PHYSICS IN MEDICINE AND BIOLOGY
Phys. Med. Biol. 52 (2007) 3089–3104
Development of a guarded liquid ionization chamber
for clinical dosimetry
K J Stewart1, A Elliott2and J P Seuntjens
Medical Physics Unit, Montreal General Hospital, Montreal, H3G IA4, Canada
E-mail: firstname.lastname@example.org and email@example.com
Received 11 December 2006, in final form 20 March 2007
Published 10 May 2007
Online at stacks.iop.org/PMB/52/3089
Liquid ionization chambers are considered superior to air-filled chambers in
terms of size, energy dependence and perturbation effects. We constructed and
tested a liquid ionization chamber for clinical dosimetry, the GLIC-03, with a
sensitive volume of approximately 2 mm3. We also examined two methods to
method relating recombination to dose per pulse. The second method can be
used even in cases where the first method is not applicable. The response
of the GLIC-03 showed a stable, linear and reproducible decrease of 1% over
a Clinac 21EX. The two methods for recombination correction agreed within
0.2% for measurements at 18 MV, 700 V, 100 MU min−1. Measurements
with the GLIC-03 in Solid WaterTMin the build-up region of an 18 MV beam
agreed with extrapolation chamber measurements within 1.4%, indicating that
the GLIC-03 causes minimal perturbation.
Air-filled ionization chambers are the most commonly used instruments for clinical radiation
dose measurements.Their ease of use, exceptional long-term stability and well-studied
characteristics make this the detector of choice for many dosimetry applications. However,
there are some characteristics of air-filled chambers that limit their accuracy for certain
measurements. First of all, because air is a low-density medium, the amount of ionization
chambers with very small volumes and therefore limits the size (and thereby the spatial
1Present address: Allan Blair Cancer Centre, Regina, SK, S4T 7T1, Canada.
2Present address: OnCURE Medical Corporation, Jacksonville, FL 32216, USA.
0031-9155/07/113089+16$30.00© 2007 IOP Publishing LtdPrinted in the UK3089
3090K J Stewart et al
resolution) of air-filled chambers. When measuring in a medium of higher density, such as
water, the introduction of a low-density cavity also perturbs the electron fluence within the
medium; thus perturbation corrections are required to determine the dose in the absence of the
become particularly important in regions requiring high resolution, in areas where charged
particle equilibrium is not present, and in regions where the average electron energy varies or
is unknown. Some examples are profile measurements of very small fields, measurements in
the build-up region for high-energy photon beams and measurements of IMRT fields.
certain advantages compared with air-filled chambers. The high density of liquids compared
to air results in a higher ionization density, allowing very small-volume chambers to be
constructed which still have a sufficiently large signal, thus providing high spatial resolution.
Second, when measuring in a medium such as water, liquids produce negligible perturbation
effects as their density is very similar to water. Finally, for certain insulating liquids, the mean
restricted collision mass stopping power ratio water-to-liquid shows almost no variation with
electron energy over the range of energies used clinically. Therefore a liquid-filled detector
should have very little energy dependence.
The most extensive work on the development of liquid ionization chambers for clinical
dosimetry has been done by Wickman et al (e.g., Wickman (1974), Wickman and Nystrom
(1992)) This group has constructed and tested many chamber designs and has also studied
issues related to ion recombination in liquids (e.g., Johansson et al (1997)). There have
also been studies performed with commercial air ionization chambers filled with insulating
liquids (Francescon et al 1998). We carried out some previous work with one of the Wickman
chambers (Stewart et al 2001). We also tested an Exradin A14P chamber that had been
modified to reduce the electrode separation to 0.5 mm and filled with isooctane (Stewart and
chamber. We examine the properties of this chamber and investigate methods of correcting for
ion recombination effects. Finally we perform measurements with this new liquid ionization
chamber and compare it to other detectors.
2. Materials and methods
Our previous work with liquid ionization chambers inspired us to design an original liquid
ionization chamber. Several prototypes were constructed and tested with the most suitable
for measurements being the GLIC-03. GLIC stands for guarded liquid ionization chamber,
as the unique feature of this chamber is the inclusion of a guard electrode. Generally, in
liquid ionization chambers, a guard electrode has not been included since it is not required
to prevent in-scatter perturbation effects as it would be in air-filled chambers. The purpose
behind including it in this design, however, was to allow testing of the chamber both with and
without liquid in the sensitive volume. The GLIC-03 (figure 1) was constructed with graphite
ring is 1.5 mm thick. The Teflon insulator between these electrodes has a thickness of 0.4 mm.
The outer cap and chamber body are made from Delrin in order to make the chamber liquid
tight. The end face of the cap consists of 1 mm Delrin and 0.5 mm graphite. A Buna-N o-ring
is used to seal the chamber volume. To prevent wear on the graphite outer electrode when the
cap is screwed on and off, a 1 mm thick stainless steel ring was glued below the graphite and
Development of a guarded liquid ionization chamber3091
Figure 1. Schematic diagram showing the cylindrical cross section of the GLIC-03.
a brass foil provides improved electrical connection with the outer cap. The outer diameter
of the chamber is 21 mm. The length of the chamber body is 34 mm and holes were drilled
to insert 0.8 mm diameter stainless steel filling tubes. Tygon tubing was attached to the outer
ends of the stainless steel tubes and the ends are clamped closed after filling the chamber with
The chamber was filled with 2,2,4-trimethylpentane, also called isooctane (Aldrich,
anhydrous, 99.8%), which was used without additional purification. Non-ultra purified liquids
have a limited ion mobility and therefore require the use of high electric field strengths in
order to reduce ion recombination. For isooctane with a purity of 99.5% without additional
purification, the measured ion mobility was 2.9 × 10−8m2s−1V−1for both positive and
negative ions (Johansson and Wickman 1997). We have used this value of ion mobility for our
calculations. Many other studies have been done with isooctane to determine properties such
as temperature dependence (Franco et al 2006) and ion recombination characteristics (e.g.,
Johansson et al (1997)) and the variation in stopping power ratio isooctane-to-air is very small
over the range of energies used clinically.
2.1. Measuring the air-filled chamber characteristics
Before filling the chamber with liquid, we examined its air-filled properties. For comparison,
measurements were also taken with an Exradin A14P chamber, which is similar to the GLIC-
03 in terms of design and sensitive volume. Measurements were done for 6 and 18 MV beams
3092 K J Stewart et al
Figure 2. Diagram showing the three Solid WaterTMphantoms used in this work. Phantom 1 was
used for reference dosimetry with the Exradin A12 and Exradin A14P. Phantom 2 was used for
reference dosimetry with the GLIC-03 and for PDD measurements with the GLIC-03. Phantom 3
was used for stability and ion recombination measurements with the GLIC-03.
from a Varian Clinac 21EX with a polarizing voltage of 300 V applied to the chambers. To
determine the absorbed dose to water calibration coefficient at each energy, NDw, a cross-
calibration procedure was followed. First of all, measurements were taken with an Exradin
A12 chamber (SN 310) which has a60Co absorbed dose to water calibration coefficient
established at NRC (National Research Council, Ottawa, Canada), the national standards lab.
This chamber was positioned with the centre of the chamber at 10 cm water-equivalent depth
in a 20 × 20 × 20 cm3Solid WaterTM(Gamex RMI) phantom (phantom 1 in figure 2).
The absorbed dose to water was determined using the procedure of Seuntjens et al (2005)
which describes a method to derive dose to water for a measurement at an equivalent point
in Solid WaterTM. The correction factor kphhas a value of 1.000 for 6 MV and 1.006 for
18 MV (Seuntjens et al 2005). Immediately following the measurements with the Exradin
A12, measurements were taken with one of the small volume chambers. For the Exradin
A14P, it was possible to place the chamber in the same Solid WaterTMphantom as was used
for the Exradin A12, using a Solid WaterTMsleeve to adjust the position to represent the same
point of measurement for each detector. Because of its larger diameter, another Solid WaterTM
phantom was needed for the GLIC-03 (phantom 2 in figure 2). This phantom consisted of
a 30 × 30 × 20 cm3Solid WaterTMblock with a hole of the same diameter as the GLIC-03
drilled in the centre. 30 × 30 cm2slabs of various thicknesses were added in front of the
chamber to adjust the point of measurement to 10 cm depth. For all measurements, the SSD
was 100 cm and the field size was 10 × 10 cm2at the surface of the phantom. Polarity and
recombination effects for the air-filled chambers were determined using the method described
Development of a guarded liquid ionization chamber3093
inTG-51(Almondetal1999). Whenmeasuringwiththesmall-volumechambers, theleakage
current was measured and corrected for.
2.2. Determining the electrode separation
The sensitive volume of the GLIC-03 is a cylinder with a cross-sectional area defined by the
size of the collecting electrode and a height defined by the separation between the collecting
and outer electrodes. Since the cap of the chamber is screwed on until it is tight, this electrode
separation is not rigorously defined by the chamber construction. In order to determine the
electrode separation we performed a capacitance test. In this test the un-filled chamber was
connected to a Keithley 6517A electrometer and the polarizing voltage was incremented in
steps of 50 V from 0 to 400 V. The charge was measured after each increment. The slope of
the linear fit to the charge as a function of voltage represents the capacitance of the chamber.
Assuming a perfect parallel plate capacitor, the separation between the plates can be expressed
as a function of the capacitance, C, and the plate area, A, by
where ε0is the permittivity of free space. Because the GLIC-03 has a large guard ring, we
have assumed that the sensitive volume behaves as an ideal parallel plate capacitor. The area
of the sensitive volume is not easily measured in our case, as the collecting electrode has a
small diameter and may not be exactly circular. As well, a small error in the determination of
the radius will be magnified by the fact that the radius is squared to determine the area. For
this reason we used the calibration coefficients obtained through cross-calibration to estimate
the sensitive volume. Since Vol= dA, where Volis the sensitive volume of the chamber:
The cavity dose calibration coefficient is related to the sensitive volume by
where W/e is the energy required to produce an electron pair in dry air and ρairis the density
of air. ND,airis derived from the measured absorbed dose to water calibration coefficient after
correcting for the stopping power ratio water-to-air and neglecting perturbation effects.
2.3. Stability and reproducibility of the liquid-filled chambers
The GLIC-03 stability was tested in the 18 MV beam of a Clinac 21EX. The polarizing
voltage was set to 500 V and the lowest available dose rate setting (100 MU min−1) was used.
The chamber was inserted into Solid WaterTMphantom 3 (see figure 2), which was oriented
such that the sensitive volume of the chamber was directed vertically downwards and the
phantom was positioned at 100 cm SSD. This phantom also has a hole in which the Exradin
A12 chamber could be inserted to monitor changes in beam output. Both chambers were at
approximately the same depth of 15 cm in Solid WaterTM.
To determine the necessary pre-irradiation dose to achieve a stable reading, a series of
measurements were taken to monitor the response of the chamber. Immediately following a
new fill of the chamber with liquid, the chamber was given a series of six to ten irradiations
of 250 MU until a set of measurements showed acceptable stability. Between each series of
irradiations 2500 MU was delivered. For all other measurements, this pre-irradiation dose
was delivered initially so that a stable response was achieved. The stability of the chamber
3094 K J Stewart et al
the same irradiation conditions with 100 MU delivered for each measurement. This test was
performed on three separate liquid fills for periods of 5 h, 7 h and 24 h with the Exradin A12
chamber being used to correct for any changes in beam output.
Reproducibility of the chamber response from one liquid fill to another over a period of
8 months was monitored under the same set-up and irradiation conditions described above.
Following each fill with new liquid a large pre-irradiation dose was given and the radiation
response to a 250 MU irradiation as well as the leakage current were measured. Again the
Exradin A12 chamber was used to correct for any changes in beam output over time.
2.4. Ion recombination
collected. Three main effects contribute to this loss of charge: initial recombination, general
recombinationanddiffusion. Ininitialrecombination, oppositelychargedionsproducedalong
a single ionizing particle track recombine. The rate of initial recombination depends on the
ionization density along a track, which will increase with the density of the ionized medium
and with the LET of the ionizing particle. In general recombination, ions from different tracks
will recombine. This depends on the density of ionizing particles in the medium, which is a
function of the dose rate in continuous beams or the dose per pulse in pulsed beams. Both
initial and general recombination also depend on the applied electric field and, in addition,
general recombination depends on the ion transit time. Charge loss due to diffusion against
the electric field has a low probability of occurring and is generally considered negligible. For
an air-filled chamber, the most commonly used technique for determining the correction for
ion recombination is the so-called two-voltage technique. This method is derived from Boag’s
theory of general recombination in gases (Boag 1952). For pulsed radiation beams in the
region near saturation (where initial recombination is negligible), the inverse of the collected
charge is proportional to the inverse of the applied voltage. The correction for recombination
losses can thus be calculated by
1 − (VH/VL)
(MH/ML) − (VH/VL),
where VH and VL are, respectively, the high and low applied voltages and MH and ML
are the chamber readings at these voltages. Several other assumptions are also made in
this formulation. First, any effects of space-charge screening and diffusion loss have been
neglected. It also assumes that the recombination during the radiation pulse is negligible (the
pulse must be short compared to the ion transit time) and that the densities of ions of opposite
charges are equal (all negative charge is carried by negative ions). Finally, the pulse repetition
frequency must be low enough that the charge generated by a radiation pulse is fully collected
before the next pulse occurs.
In operating an ionization chamber, not all of the ions produced will be
2.4.2 Ion recombination in liquids.
straightforward as determining it in gases. Equation (4) is not applicable as key conditions
cannot be met in liquids. Charge collection in liquids does not saturate with increasing applied
voltage. In fact, increasing the voltage increases the number of free ions available in the
volume. As well, the effects of initial recombination are large in liquids even at very high
electric field strengths, so it is not possible to operate in a region where initial recombination
can be neglected. The condition that all charges from one radiation pulse must be collected
before the next radiation pulse occurs can also be difficult to achieve in liquids, as their ion
mobilities are much lower than those in gases.
Determining ion recombination in liquids is not so
Development of a guarded liquid ionization chamber3095
Attempts have been made, however, to apply a modified form of Boag’s theory to liquids
current and the applied voltage in conditions of low dose per pulse and where the charge
generated by one pulse is completely collected before the next pulse occurs. In this case,
above a threshold voltage, the current can be observed to increase linearly with applied
voltage. In this region it is assumed that general recombination is negligible, so fitting a
straight line to this region provides an expression for the expected current in the absence of
recombination (labelled itheor):
itheor= (c1+ c2E)
where E is the electric field strength,˙D is the dose per pulse and c1and c2are constants
obtained from the linear fit to the measured charge as a function of applied voltage. The
theoretical general collection efficiency, ftheor, can be determined by making use of Boag’s
with u being defined as
where ε is the relative permittivity of the liquid, ε0is the permittivity of free space, rchamberis
the radius of the sensitive volume of the chamber, and ν is the pulse repetition frequency.
There are, however, also limitations to the applicability of this method. First of all, more
detailed study (Pardo et al 2004) has revealed that the collected charge does not continue
to increase linearly with increasing electric field. This can be understood from Onsager’s
theory of electrolytic dissociation in external electric fields (Onsager 1934). At higher electric
field strengths, higher order terms in the equation relating initial recombination to applied
electric field become significant, so the response is no longer linear. For isooctane, this
departure from linearity occurs at an electric field strength of about 1.25 × 106V m−1(Pardo
et al 2004). For a chamber with an electrode separation of 0.56 mm, as is the case for the
GLIC-03, this corresponds to an applied voltage of 700 V. A confounding limitation is the
requirement that all charge produced in a pulse be collected before the next pulse occurs. This
condition sets a limit on the lowest voltage which can be applied for a given pulse repetition
frequency. For a Varian Clinac 21EX, the lowest available pulse repetition frequency is at the
100 MU min−1setting. For 18 MV the lowest frequency is 30 Hz and for 6 MV it is 60 Hz.
From this, considering a 0.56 mm electrode separation, based on the ion mobility of isooctane,
the minimum voltage required is 325 V for 18 MV and 650 V for 6 MV.
Note that for the 6 MV beam we have only a very small range of applicable voltages
(650–700 V). This fact makes the evaluation of recombination based on the method described
by Johansson et al (1997) impossible for our chamber in this beam. There is, however, another
possible approach. General recombination also varies as a function of dose rate or dose per
pulse, while initial recombination is not affected by changes in dose per pulse. If the voltage
and pulse repetition frequency are kept constant, an empirical relationship between collected
charge and dose per pulse can be determined. Although this approach will not provide an
absolute value of the ion recombination, it can be used for relative measurements to correct
for different levels of recombination present at different dose rates. This method should be
applicable for comparing beams of different pulse frequencies so long as there is complete
charge collection from one pulse before the arrival of the next pulse. In a single beam it can
also be applied for various levels of dose per pulse even at a high pulse frequency where there
is overlap of charge from successive pulses.
uln(1 + u),
3096 K J Stewart et al
Table 1. Beams, voltages, dose rate settings and corresponding pulse repetition frequencies used
in the evaluation of the relative efficiency as a function of dose per pulse for the GLIC-03. Note
that the 500 Mu min−1dose rate on the Varian Clianc 21EX is obtained from a series of pulses at
180 Hz with one pulse in every six being turned off.
Dose rate setting
We performed tests with the GLIC-03 to evaluate the recombination using both of
the methods described above. For the method of Johansson et al (1997) we used the
100 MU min−1setting with the 18 MV beam and performed measurements with the GLIC-03
and Exradin A12 chambers inserted into phantom 3 (figure 2). Measurements were performed
at different SSDs to achieve different values of dose per pulse, which were determined based
on the Exradin A12 readings. The applied voltage was varied between 100 and 700 V for all
SSDs and the measurements with applied voltages between 500 and 700 V at the farthest SSD
were used to determine the constants c1and c2. Equations (5)–(7) were then used to determine
For the second method, the GLIC and Exradin A12 chambers were again inserted in
phantom 3 (figure 2) and the SSD was changed in order to vary the dose per pulse. The
variation of response as a function of dose per pulse was used to obtain a relative efficiency.
The beams, voltages, dose rate settings and associated pulse repetition frequencies tested with
this method are listed in table 1.
2.5. Cross-calibration of the liquid-filled GLIC-03
In order to evaluate the energy dependence of the GLIC-03, the chamber was cross-calibrated
in the 6 and 18 MV beams of a Clinac 21EX using the 100 MU min−1dose rate setting. The
absorbed dose to water in Solid WaterTMwas measured using the Exradin A12 chamber in
the same manner as described in section 2.1. The response of the GLIC-03 was measured in
Solid WaterTMphantom 2 (figure 2) for the same irradiation conditions (10 × 10 cm2field,
100 cm SSD, 10 cm depth) as were used for the air-filled calibration. The voltage applied to
Corrections for relative differences in ion recombination between 6 and 18 MV were made
using the curves of response as a function of dose per pulse for this voltage and these beams.
2.6. PDD measurements in Solid WaterTM
The PDD in Solid WaterTMwas measured for the build-up region of an 18 MV beam in using
the GLIC-03 both with and without liquid in Solid WaterTMphantom 3 (figure 2). The front
face of the chamber was flush with the surface of the 30 × 30 × 20 cm3block and then
various thicknesses were added in front of the detector to vary the depth. The phantom was
placed on a translatable stage and was moved as each thickness was added so that the phantom
surface remained at 100 cm SSD. The accuracy of the positioning is estimated to be ±0.2 mm.
The Solid WaterTMslabs used in front of the chamber have certification documents which
list their thicknesses to the nearest 0.01 mm. The dose rate setting was 500 MU min−1and
Development of a guarded liquid ionization chamber3097
Table 2. Properties of the air-filled GLIC-03 and Exradin A14P chambers for an applied voltage
of 300 V.
Sensitive volume thickness (mm)0.56
NDw(cGy nC−1) 18 MV
Ratio of (NDw18 MV)/(NDw6 MV)
the field size was 10 × 10 cm2. Measurements were taken at both + and −300 V with the
air-filled GLIC-03 and averaged to correct for polarity effects. For the liquid-filled GLIC-03,
the voltage applied was 500 V and relative corrections for recombination as a function of dose
rate were determined from the 500 V, 500 MU min−1, 18 MV curve. Depths in Solid WaterTM
were scaled by density to represent the equivalent depth in water.
3. Results and discussion
3.1. Air-filled properties of the GLIC-03
The air-filled characteristics of the GLIC-03 chambers are listed in table 2.
characteristics of a commercial chamber of similar dimensions, the Exradin A14P, are listed
for comparison. Reproducibility on each of these values was within ±0.25%. The leakage
current was always less than 0.3% of the measured signal for the Exradin A14P chamber. For
the GLIC-03, theleakage signalwas larger, but stableand lessthan 4% of themeasured signal.
3.2. Electrode separation
determined from the cross-calibration to find the plate separation according to equation (2).
The electrode separation of the GLIC-03 is 0.56 mm. This value is relatively insensitive to
uncertainties in the volume determination. A 10% error in the volume results in an error in
the separation of only 0.03 mm. In contrast, if the physical radius is used to determine the
thickness, then a 0.1 mm error in the radius will result in an error in electrode separation of
0.15 mm. For the GLIC-03, based on the volume determined using the calibration coefficient
and the capacitance measurements, the sensitive volume has a radius of 1.1 ± 0.05 mm. This
is much larger than the nominal physical radius of 0.75 mm.
3.3. Stability and reproducibility
With the GLIC-03, we were able to achieve a stable response following a large pre-irradiation
dose. Figure 3 shows that, although changes in chamber response of over 30% could be
observed initially over the delivery of the first 14500 MU (sets (a)–(c)), the measurements
that followed (sets (d) and (e)) had a variation of less than 5% while an additional 5750 MU
was delivered and the final nine measurements (set (e)) have a variation of less than 1%. The
large initial change in response indicates that there are some impurities present in the liquid,
however, with increased radiation dose a relatively steady state is reached.
3098K J Stewart et al
Figure 3. Relative response of the GLIC-03 over the initial 41 measurements after a new liquid
fill. The five different sections (a)–(e) represent series of readings for an irradiation of 250 MU.
Before each series 2000 MU was delivered.
Figure 4. Response of the GLIC-03 for three different liquid fills tested over periods of 5, 7 and
of 100 MU with error bars indicating one standard deviation. The straight line is a linear fit to all
of the data points.
When monitoring the response over time following a pre-irradiation dose, it can be seen
from figure 4 that, although there is a decrease in response of more than 2% over 24 h, this
decrease isstableandreproducible, soitwouldbepossibletocorrectthereadings accordingly.
Additionally, the change is only 1% over 10 h, so the correction would be insignificant for
measurements over a short time interval.
We examined the reproducibility of the chamber response each time the chamber is filled
with new liquid. Figure 5 indicates that this is never more than 5% different from the mean.
The leakage current is also consistently less than 1% of the signal measured with a dose rate of
approximately 63 cGy min−1. The collection of data examining the stability as a function of
Development of a guarded liquid ionization chamber3099
Figure 5. Relative response of the GLIC-03 after each new liquid fill. Filled symbols indicate the
relative response and correspond to the values on the left axis. The dose rate was approximately
63 cGy min−1and each point represents the average of five measurements with one standard
deviation indicated by the error bars. Open symbols indicate the leakage current expressed as a
percentage of the chamber response and corresponding to the values on the right axis.
fill provides a good indication whether or not drastic changes in chamber or liquid properties
have occurred, for example, high levels of impurities in the liquid.
3.4. Ion recombination
Figure 6 shows the theoretical general collection efficiency, ftheor, as a function of electric
field strength for the GLIC-03 calculated according to equations (5)–(7) for different values
of dose per pulse. For the lowest dose per pulse (0.05 mGy pulse−1), ftheoris greater than
0.997 at all field strengths. Note that for the highest dose per pulse value shown here
(0.36 mGy pulse−1) the collection efficiency is lower than 0.986 even at the highest field
strength, corresponding to 700 V. This decrease in collection efficiency is important, as many
typical clinical measurements are done with a dose per pulse as high as 0.6 mGy pulse−1,
where ftheorwould be reduced to below 0.978 according to this method of calculation.
As was mentioned previously, there are cases where the method of Johansson et al (1997)
is not applicable to measurements with the GLIC-03. In particular, measurements in the 6 MV
beam or other measurements where we would like to use a higher pulse repetition frequency.
In order to correct for ion recombination in these cases, we used the second method described
in section 2.4.2, where the relative efficiency is determined as a function of dose per pulse
and normalized at an arbitrary value of dose per pulse. The lines in figure 7 show this relative
efficiency for the four cases listed in table 1 normalized to unity at a dose per pulse of 0.1 mGy
pulse−1. This is a linear fit to the data measured when varying the dose per pulse by changing
the SSD of the Solid WaterTMphantom. The symbols in figure 7 indicate the theoretical
general collection efficiency calculated using equations (5)–(7) for the two cases where this
100 MU min−1), also normalized to unity at 0.1 mGy min−1.
For case II (18 MV, 700 V, 100 MU min−1), ftheoragrees with the measured relative
efficiency within 0.2%; however, the difference is larger (up to 0.6%) for case III (18 MV,
3100K J Stewart et al
equations (5)–(7) as a function of electric field strength. The range of electric field strengths
corresponds to a range of polarizing voltages between 450 V (0.80 × 106V m−1) and 700 V
(1.25 × 106V m−1). Values are shown for different dose per pulse rates: (from top to bottom)
0.05 mGy pulse−1, 0.07 mGy pulse−1, 0.10 mGy pulse−1, 0.18 mGy pulse−1and
0.36 mGy pulse−1.
Theoretical general collection efficiency of the GLIC-03 calculated according to
Figure 7. Relative efficiency for the GLIC-03 as a function of dose per pulse. Lines indicate the
relative efficiency based on a linear fit to measurements done with different dose per pulse values
for the conditions given in table 1: case I, 6 MV, 700 V, 100 MU min−1(solid line); case II, 18 MV,
700 V, 100 MU min−1(dashed line); case III, 18 MV, 500 V, 100 MU min−1(dotted line) and case
IV, 18 MV, 500 V, 500 MU min−1(dot-dashed line). Symbols indicate the theoretical efficiency,
ftheor, calculated using equations (5)–(7) and normalized to 1 at a dose rate of 0.1 mGy pulse−1for
case II, 18 MV, 700 V, 100 MU min−1(open symbols) and case IV, 18 MV, 500 V, 100 MU min−1
500 V, 100 MU min−1). Because there is incomplete charge collection from one pulse before
the next pulse occurs, the efficiency as a function of dose per pulse is much lower for case IV
(18 MV, 500 V, 500 MU min−1), and the theoretical efficiency calculation cannot be applied.
In the case of the 6 MV beam (case I), since the pulse frequency is double that at 18 MV, it
was not possible to apply a large enough range of electric field strengths to use the method of
Johansson et al (1997) to determine c1and c2, so values of ftheorcould not be calculated.
Development of a guarded liquid ionization chamber3101
Figure 8. Per cent difference between the 18 MV PDD measured with the liquid-filled GLIC-03
and other detectors: IC-10 (thin solid line) (Abdel-Rahman et al 2005), Roos chamber (dotted
line) (Abdel-Rahman et al 2005), PEEC extrapolation chamber (dot-dash line) (Abdel-Rahman
et al 2005), and air-filled GLIC-03 (thick solid line).
3.5. Energy dependence
For the 6 and 18 MV beams with 700 V and 100 MU min−1, the relative efficiencies agree
within 0.3%, indicating that there is not a substantial difference between the recombination
behaviour for these two beams although they have very different pulse frequencies. We
therefore considered it valid to apply corrections for the relative efficiency for each beam
based on the linear fits shown in figure 7. Since the voltage applied to the GLIC-03 was 700 V
used were 1.0106 for 18 MV (0.44 mGy pulse−1) and 1.0010 for 6 MV (0.19 mGy pulse−1),
correcting to 0.1 mGy pulse−1. Since, in this work, we looked only at the relative energy
response between 6 and 18 MV, what we are interested in is the ratio of Pioncorrections, which
is 1.0095 for 18 MV/6 MV, so the point of normalization is not important. If we were to
calculate the correction ratio for these two dose rates based on the ftheorvalues for 18 MV,
700 V, 100 MU min−1, we would obtain 1.0101 for 18 MV/6 MV, a difference of only 0.06%.
Using the Pioncorrections and the data obtained from cross calibration against the Exradin
A12 chamber, the ratio of the absorbed dose calibration factors (NDw18 MV)/(NDw6 MV)
is 0.989 ± 0.004. This is 1% closer to unity than the same ratio calculated for the air-filled
GLIC-03 of 0.979 ± 0.003, indicating that there is less energy dependence for the liquid-filled
3.6. Build-up region measurements
03 and the air-filled GLIC-03 measurements from this study. The readings of the liquid-filled
GLIC-03 were corrected for relative differences in recombination using the 18 MV, 500 V,
500 MU min−1linear fit (case IV) from figure 7. Note that the range of dose rates is from 0.25
to 0.56 mGy pulse−1, so the relative difference in Pioncorrections used is 3.7%. As well, the
liquid-filled GLIC-03 is compared with PDD measurements in Solid WaterTMfrom Abdel-
Rahman et al (2005). These measurements were taken with the IC-10 cylindrical chamber, the
3102K J Stewart et al
Roos plane-parallel chamber and the PEEC extrapolation chamber. All measurements have
been scaled to an equivalent depth in water.
The liquid- and air-filled GLIC-03 measurements agree within 0.5% over the entire
measured PDD. In comparison to the measurements from Abdel-Rahman et al (2005), the
largest difference between the liquid-filled GLIC-03 and the PEEC extrapolation chamber is
1.4% and beyond a depth of 6 mm, the agreement is better than 0.7%. Similarly for the Roos
chamber, the largest difference is 2.5% near the surface and agreement is better than 0.5%
beyond 4 mm depth. The situation is different for the IC-10 chamber, which has a response
to less than 1% beyond 22 mm depth.
The good agreement between both the liquid- and air-filled GLIC-03 measurements and
those taken with the PEEC extrapolation chamber indicates that the GLIC-03 chamber has a
verysmallperturbationeffect, whetherornotitisfilledwithliquid. Sincethechambervolume
is so tiny and the guard ring is large compared to the radius of the collecting electrode, this is
to be expected. The IC-10 chamber, on the other hand, causes significant perturbation to the
radiation field and therefore shows a large over-response at shallow depths. The fact that the
liquid- and air-filled GLIC-03 measurements agree within 0.5% indicates that our method of
correcting the liquid-filled chamber response for general recombination effects is appropriate
even though the relative corrections for general recombination are large (up to 3.7%).
The purpose of this work was to construct and test the properties of a liquid ionization
chamber. A re-fillable detector with a guard electrode was constructed, the GLIC-03. Using a
well-guarded chamber is advantageous as the properties can be tested first without filling the
chamber with liquid. The air-filled characteristics of the GLIC-03 are equivalent or superior
to a similar commercial air-filled chamber, the Exradin A14P. Impurities in the liquid can
affect the chamber response, stability and leakage current. Monitoring the chamber response
and leakage current each time it is filled with liquid provided a way to detect changes in
liquid purity or chamber behaviour. The response of the GLIC-03 decreased by 1% over 10 h;
however, this decrease in response was stable, linear and reproducible. Further investigation
of impurities in the liquid should be done to see if the stability of the chamber response can
be improved, as other liquid ionization chambers have shown superior long-term stability
(Wickman and Nystrom 1992).
One of the significant issues related to measurements with liquid ionization chambers is
correction for ion recombination, since this is much larger in liquids than it is in gases. We
first used the method described by Johansson et al (1997) which modifies Boag’s theory of
general recombination in gases and applies it to liquids. Because this theory has limitations
related to the maximum pulse repetition frequency and the nonlinear behaviour of Onsager’s
theory of initial recombination at high electric field strengths, we proposed a second method
for general recombination corrections in cases where the method of Johansson et al was not
applicable. This second method relates the general collection efficiency to the dose per pulse
within 0.2% for the 18 MV, 700 V, 100 Mu min−1case and within 0.6% for the 18 MV, 500 V,
100 Mu min−1case.
One advantage expected with liquid ionization chambers is a lower energy dependence,
since the ratio of the mean restricted collision mass stopping power water-to-isooctane varies
range, the stopping power ratio water-to-air varies by 16%. We found that the liquid-filled
Development of a guarded liquid ionization chamber3103
GLIC-03 had an energy dependence of 1.1 ± 0.4% while the air-filled GLIC-03 had a 2.1 ±
0.3% energy dependence when comparing the response between the 6 and 18 MV beams from
a Clinac 21EX.
We also compared PDD measurements of the build-up region of the 18 MV beam from
a Clinac 21EX. This served two purposes. First of all, since we used a high pulse repetition
frequency where the method of Johansson et al (1997) was not applicable, we were able to
test whether our method of correcting for relative differences in general recombination based
on response as a function of dose per pulse was valid. The measurements were taken with
values of dose per pulse ranging from 0.25–0.56 mGy pulse−1, and the general recombination
correction varied by 3.7% over this range of dose per pulse. The second purpose of these
measurements was to examine the amount of perturbation caused by the liquid-filled or air-
filled GLIC-03. We compared our measurements to measurements taken with the PEEC
extrapolation chamber, which should produce negligible perturbation as well as the IC-10
chamber which is known to have a large perturbation effect at depths near the surface.
The excellent agreement between the liquid- and air-filled GLIC-03 measurements and
the measurements with the PEEC extrapolation chamber indicates that, whether this chamber
isfilledwithairorliquid, itproducesnegligibleperturbation. Aswell, thisagreementprovides
confirmation of the validity of correcting for relative general recombination as a function of
dose per pulse. This method could then be used for other relative measurements in cases
where the method of Johansson et al (1997) is not applicable.
electronics. We would also like to thank W Abdel-Rahman for sharing his PDD data with us
and C Ross and N Klassen for their helpful ideas. Funding has been provided by the Canadian
Institutes of Health Research through a Doctoral Research Award and the Natural Science
and Engineering Research Council of Canada through grant RGPIN 298191. J Seuntjens is a
research scientist of the National Cancer Institute of Canada appointed with funds provided
by the Canadian Cancer Society.
Abdel-Rahman W, Seuntjens J P, Verhaegen F, Deblois F and Podgorsak E B 2005 Validation of Monte Carlo
calculated surface doses for megavoltage photon beams Med. Phys. 32 286–98
Almond P R, Biggs P J, Coursey B M, Hanson W F, Huq M S, Nath R and Rogers D W O 1999 AAPM’s TG-51
protocol for clinical reference dosimetry of high-energy photon and electron beams Med. Phys. 26 1847–70
Boag J W 1952 The saturation curve for ionization measurements in pulsed radiation beams Br. J. Radiol. 25 649–50
Francescon P, Cora S, Cavedon C, Scalchi P, Reccanello S and Colombo F 1998 Use of a new type of radiochromic
film, a new parallel-plate micro-chamber, MOSFETs, and TLD 800 microcubes in the dosimetry of small beams
Med. Phys. 25 503–11
Franco L, Gomez F, Iglesias A, Pardo J, Pazos A, Pena J and Zapata M 2006 Liquid-filled ionization chamber
temperature dependence Nucl. Instrum. Methods Phys. Res. A 560 584–8
Johansson B E and Wickman G 1997 General collection efficiency for liquid isooctane and tetramethylsilane used as
sensitive media in a parallel-plate ionization chamber Phys. Med. Biol. 42 133–45
Johansson B E, Wickman G and Bahar-Gogani J 1997 General collection efficiency for liquid isooctane and
tetramethylsilane in pulsed radiation Phys. Med. Biol. 42 1929–38
Onsager L 1934 Deviations from Ohm’s law in weak electrolytes J. Chem. Phys. 2 599–615
Pardo J et al 2004 Free ion yield observed in liquid isooctane irradiated by gamma rays. Comparison with the Onsager
theory Phys. Med. Biol. 49 1905–14