. The location and dimension of each item can be found in table 1.

Johns 2009). Instead of measuring the scattering as a function of using monochromatic


x rays, we measure the scattering at a fixed using polychromatic x rays and an energy
sensitive detector. Via equation (2), the form factor is determined as a function of x.
2. Apparatus
Figure 1 is a schematic of our experiment. The measurement of Fcoh requires knowledge of
the transmission, as well as of the scatter spectrum. The geometry shown allows both to be
measured without a change in the source or the detector location, by translating the target
lateral position, Y. This configuration is based on earlier work by our group (Hasan 2003,
Hasan and Johns 2004). An x-ray tube is used as the source and a high purity germanium
(HPGe) spectrometer measures the energy of each photon. The distance from the source to
the target plane is Lst = 69.3 cm, and from the target plane to the detector is Ltd = 75.6 cm.
As shown, the rays make angles and with the source and detector, respectively; their sum
gives the scattering angle, . The target is placed in a custom-made station with Pb apertures
at either end to allow measurement at seven different . This extends the range of x accessible.

2.1. Apertures
All apertures in the system, except aperture G in figure 1, were punched out of Pb sheet with
a Di-Acro mechanical punch (Acrotech, Lake City, MN). Five punches were available. The
dimensions were measured with a travelling microscope and are given in table 2.
Aperture plane D was fabricated with a circular aperture for the transmission configuration
and seven different rectangular apertures, allowing seven different scatter configurations.
Apertures C and F can be translated laterally for alignment in any one of these eight geometries.
This allows for independent measurements of the form factor in different x regions. The
punches used in each configuration are shown in table 3. Narrower apertures were used for

858

B W King and P C Johns


Table 1. Location of objects in transmission and scatter configurations as shown in figure 1. The
Z position refers to the centre of each object. For more information on the transmission aperture
G see section 2.1.

Object

Description

Z position (cm)

Pb thickness (mm)

A
B
C
D
E
F
G
I
J

Source
Ion chamber
Aperture
Aperture
Target
Aperture
Aperture
Aperture
Detector

0
12.8 0.6
13.8 0.6
64.5 0.3
69.3 0.4
76.9 0.3
129.3 0.6
145.1 0.3
145.8 0.6

N/A
N/A
2.0 0.1
3.9 0.1
40.0 0.5
3.9 0.1
See text
3.9 0.1
N/A

Table 2. Available aperture dimensions.

Punch

Shape

W , width (mm)

H, height (mm)

C1
C2
R1
R2
R3

Circular
Circular
Rectangular
Rectangular
Rectangular

1.66 0.03
3.32 0.05
2.08 0.04
3.12 0.07
3.09 0.04

7.66 0.03
7.62 0.06
10.11 0.09

Table 3. Aperture configurations used for form factor measurements. The detector aperture in the
transmission configuration is defined by a separate transmission aperture (G) as shown in figure 1;
aperture I is not the limiting aperture when G is present. The dimensions of the apertures can be
found in table 2.

Configuration

Aperture C

Aperture D

Aperture F

Aperture I

Transmission
Scatter 1
Scatter 2
Scatter 3
Scatter 4
Scatter 5
Scatter 6
Scatter 7

C2
C2
C2
C2
C2
C2
C2
C2

C1
R1
R1
R1
R1
R2
R2
R2

R1
R1
R1
R1
R1
R2
R2
R2

See text
R1
R1
R1
R1
R3
R3
R3

the first four scatter configurations to improve the resolution at small x where more structure
is expected for amorphous materials. Wider apertures were used for the last three scatter
configurations to improve the counting statistics.
Optimal alignment of aperture C was accomplished by using a PRM model D-15 pancake
style ion chamber (Precision Radiation Measurement Inc., Nashville, TN) in place of the target
at plane E and measuring the exposure for each configuration as the aperture was translated
laterally.
Optimal alignment of aperture F was accomplished using a short (6.6 mm thick) water
target and measuring the total number of counts observed by the HPGe detector for each

An energy-dispersive technique to measure x-ray coherent scattering form factors

859

1.3
Measured
Fit

1.1

Aperture Area (cm )

1.2

1.0

0.9

0.8

0.7
30

40

50

60

70
Energy (keV)

80

90

100

110

Figure 2. Typical energy dependence of effective area of detector transmission aperture, Adt ,
caused by tungsten spindles. The fitting function is computed from separate straight lines in the
energy ranges 3069 keV and 69110 keV. Note that the tungsten K-edge is at 69 keV.

configuration as the aperture was translated laterally. It is important that a short target be used
to ensure that the optimal value is found for the centre of the target. If a long target is used,
multiple scattering can cause a false optimum to be found for larger scattering angles.
For measurements in the transmission configuration, where the fluence from the primary
beam is very high, a coarse 5 mm diameter aperture is used at plane I and an additional
aperture (labelled G in figure 1) is used to limit the countrate. This aperture comprises a
5 mm thick Pb sheet with a small pinhole (made with a nominal 0.5 mm drill) followed by an
opposed pair of 6 mm diameter tungsten carbide tipped micrometer spindles mounted directly
behind the pinhole. The effective area of the pinhole referenced to aperture plane I was
determined by computing the ratio of the total counts measured by the spectrometer through
the pinhole to those measured through a larger aperture with known dimensions, incorporating
the inverse square law. A 65 kVp spectrum with 3 cm Al filtration and a current of 0.2 mA was
used. This gave a countrate of 4000 counts per second and 1% deadtime through the larger
aperture.
The micrometer spindles allow precise control of the aperture size for a very small
aperture. The effective area of the detector aperture in transmission mode, referenced to plane
I and denoted by Adt , was measured by computing a further ratio of counts in each energy
bin, measured at the lowest possible current (0.2 mA) at 121 kVp with the spindles in the
open position and in their final position. Countrates and deadtimes were 20 000 counts per
second and 5% in the open position and much smaller in the closed position. Ideally, the
ratio would be energy independent, giving a constant value for Adt . In practice, the tungsten
spindles add a slight energy dependence which is compensated for by fitting to the measured
area as a function of energy. This ensures that the tungsten spindles do not affect the measured
form factors. The effective area corresponds roughly to a rectangular aperture of width
0.02 mm and height 0.5 mm but this can vary slightly each time the aperture is emplaced, so
Adt must be characterized before each transmission spectrum is measured. A typical result of

860

B W King and P C Johns

the measurement of Adt is shown in figure 2. For measurements in the scatter configuration,
transmission aperture G is removed from the bench completely.
2.2. Source, detector and target
The x-ray tube is a MachlettDynamax rotating tungsten anode tube, powered by a Picker
GX550 single phase generator. It was operated at 121 kV and 2 mA using the nominal
0.6 mm focal spot for a total exposure of approximately 600 mAs. The tube is mounted
vertically such that the heel effect does not affect the tube spectrum as seen at different
scattering angles. The tube has a target angle of 12 but has been tilted towards the
detector by 9.83 to reduce the apparent height of the focal spot through foreshortening.
The output of the x-ray tube is monitored using a PTW N30001 ion chamber (PTW Freiburg,
Germany) permanently placed directly in front of the x-ray tube, out of the beam path for
the measurements. The ion chamber position is unaltered across the eight experimental
geometries. Its readings are used to normalize all measured spectra to the same tube output.
All spectra were measured using an Ortec planar HPGe detector (Ortec, Oak Ridge, TN),
with a detector crystal 16 mm in diameter and 10 mm thick. The full width at half maximum
of the detector was measured to be 0.41 keV at 59.5 keV and 0.53 keV at 122 keV (King and
Johns 2007). The detector is located in a Pb casing to reduce the effect of stray radiation. A Pb
aperture, labelled I in figure 1, is placed in the casing to limit the area of the detector exposed
to radiation and to ensure that the radiation is absorbed in the central region of the crystal. The
dimensions of the apertures are given in table 3. The signal chain consists of a Canberra 2024
spectroscopy amplifier with time constant 0.5 s, 8077 ADC and an Aptec 2000 series MCA,
all from Canberra (Canberra Industries, Meriden CT). The MCA is integrated into a personal
computer. Deadtimes in the transmission configuration were between 1.5 and 4% while in the
scatter configurations were 0.2%.
The target is placed on a plastic holder which can be translated laterally using a precision
slide assembly (Velmex, Bloomfield, NY). Liquid targets are placed in a 40 mm long
rectangular container made from polycarbonate with 0.5 mm thick entrance and exit windows.
The container fits securely in place on the plastic holder to ensure a consistent target location.
Solid targets are placed directly on the plastic holder.
2.3. Spectral measurements
Spectra are acquired with bin sizes 0.1 keV and then rebinned so that the spectra for all
configurations are measured on a consistent x mesh. The form factors were measured on a
200 point logarithmic mesh ranging from 0.25 nm1 to 14 nm1 . This allows finer resolution
at small x where more structure is expected in Fcoh .
Some of the x-ray photons that reach the detector will not come from scattering within the
target. In order to remove the effect of these photons, background spectra must be measured
for each . There are two distinct sources of unwanted photons. First, the number of stray
photons that are present throughout the room can be determined by closing all of the apertures
at plane D and measuring the spectrum. Second, the number of background photons that
scatter along the path of the beam, in the walls of the target container, or from the edges of the
apertures, can be determined by measuring the spectrum with an empty target container. In
this case, the spectrum measured must be corrected for the attenuation that these background
photons would experience when travelling through the target in the actual experiment. The
transmission factor is determined by measuring the spectra along the transmission path with
the target present and without.

An energy-dispersive technique to measure x-ray coherent scattering form factors

861

Taken all together, this means that five spectra are required in order to compute the
coherent scattering form factor with a given configuration: two spectra along the transmission
path, one with the target present and one without, two spectra along the scatter path, one
with the target present and one without, and one spectrum with closed apertures to measure
the stray radiation. All measured spectra are normalized to the tube monitoring ion chamber
signal and corrected for MCA deadtime.
Three spectral effects could cause artefacts in the measured form factors. Some of
the photon energy can escape from the spectrometer through interaction processes such
as Compton scattering or fluorescence escape. When this happens, the measured photon
energy is lower than the true energy. It is possible to correct for this effect by measuring a
number of monochromatic spectra to determine a detector response function. However, for Ge
spectrometers, this effect is mostly insignificant except in the very low-energy region of the
spectrum. We investigated two options for dealing with this effect. Option number 1 entailed
implementing a detector response correction based on measured monochromatic spectra while
option number 2 was simply to exclude the low-energy region, below 30 keV, where the effect
is strongest. Both options gave virtually identical results for the form factor measurements.
As a result, we chose to simply exclude the low-energy region from all calculations. We
also excluded energies above 110 keV where a small amount of pulse pileup was present in
the transmission spectra due to their much higher countrates. Furthermore, we will show in
section 3 that the form factor can be calculated by taking the ratio of spectra measured in the
scatter configuration and the transmission configuration. In strongly peaked regions of the
spectrum, such as in the vicinity of the tungsten characteristic peaks, the ratio will be quite
sensitive to minute differences in the locations of the peaks. To avoid artefacts caused by
these slight misalignments, energies between 5760 keV and 6770 keV are excluded from
the analysis.
3. Theory
3.1. Model development
Here, we derive expressions for the number of photons measured by the detector in each
energy range E E + dE in both the transmission and scatter configurations.
In the transmission configuration, if the differential incident fluence per energy interval
at distance Lst is dt0 /dE, then in the energy range E E + dE, the detector will measure
L2st
Adt exp[t (E)Lt ],
(3)
(Lst + Ltd )2
where t is the attenuation coefficient of the target and Adt is the effective area of the detector
exposed to the source in transmission mode as determined, including energy dependence, in
section 2.1.
For the scattering configurations, Y > 0. Let the differential incident fluence per energy
interval at distance Lst be ds0 /dE. The differential x-ray fluence at a particular depth l of
the target before any scattering has occurred is


ds
ds0
L2st
t (E)l

 exp
=
.
(4)
dE
dE L2st + Y 2
cos
dNt (E) = dt0 (E)

The probability of these photons scattering at depth l into a given solid angle  and travelling
through the rest of the target is


de e dV (l)
t (E)(Lt l)
,
(5)
ps (l) =
exp
d At (l) cos
cos

862

B W King and P C Johns

where At (l) is the area of the target at depth l from which scattering can be observed at the
detector, e is the electron density of the target material, dV (l) = At (l) dl is an element of
target volume at depth l, and the solid angle of the detector is given by
 =

Ads cos
,
L2td + Y 2

(6)

where Ads is the physical area of the detector aperture I corrected for its finite thickness for
photons incident at angle . The number of photons observed by the detector in the energy
range E E + dE from the layer dl at depth l is ds At (l) cos ps (l):


ds0 (E)L2st e Ads cos
t (E)Lt de
 2
 exp
d2 Ns =  2
cos
d
Lst + Y 2 Ltd + Y 2



1
1
exp t (E)l
(7)

At (l) dl.
cos
cos
The total number of photons observed by the detector is given by integrating this expression
over all depths:


ds0 (E)L2st e Ads cos
t (E)Lt de
 2
 exp
dNs (E) =  2
cos
d
Lst + Y 2 Ltd + Y 2




Lt
1
1
At (l) dl.

exp t (E)l
(8)

cos
cos
0
The exponential factor in the integral is very nearly 1 for the small angles present in our case
and can be neglected.
The integral then resolves to simply the scattering volume within the

target, Vt = At (l) dl.
The ratio of dNs (E) to dNt (E) in each energy range E E + dE is then







(Lst + Ltd )2 e Vt cos
Ads
ds0
1
de
dNs (E)
 2

exp t (E)Lt
=  2
1
.
2
2
dNt (E)
Adt
dt0
cos
d
Lst + Y Ltd + Y
(9)
For small , the exponential factor can be neglected. We checked the validity of neglecting the
exponential factors in equations (8) and (9) by numerically evaluating the full integral in the
worst case scenario for water: low-energy photons (30 keV) with the largest scattering angle
used in our experiment. We found that the two exponential factors in equation (8) constituted
less than a 1% effect on the volume calculation, justifying our decision to ignore them. Using
the definition of the scattering cross section from equation (1), an expression for the coherent
scattering form factor is arrived at based entirely on measurable and tabulated parameters:
 
1/2

 

L2st + Y 2 L2td + Y 2
Adt dt0 dNs (E)
FKN (E, )Finc (x)
.
Fcoh (x) =
e 0
Ads ds0 dNt (E)
e Vt cos
(Lst + Ltd )2 dd
(10)
3.2. Parameter calculation
The scattering volume Vt and the scattering angles cannot be measured directly in the
experiment. Instead, we have developed a Monte Carlo ray tracing simulation of our
experiment in Matlab (The Mathworks, Natick, MA). We perform a numerical integration
of the volume by uniformly sampling N points within a rectangular parallelepiped target
volume Vtot . For each, a point is chosen within the acceptance area of the detector Ads (this

An energy-dispersive technique to measure x-ray coherent scattering form factors

863

Table 4. Calculated parameters for scattering apertures from the Monte Carlo calculation. The
calculated uncertainties include variations in all experimental parameters.

Configuration

(deg)

(deg)

(deg)

Y (cm)

Vt (cm3 )

Scatter 1
Scatter 2
Scatter 3
Scatter 4
Scatter 5
Scatter 6
Scatter 7

0.91 0.11
1.70 0.11
2.67 0.12
3.33 0.12
5.31 0.13
6.63 0.13
7.94 0.13

0.77 0.09
1.48 0.09
2.36 0.08
2.97 0.09
4.74 0.11
5.95 0.11
7.15 0.11

1.67 0.17
3.17 0.15
5.03 0.15
6.30 0.15
10.05 0.17
12.58 0.17
15.09 0.17

1.07 0.11
2.04 0.11
3.22 0.12
4.03 0.11
6.45 0.11
8.07 0.12
9.69 0.11

0.61 0.13
0.51 0.10
0.39 0.05
0.31 0.03
0.42 0.03
0.31 0.02
0.241 0.017

acceptance area depends on because of transmission effects through the thick aperture I
directly in front of the detector), and a point is chosen from the spatial distribution of the x-ray
tube source strength over the focal spot area Af . We modelled with a two-dimensional
normal distribution with widths determined from pinhole images of the focal spot (Johns and
Schulze 2009). The number of locations Nt for which the path defined by the source, target and
detector points passes through all of the apertures in the system is then calculated. The target
volume Vt is then obtained as Vt = Nt /N Vtot in the limit of large N. Strictly speaking, because
we have included the finite sizes of the source and detector in the Monte Carlo calculation,
the integral actually calculated is the product of Vt , and Ads , but since it is this product that
appears in equation (10), the distinction is unimportant. This volume calculation is similar to
that of Balogun et al (2000) with the exception that we have extended the calculation to model
the effect of rectangular apertures with finite thickness. The effect of thick apertures can be
quite significant for the larger angles and must be accounted for properly.
The calculation treats all apertures as ideal, ignoring any penetration through the collimator
edges. In fact, there would be a slight effect from penetration through the edges of each
aperture, which would become more significant for larger . This would introduce an
additional energy-dependent effect to our measurements as more x rays will penetrate the
aperture edges at higher energies. If this were a significant effect, the Pb K-edge would be
apparent in our measured spectra where the penetration would be most significant. As our
measured spectra did not show any significant effect around the Pb K-edge, we concluded that
penetration through the apertures was insignificant.
The volume calculation has a precision better than 0.5%. The same Monte Carlo
calculation is also used to determine , , and the offset Y as well as the statistical
uncertainties and correlations of all parameters. Uncertainties in the location and dimensions
of every element in the experiment are factored into the uncertainty estimates. The calculated
values are given in table 4.

3.3. Overall fluence ratio


The Monte Carlo calculation of section 3.2 models the finite size of the x-ray tube focal spot
but it does not include any variations in the size or strength of the x-ray beam at different .
We account for this by measuring the overall fluence ratio dt0 /ds0 separately.
We measure the exposure from the x-ray tube with a PRM model D-15 pancake style
ion chamber (PRM, Nashville, TN) in place of the target at plane E for each configuration in
table 3. The signal, S, in each case is proportional to the fluence at the ion chamber and to

864

B W King and P C Johns


Table 5. Measured fluence ratios (transmission/scatter) for scattering apertures. Measurements
for the ratios were performed three times and averaged. The overall fluence ratio is the ratio of the
expected to the measured values.

Plane D aperture

Expected ratio

Measured ratio

dt0 /ds0

Scatter 1
Scatter 2
Scatter 3
Scatter 4
Scatter 5
Scatter 6
Scatter 7

7.2 0.3
7.0 0.3
6.7 0.3
6.6 0.3
9.7 0.4
9.3 0.4
8.9 0.4

7.50 0.12
7.83 0.07
7.26 0.06
7.78 0.10
11.55 0.07
11.03 0.11
10.9 0.2

0.96 0.04
0.90 0.04
0.93 0.04
0.84 0.04
0.84 0.04
0.85 0.04
0.82 0.04

the effective volume, v, of the ion chamber visible through the apertures. In the transmission
configuration,
dt0
St 2 vt
(11)
Lst
while in the scatter configuration,
ds0
Ss 2
vs .
(12)
Lst + Y 2
Taking the ratio of equations (11) and (12) allows an expression to be derived for the ratio
dt0 /ds0 :
 L2st  vs
re
dt0
L2 +Y 2 vt
= st Ss
=
.
(13)
ds0
rm
S
t

The ratio re is the expected scatter to transmission ratio under the assumption that the x-ray
tube fluence is independent of . The ratio rm = Ss /St is the actual scatter to transmission
ratio measured with the ion chamber. The expected ratio was calculated using a Monte
Carlo calculation similar to that in section 3.2. The focal spot was modelled using the same
dimensions as in section 3.2 for consistency.
The overall fluence ratio measured using this procedure is shown in table 5. The ratio
decreases with larger angles, meaning that the x-ray tube output increases for larger . An
assumption with this procedure is that any variation in x-ray tube output is not energy dependent
as the ion chamber is energy insensitive.
4. Results
4.1. Individual scattering angles
We have measured the coherent scattering form factors of water (H2 O, e = 3.3348
1023 cm3 based on mass density at 22.5 C (Lide 2009)) and polyethylene ([C2 H4 ]n , e =
3.255 1023 cm3 , based on measurement of mass density of our sample) to verify that the
energy-dispersive method is configured correctly. The individual form factor measurements
from each scattering configuration are shown in figure 3. Although not shown on the graphs,
uncertainties were also calculated for each point, taking into account uncertainties in all
experimental parameters. To confirm the consistency of the individual measurements, 2
testing was performed, based on the estimated uncertainty in each measurement, for each

An energy-dispersive technique to measure x-ray coherent scattering form factors

2.5

865
Scatter #1
Scatter #2
Scatter #3
Scatter #4
Scatter #5
Scatter #6
Scatter #7

(a)

2.0

form factor (free e / bound e )

0.5

1.5
1.0
0.5
0

(b)
5
4
3
2
1
0
0

x (nm )

Figure 3. Form factor measurements for (a) water and (b) polyethylene from each scattering
configuration. Error bars are not shown for clarity. Note the different range of x displayed for each
material.

value of x. The vast majority of x values were found to be consistent, within a confidence
interval of 95%. For water, only 2 points out of 145 were found to be inconsistent for x between
0.5 nm1 and 9.5 nm1 . Since there are no adjustable parameters in our measurement, this
suggests that our calculations have been implemented correctly. The measured value for each
configuration drops off towards zero at large x as the upper energy limit of the x-ray spectrum
is approached. We truncate Fcoh for x > 9.5 nm1 where the uncertainty is too large to provide
meaningful results.
The measurements for the more structured material, polyethylene, were made to examine
the limits of the x resolution of our equipment. We again found that the vast majority of points
were consistent, based on the same 2 criteria as for water. Excluding the sharply peaked
region around 1.2 nm1 , only 3 points were inconsistent. The differences around 1.2 nm1
are a result of the limited x resolution for the smaller angles. As can be seen in table 4, the
angular resolution of scatter configuration 1 is around 10%, meaning that the sharp peaks
in the polyethylene form factor are poorly resolved for this configuration. As the resolution
improves at higher , the sharper peaks become apparent.

866

B W King and P C Johns

0.25

Fractional error

0.20

0.15

0.10

0.05

ZI

ZE

ZA

r#

Adt

te

dNt

at

dNs

Sc

Vt

Figure 4. Uncertainties in experimental parameters for each configuration from the measurement
of the form factor of water at 50 keV. The subscripts on the W, H, and Z parameters refer to the
aperture plane in figure 1. The uncertainty in dNs /dNt contains the effect of the fluence ratio as
well. All parameters, with the exception of the ratio of the two measured spectra dNs /dNt , Adt
and E itself are energy independent.

The uncertainties were calculated by repeating the Monte Carlo calculation of


section 3.2 as the geometric parameters were varied and constructing a covariance matrix
containing all 29 of the geometric uncertainties. Variances of the ratio dNs /dNt and the
effective transmission aperture area Adt were determined using counting statistics. The
uncertainties in each experimental parameter at a specific energy are shown in figure 4.
It can be seen from this that the most significant source of error in these measurements comes
from variation in Vt . The resulting error in the coherent scattering form factor for water is
shown in figure 5.
The resolution in x is dependent on both the energy and angular resolution of the
measurements via equation (2). The angular resolution is most significant for small while
the energy resolution becomes more significant for large .
4.2. Combining results from individual scattering angles
The individual measurements made with each configuration are combined to yield one dataset
over a larger x range. The energy bins of our individual measurements were chosen such
that the form factor measurements are on a consistent scale. In the overlap regions, we apply
a 2 test to check our individual measurements at each x value for consistency with a 95%
confidence interval. Consistent values were combined using a weighted average based on the
uncertainty of each measurement.
For points where the 2 test shows that the results are inconsistent, we do one of two
things: if an isolated point is inconsistent, we compute an unweighted average of the individual
results. If a region of points (defined as three or more adjacent points) is inconsistent, we use the
single measurement with the highest x resolution under the assumption that the inconsistency
is caused by a difference in resolution, such as in the 1.2 nm1 region of polyethylene.
The combined results are shown in figure 6 compared to previously published values.
Water has been studied by several groups, most extensively by Narten (1970). Although some

An energy-dispersive technique to measure x-ray coherent scattering form factors

867

0.50
0.45
0.40

Fractional error in Fcoh

0.35

Scatter #1
Scatter #2
Scatter #3
Scatter #4
Scatter #5
Scatter #6
Scatter #7

0.30
0.25
0.20
0.15
0.10
0.05
0

5
x (nm1)

Figure 5. Error estimates for individual coherent scattering form factor measurements of water.

Table 6. Average absolute relative difference between the energy-dispersive technique and
measurements from the literature.

Material

Comparison

x range (nm1 )

Dav

Water

Narten (1970)
Kosanetzky et al (1987)
Peplow and Verghese (1998)
Hura et al (2000)
Kosanetzky et al (1987)

0.59.5
0.54.25
0.59.5
0.59.5
0.54.25

0.101
0.086
0.140
0.137
0.198

Polyethylene

differences exist between the measurements by different groups, the general behaviour of the
scattering form factor is known. Polyethylene measurements are compared to those made
using a diffractometer by Hardings group (Kosanetzky et al 1987).
Quantitative comparisons are made to various published results (Fref ) in table 6 by
computing the average absolute relative difference

N 

1   Fcoh (xi )
(14)
1
Dav =

N i=1 Fref (xi )
where N is the number of points that were measured.
5. Discussion
The results of the energy-dispersive measurements are quite encouraging. The good agreement
of the individual measurements with each other, as well as the fact that the measurements
approach the IAM value at large x is very strong evidence of the accuracy of the energydispersive measurement. The good correspondence with previously published diffractometer
measurements is also promising.

868

B W King and P C Johns

2.5

(a)

This work
Narten

2.0
1.5

form factor (free e / bound e )

0.5

1.0
0.5
0

6
This work
Kosanetzky et al + IAM

(b)
5
4
3
2
1
0
0

x (nm1)

Figure 6. Combined form factor measurements of (a) water and (b) polyethylene compared to
published values (Narten 1970, Kosanetzky et al 1987). Error bars are shown for only every 25th
point for clarity.

Setup and characterization of the experiment is quite time consuming. After the system
has been characterized, however, measurements are relatively quick. Each set of measurements
requires a 5-min low-current transmission spectrum to characterize the detector transmission
aperture (aperture G in figure 1), two 1-min transmission spectra, and seven 5-min scatter
spectra. Including the time for adjustment for each configuration, a complete measurement
can be made in approximately 1 h.
Our technique measures the form factor absolutely, with no scaling parameters.
Diffractometer measurements are placed on an absolute scale by fitting the measured values
to the IAM results at large x. It is not readily apparent, however, exactly where the IAM
region begins. If the range of x is not large enough then the scaling can introduce a systematic
shift in the measurement. This was certainly a problem in the results of Johns and Wismayer
(2004) and also possibly with some of the results of Kosanetzky et al (1987). Our energydispersive measurement avoids this problem entirely and allows an independent check that the
experiment is configured correctly.
It is certainly possible to scale the energy-dispersive results to the IAM value at large
x once the region of validity of the IAM has been established. This could be achieved by
scaling the results for the largest scattering angle to the IAM value at large x, then scaling
the next largest results to the refined value in the overlap region, and so on until all of the
scattering angles have been refined. Doing this may provide a more precise value of the

An energy-dispersive technique to measure x-ray coherent scattering form factors

869

geometrical parameters (primarily Vt ) than can be achieved directly from the measured values,
thus giving a more precise measurement of Fcoh . The series of cascaded scaling parameters,
however, would mean that the uncertainties would increase as the calculation progresses to
smaller x values. Additional study would be required to determine whether this would be an
improvement over the current method.
Even if it is advantageous to scale the measured values to the IAM, it should not be done
lightly. The full calculation should be performed first and the self-consistency of the individual
scatter results verified, as well as the region of validity of the IAM. Introducing an arbitrary
scaling parameter would allow potential experimental configuration problems to be masked
if it is used indiscriminately. The scaling should be used as a refinement of the measured
parameters, not a replacement.
Currently, there is very little benefit to be gained from longer exposures to improve the
counting statistics as the uncertainty is dominated by the volume calculation. If scaling to the
IAM value at large x is found to be beneficial, the counting statistics in the scatter spectra may
become the dominant source of uncertainty and longer exposures could provide more accurate
results.
The fairly long target used in these measurements was chosen to provide the best
balance between scattering volume and target attenuation. As a first approximation, using
equation (8), the scatter countrate is proportional to the product of the target length and the
transmission factor. The maximum countrate is then for a target length of 1 which for water
gives a value between 3 and 6 cm. For larger this approximation breaks down, however. The
Monte Carlo calculation shows that for the largest scattering angle used in our experiment,
the scattering volume is contained entirely within about half of the target length. A shorter
target would therefore provide a better countrate at larger angles, at the expense of a reduced
countrate for the smaller angles and larger uncertainties in the volume calculation.
The long target used here may not be applicable to measurements of some tissues as it
might not be possible to find large enough homogeneous samples of a given type of tissue.
In these cases, a shorter target could be used at the expense of larger uncertainties due to the
reduced countrate and larger volume uncertainties. For materials that are available in larger
quantities, such as fat or muscle, the longer target would give better results.
The combination of the individual measurements into a single dataset has advantages and
disadvantages. Since each measured x value (except at very high and low x) was measured at
least twice, in most cases three or more times, any variations that might be present tend to be
reduced in the combined measurement. On the other hand, the difference in the x resolution
between different configurations means that, in some cases, the weighted average will not be
appropriate. Our method of determining when to combine the results works fairly well but
can sometimes cause small jumps in the combined measurements as adjacent points in x space
switch from a single measurement to a combined value.

6. Conclusion and future work


We have developed and characterized a general purpose system to measure coherent scattering
form factors of amorphous materials such as tissues and phantom materials used in imaging.
We have compared our measured results to previous work and found good agreement. Our
energy-dispersive measurements are more reliable than diffractometer based systems for
several reasons: the measurements are made without the need for any scaling parameters,
a wider range of x values is accessible, well into the IAM region, and the transmission
configuration allows a more well-defined and quantifiable background to be measured. In

870

B W King and P C Johns

addition, the thorough characterization of the system has provided quantitative uncertainty
estimates.
We are currently using this system to make accurate measurements of several types of
tissue to develop an accurate form factor library that will aid in the development of the field
of x-ray scatter imaging.
Acknowledgments
This work was supported financially by the Natural Sciences and Engineering Research
Council of Canada. Measurements of spectra were performed by Karl Landheer. The authors
thank Philippe Gravelle in the Physics Department machine shop for general fabrication
assistance and Dr David Rogers for loan of the dual channel electrometer used for measuring
the overall fluence ratio.
References
Arendtsz N V and Hussein E M A 1995 Energy-spectral Compton scatter imagingpart 2: experiments IEEE T.
Nucl. Sci. 42 216672
Balogun F A, Brunetti A and Cesareo R 2000 Volume of intersection of two cones Radiat. Phys. Chem. 59 2330
Brateman L, Jacobs A M and Fitzgerald L T 1984 Compton scatter axial tomography with x-rays: SCAT-CAT Phys.
Med. Biol. 29 135370
Davidson M M T, Batchelar D L, Velupillai S, Denstedt J D and Cunningham I A 2005 Analysis of urinary stone
components by x-ray coherent scatter: characterizing composition beyond laboratory x-ray diffractometry Phys.
Med. Biol. 50 377386
Elshemey W M and Elsharkawy W B 2009 Monte Carlo simulation of x-ray scattering for quantitative characterization
of breast cancer Phys. Med. Biol. 54 377384
Griffiths J A, Royle G J, Hanby A M, Horrocks J A, Bohndiek S E and Speller R D 2007 Correlation of energy
dispersive diffraction signatures and microCT of small breast tissue samples with pathological analysis Phys.
Med. Biol. 52 615164
Harding G, Kosanetzky J and Neitzel U 1987 X-ray diffraction computed tomography Med. Phys. 14 51525
Hasan M Z 2003 Measurement of x-ray scattering form factors over a wide momentum transfer range Masters Thesis
Department of Physics, Carleton University, Ottawa, Ontario, Canada
Hasan M Z and Johns P C 2004 Energy-dispersive technique to measure x-ray scattering form factors over a wide
momentum transfer range Phys. Can. 60 145 (abstract)
Hubbell J H and verb I 1979 Relativistic atomic form factors and photon coherent scattering cross sections J. Phys.
Chem. Ref. Data 8 69105
Hura G, Sorenson J M, Glaeser R M and Head-Gordon T 2000 A high-quality x-ray scattering experiment on liquid
water at ambient conditions J. Chem. Phys. 113 91408
Johns P C and Schulze J S C 2009 Pinhole image characterization of x-ray tube focal spots using storage phosphor
imaging technology Med. Phys. 36 4306 (abstract)
Johns P C and Wismayer M P 2004 Measurement of coherent x-ray scatter form factors for amorphous materials
using diffractometers Phys. Med. Biol. 49 523350
King B W and Johns P C 2007 Laboratory experience with a mechanically cooled high-purity germanium x-ray
spectrometer Radiother. Oncol. 84 (Suppl. 2) S56 (abstract)
King B W and Johns P C 2008 Measurement of coherent scattering form factors using an image plate Phys. Med.
Biol. 53 597790 (corrigendum 54 6437)
King B W and Johns P C 2009 An energy-dispersive technique to measure tissue x-ray coherent scattering form
factors Med. Phys. 36 2786 (abstract)
Kosanetzky J, Knoerr B, Harding G and Neitzel U 1987 X-ray diffraction measurements of some plastic materials
and body tissues Med. Phys. 14 52632
Leclair R J, Boileau M M and Wang Y 2006 A semianalytic model to extract differential linear scattering coefficients
of breast tissue from energy dispersive x-ray diffraction measurements Med. Phys. 33 95967
Leclair R J and Johns P C 1998 A semianalytic model to investigate the potential applications of x-ray scatter imaging
Med. Phys. 25 100820

An energy-dispersive technique to measure x-ray coherent scattering form factors

871

Leclair R J and Johns P C 1999 Fundamental information content accessible with medical x-ray scatter imaging Phys.
Med. Imaging, Proc. SPIE 3659 67281
Lide D R (ed) 2009 Standard density of water CRC Handbook of Chemistry and Physics 90th edn (Boca Raton, FL:
CRC Press) section 6, p 7
Narten A H 1970 X-ray diffraction data on liquid water in the temperature range 4 C200 C Technical Report ORNL
4578 (Oak Ridge, TN: Oak Ridge National Laboratory)
Peplow D E and Verghese K 1998 Measured molecular coherent scattering form factors of animal tissues, plastics
and human breast tissue Phys. Med. Biol. 43 243152
Poletti M E, Goncalves O D, Schechter H and Mazzaro I 2002 Precise evaluation of elastic differential scattering
cross-sections and their uncertainties in x-ray scattering experiments Nucl. Instrum. Methods B 187 43746
Van Uytven E, Pistorius S and Gordon R 2007 An iterative three-dimensional electron density imaging algorithm
using uncollimated Compton scattered x rays from a polyenergetic primary pencil beam Med. Phys. 34 25665