Radiation Physics and Chemistry 81 (2012) 599–608

Contents lists available at SciVerse ScienceDirect

Radiation Physics and Chemistry

0969-80

doi:10.1

n Corr

E-m

journal homepage: www.elsevier.com/locate/radphyschem

Characteristics of radiation induced light in optical fibres for portalimaging application

I. Silva a, G. Pang b,n

a Department of Medical Biophysics, University of Toronto, 2075 Bayview Avenue, Toronto, Canada M4N 3M5b Odette Cancer Centre, Departments of Radiation Oncology and Medical Biophysics, University of Toronto, 2075 Bayview Avenue, Toronto, Canada M4N 3M5

a r t i c l e i n f o

Article history:

Received 28 February 2011

Accepted 23 January 2012Available online 8 February 2012

Keywords:

Cherenkov radiation

Portal imaging

6X/$ - see front matter & 2012 Elsevier Ltd. A

016/j.radphyschem.2012.01.044

esponding author.

ail address: geordi.pang@sunnybrook.ca (G. P

a b s t r a c t

The purpose of this paper is to characterize the radiation induced light in optical fibres to optimise the

design of a new Cherenkov detector for portal imaging application in radiation therapy. Experiments

were performed using a single optical fibre to evaluate the angle dependence, spectrum and temporal

properties of the radiation induced light in the optical fibre in comparison with that of Cherenkov

radiation. A theoretical model was also developed to compare with experiments. It has been found that

radiation-induced light output from the optical fibre under megavoltage (MV) x-ray irradiation is

significantly (about 45 times) higher than that under 100 kVp x-ray irradiation for the same dose rate at

the fibre. The angular-dependence, spectrum and temporal properties of the radiation induced light in

the optical fibre under MV x-ray irradiation match that of Cherenkov radiation. Different angular

dependence and spectrum results from that of previous studies on radiation induced light in optical

fibres have also been found. The result of the theoretical model agrees with the angle-dependence

measurements.

& 2012 Elsevier Ltd. All rights reserved.

1. Background

Most electronic portal imaging devices (EPIDs) in radiotherapyuse a thin Cu plate/phosphor screen to convert x-ray energiesinto light photons while maintaining a high spatial resolution(Antonuk, 2002). This results in low x-ray absorption (i.e., lowquantum efficiency (QE)) of approximately 2–4% for high energymega-voltage (MV) x rays (Antonuk, 2002; Pang and Rowlands,2004). A considerable increase in QE for MV x rays is needed forapplications, such as MV fluoroscopy and MV cone-beam com-puted tomography (MV-CBCT) (Pang and Rowlands, 2004; Seppiet al., 2003; Samant and Gopal, 2006a). Furthermore, current EPIDsystems use high atomic number (high-Z) materials which havean undesirable over-response to low energy x rays when used fordosimetric verification (Pistorius and McCurdy, 2002; Warkentinet al., 2003). Thus, there is a need to develop new EPIDs that havea high QE and use low-Z materials to overcome the obstaclesfaced by current x-ray imaging technologies (Pang and Rowlands,2004; Samant and Gopal, 2006a, 2006b; Sawant et al., 2006).A novel detector design was recently proposed (Mei et al., 2006)for a thick high QE detector using a matrix of optical fibresfocused towards an x-ray source, as shown in Fig. 1. This detector

ll rights reserved.

ang).

uses radiation induced light in the optical fibres in the form ofCherenkov radiation as the primary imaging signal (Mei et al., 2006).When MV x rays interact with the optical fibre array, Comptonscattering and pair-production processes will produce energeticelectrons. Those with sufficient energies in the optical fibres areexpected to produce Cherenkov light along their tracks. The lightphotons produced in the fibre core and emitted within the accep-tance angle of the fibre are guided by total internal reflectiontowards the optically sensitive amorphous silicon (a-Si) activematrix flat panel imager (AMFPI) for image readout. The activematrix is made optically sensitive either with an a-Si PIN photodiodeat every pixel or a continuous layer of amorphous–selenium (a-Se).The gap between fibres can be filled with, for example, a blackepoxy resin to absorb light which is not guided by the fibres butrather escapes from the sides of the fibres. Preliminary findings (Meiet al., 2006) have shown that such a detector can yield a zero-frequency detective quantum efficiency (DQE) an order of magni-tude higher than current low QE EPIDs, while maintaining a spatialresolution comparable to video-based EPIDs (Herman et al., 2001).

An important issue with the design of a Cherenkov detector forportal imaging applications is whether the radiation induced lightin the optical fibres is predominantly Cherenkov radiation. Thisis due to the fact that: (1) there could be other light sources,e.g., fluorescence, in optical fibres; and (2) the design dependsstrongly on the type or the properties of the light sources in theoptical fibres. The purpose of the present work is to characterize

Fig. 1. A cross section of the Cherenkov detector proposed in Mei et al. (2006)

with optical fibres focused towards an MV x-ray source. Incident MV x rays

interact with the detector and eject Compton electrons within the optical fibre,

and these high speed electrons result in Cherenkov radiation. This light is

transmitted down the fibre core towards a two-dimensional active matrix flat

panel imager (AMFPI) for image readout.

I. Silva, G. Pang / Radiation Physics and Chemistry 81 (2012) 599–608600

the radiation induced light in optical fibres to optimise the designshown in Fig. 1 and demonstrate that the predominant light sourcein the optical fibre under MV irradiation is indeed Cherenkovradiation. Although radiation induced light in optical fibres waspreviously investigated (angular dependence light-output curvesand radiation induced light-output spectra were measured withradiation therapy beams for some optical fibres) (Beddar et al.,1992; de Boer et al., 1993; Law et al., 2007; Lambert et al., 2009), allprevious studies for radiation therapy applications focused oneliminating the effect of Cherenkov light in, for example, scintillatordosimeters coupled with optical fibres (Letourneau et al., 1999).In contrast, here we want to maximise the use of Cherenkovradiation in optical fibres. In this work, experiments have beenperformed with the optical fibre used in the Cherenkov prototype inMei et al. (2006). These experiments were used to evaluate both thecontribution of Cherenkov radiation to the total light output, and theangle dependence, spectrum and temporal properties of the radia-tion induced light in the optical fibre: all of which are crucial for thedesign of the proposed detector. Discussion on the use of the resultsof the present study to optimise the design of a high QE Cherenkovdetector for portal imaging applications, as well as preliminaryresults on a prototype array detector are also included.

2. Characteristics of Cherenkov radiation

Cherenkov radiation (or Cherenkov light), discovered by PavelCherenkov, is an electromagnetic shock wave of light resulting froma charged particle (e.g., an electron) moving through a dielectricmedium at a velocity greater than the speed of light in the dielectricmedium (Zrelov, 1970; Jelley, 1958). The speed of light in thedielectric medium is c/n, where c is the speed of light in a vacuumand n is the refractive index of the medium, such that the thresholdcondition for Cherenkov radiation to be emitted is given by

nb41 ð1Þ

where b is the ratio of the particle velocity to the speed of light ina vacuum (v/c). The kinetic energy Ek of the charged particle isgiven by

Ek ¼m0c2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1�ðn=cÞ2q �m0c2 ð2Þ

where m0 is the rest mass of the charged particle. Thus, based onEqs. (1) and (2) we can determine the threshold energies requiredto produce Cherenkov radiation in a known medium. For silica(n¼1.46), the threshold energy for electrons is 191 keV.

Several unique properties of Cherenkov radiation separate itfrom other sources of light, namely: (1) The spectrum of Cherenkovradiation is broad and continuous: (Zrelov, 1970; Jelley, 1958) itspans the ultraviolet to infrared spectral regions. The relativenumber of light photons emitted for a given optical wavelengthl varies according to l�2 (the total energy of light photons for agiven l is proportional to l�3). Thus, its highest intensity is in theblue–violet portion of the spectrum; (2) The radiation emitted has astrong angle dependence: it is only emitted on the surface of a coneat an angle yC with respect to the charged particle’s trajectory (seeFig. A1 below with y¼yC), where yC satisfies the equation:

cosyC ¼1

nbð3Þ

This is in contrast to other sources of light, such as fluorescence,whose light output is expected to be uniform in all directions(Zrelov, 1970; Prasad, 2004). (3) The radiation is created instanta-neously with a very short time delay of about 10�11 s betweenthe onset of the interaction between the charged particle and thedielectric medium and the creation of Cherenkov radiation. (4) Theradiation has a strong dependence on the charged particle’s energy:no Cherenkov radiation will be generated below the thresholdenergy and (5) when the charged particle’s energy is above thethreshold energy, the total number of light photons generated isproportional to the product of the charged particle’s path length inthe medium and sin2 yC (Zrelov, 1970; Jelley, 1958).

In the following we will discuss the experiments to character-ize the radiation induced light in optical fibres in comparison withCherenkov radiation.

3. Materials and methods

3.1. Magnitude of radiation induced light in optical fibres

The purpose of this experiment was to determine the magni-tude of light output created in optical fibres under high energyirradiation (MV energy) as compared to that created under lowenergy irradiation. In all experiments, except for the temporalproperty measurement, the high energy irradiation refers to theirradiation with either a 6 MeV or 15 MeV electron beam gener-ated from a linear accelerator (LINAC). An electron beam was usedinstead of a photon beam to simplify analysis and interpretationof the results while characterizing the radiation induced light.This is mainly due to the fact that it is the electrons that generatethe Cherenkov radiation. Thus, using an electron beam avoids theadditional stage of photons creating recoil electrons in the opticalfibre. However, in clinical applications, a MV photon beam isusually used for imaging. The feasibility of using the proposedCherenkov detector for portal imaging, as well as the angulardependence of radiation-induced light in a single pixel prototypeirradiated with a 6 MV beam, have been investigated in Mei et al.(2006) and will be further discussed in Section 5.

The low energy irradiation refers to irradiation with a 100 kVpx-ray beam (since no kV electron source is available) with x-rayenergies intentionally chosen below the threshold energy forCherenkov radiation. Given the strong energy dependence ofCherenkov radiation, it is expected that different magnitudes oflight output should be produced between the high energy (i.e,above the threshold energy for Cherenkov radiation) and lowenergy (below the threshold energy) cases.

The optical fibre used in all experiments was a 10 m long, singlefused silica optical fibre (JTFSH, Polymicro Technologies) with a corediameter of 600 mm and coated with a polymer cladding layer to anouter diameter of 630 mm. The fibre core density is 2.2 g/cm3, witha core refractive index of 1.46 and cladding refractive index of 1.41

Fig. 3. Experimental setup for angular dependence of light output measurements

using high energy electrons. An optical fibre was placed on a rotating stage and

irradiated with an electron beam collimated by an electron applicator. The light

output at various beam-incident angles between 01 and 1801 was measured with a

PMT.

I. Silva, G. Pang / Radiation Physics and Chemistry 81 (2012) 599–608 601

(at wavelength of 550 nm). The fibre is also enclosed in a 3 mmpolymer jacket for mechanical protection.

The optical fibre was irradiated with a LINAC (Primus, Sie-mens) for high energy irradiation and an orthovoltage x-raymachine (D3000, Gulmay Medical) for low energy irradiation.A 10 cm segment of the optical fibre was placed in the radiationfield (perpendicular to the incident beam) between blocks of solidwater, as seen in Fig. 2. The thickness X of the solid water abovethe fibre was set at 2.5 cm for 6 MeV, 6.5 cm for 15 MeV, and 0 cmfor the low energy case. Except for the low energy case, thethickness of X was chosen to be beyond the dmax (i.e., the depth ofthe maximum dose) so that the incident electrons were suffi-ciently scattered at the location of the fibre. Thus, the measuredlight output from the fibre was the mean light output averagedover all incident angles of electrons at the fibre. We note that thedose rate at the fibre may vary for different values of X and acorrection for the dose rate variation was made in the final result(see below). The field size was set at 10 cm�10 cm (with anelectron applicator) for the MV case and 10 cm in diameter for thekV case. The centre of the top surface of the solid water blockswas at 100 cm and 30 cm away from the radiation source for theMV and kV case, respectively. In all cases, the length of irradiatedportion of the fibre was the same (i.e., 10 cm).

The magnitude of the light output from the fibre was mea-sured with a PMT (photomultiplier tube, type R6094, Hama-matsu) that was placed outside the treatment (or radiation)room to minimise the signal from direct interaction betweenthe radiation and PMT. The supply voltage to the PMT was set at850 V (with a nominal PMT gain of 6�105). The measured PMTvalues (in current) were normalised by the mean dose rate alongthe length of the fibre in the beam, i.e., the measured output wasdivided by the mean dose rate at the fibre in order to compare thelight outputs obtained with different dose rates at the fibre. Themean dose rate along the length of the fibre in the beam wascalculated using the dose rate at the reference depth (300 cGy/min for MV electron beams and 280 cGy/min for 100 kVp x-raybeam under the calibration conditions), the percentage depthdose (PDD—the ratio of the dose at a given depth on the centralaxis to that at the reference depth), and the beam profile acrossthe fibre (the ratio of the dose off-axis to that at the central axisfor the given depth). The PDD values at the location of the fibrewere 38%, 35% and 100% for 6 MeV, 15 MeV electrons and100 kVp x rays, respectively.

An important difference between the high energy and lowenergy cases is that the high energy electron beam from theLINAC was pulsed, while the x rays from the kV source were not.As a result, the light output from the fibre irradiated with highenergy electrons was pulsed in time, while the light output for thekV case was essentially constant over time. Thus, in order tocompare the magnitudes of light output between the MV and kVcases, the measured pulsed light output for the MV case was

Fig. 2. Experimental setup for magnitude of light output measurement using low

energy x rays and high energy electrons. In each case the optical fibre was placed

between blocks of solid water. The total light output was measured with a PMT

(photomultiplier tube) and normalised by the mean dose rate at the optical fibre.

integrated over the duration of a pulse (about 3.6 ms) and the totalarea (signal� time) was divided over the time between pulses(about 6.2 ms), to yield an equivalent kV light output that is‘constant’ over time. To account for potential pulse to pulse varia-tions in the output of LINACs, an average of light output over 128pulses was obtained.

3.2. Angular dependence of radiation induced light in optical fibres

In this experimental setup, shown in Fig. 3, the optical fibre(attached to a foam block for elevation) was placed on a rotatingstage and was irradiated at angles between 01 and 1801 with a15 MeV electron beam. The foam block raises the optical fibreaway from the stage, minimising the effect of scattered radiationfrom the stage onto the optical fibre. Fig. 3 shows the fibre angledparallel and pointed towards the electron beam source, repre-senting the 01 angle. As the stage rotated counter clockwise, theangle of irradiation increased from 01 to 1801, finishing with thefibre parallel and directed away from the electron beam source.The beam was collimated by an electron applicator and additionallead blocks to approximately 2 cm at the optical fibre, and thelight output of the fibre was measured twice at each anglebetween 01 and 1801: (a) with the fibre 1 cm past the centre ofthe stage, and (b) with the fibre pulled back 2 cm from theoriginal placement (dotted outline), as shown in Fig. 4. Subtract-ing (b) from (a) yielded light output from a 2 cm segment of theoptical fibre (Section (c) in Fig. 4) whose volume irradiated wasconstant at every beam angle. We note that the method used herefor the angular dependence measurement is different from thatdetailed in Beddar et al. (1992), Law et al. (2007) and Lambertet al. (2009), where no second set of measurements with the fibrepulled back was performed. We found that in our case the second

Fig. 4. Method of keeping irradiated volume constant during angular dependence

measurement. (a) The optical fibre was placed on a foam block and centred on the

rotating stage with the tip of the fibre 1 cm past the centre of the stage. (b) The

fibre was pulled back 2 cm such that the tip of the fibre was 1 cm behind the

centre. (c) Measurement b was subtracted from a to yield the light output only

from the 2 cm segment of the fibre centred on the rotating stage at any angle of

rotation (i.e., at any incident beam angle).

I. Silva, G. Pang / Radiation Physics and Chemistry 81 (2012) 599–608602

set of measurements was necessary in order to correct for theeffect of possible variation of irradiated volume with angle as wellas the effect of scattered radiation (outside the field) on the resultof measurement. Since the distance between the radiation sourceand the centre of the 2 cm segment is 100 cm, the varyingdistance to the source over the length of this 2-cm segment fibreis negligible.

The optical fibre was coupled to the PMT that converted theoptical signal into electrons with additional gain provided byimpact ionisation. The electrical signal was shaped using a pulseshaping circuit, such that the peak voltage from the circuit wasproportional to the total charge in each PMT pulse (Mei et al.,2006). The time constant of the circuit was 250 ms, such that itwas much longer than each LINAC pulse (about 3.6 ms), but alsoshorter than the time between each LINAC pulse (�6.2 ms)(Mei et al., 2006). The output of the pulse shaping circuit wasmeasured approximately 128 times per angle, and each measure-ment was repeated three times over different days. The averagepeak voltage was calculated in all measurements at each angleand was used as the light output for that angle. This was repeatedfor all desired angles between 01 and 1801 to yield the relativelight output as a function of angle, and these data points werenormalised to the maximum light output.

The accuracy of the results was highly dependent on theprecise alignment of the optical fibre with the central axis ofthe electron beam. To ensure proper alignment, a hollow steel rodof the size of the fibre was placed in line with the treatment roomlasers. The stage was rotated in small increments until the highesttransmission of the electron beam through the hollow rod wasobtained, as verified using a film. The rod was then replaced bythe optical fibre to align the fibre with the central axis of theelectron beam.

Based on the experimental setup, at the angle of 01 the opticalfibre was facing the electron beam and aligned with its centralaxis while at 1801 the fibre was in the opposite direction. If weconsider a as the incident angle of the electron beam relative tothe centerline of the optical fibre, then a was equal to the angle ofrotation of the rotating stage.

A similar experiment was repeated with a low energy(100 kVp) x-ray beam from the orthovoltage unit to assess theangular dependence at energies below the Cherenkov radiationthreshold energy. Here, the optical fibre was coiled five timesaround an acrylic block on the stage to overcome the otherwisetoo-weak-to-measure low light output at lower energies and wasrotated between 301 and 1501, as shown in Fig. 5. We note thatthis setup is less ideal for the angular dependence measurementas compared to the MV case, but will not affect the conclusions of

Fig. 5. Angular dependence measurement setup using 100 kV x-ray source. The

optical fibre was wrapped around the acrylic block and was lead away from the

block towards a PMT outside the treatment room. The acrylic block was on a

rotating stage, and the block was rotated to measure the light output at various

angles of rotation. The light output was measured with and without the lead

shielding at every angle. The difference between these two sets of measurements

yielded light output only from the segments of fibres behind the lead shielding,

which has a constant irradiated volume at all beam incident angles.

this paper. The x-ray beam was collimated by a 5 cm diameterapplicator, and the light output from the fibre was measured withthe PMT. To achieve a constant irradiated volume of fibre atdifferent angles, lead shielding was used. The measurements wereperformed twice: once with lead shielding (1 cm wide and 2 mmthick) covering the middle 1-cm segment of fibres, and again withthe lead removed. The first scenario served as a backgroundmeasurement, while the difference between the two cases wasthe angular dependence of only the 1 cm portion of optical fibresin the field. The experiment was performed four times onseparate days and the average of all four trials was calculated.

3.3. Spectrum of radiation induced light in optical fibres

In this experiment, the energy intensity spectrum of radiat-ion induced light was measured and compared with that ofCherenkov light. To overcome the relatively low sensitivity ofthe spectrometer used (Ocean Optics, USB2000) the optical fibrewas coiled at the base of a water tank to increase the light outputof the fibre, as shown in Fig. 6. The optical fibre was irradiatedwith a 15 MeV electron beam for the high energy case (with alarge 20�20 cm2 field) and a 100 kVp x-ray beam for the lowenergy case (with a 20 cm diameter applicator), with the largefield ensuring that the entire coiled area was irradiated. Theoptical fibre continued outside the tank into the spectrometer(integration time of 30 s), for measurement of the energy inten-sity spectrum. A depth of 4 cm of water was used for the highenergy case while only a few millimeter of water (with the watersurface just covering the fibre) was used for the low energy casein order to maximise the light output. Approximately 100 spec-trum measurements were taken, both with the beam on andbeam off (background), and the difference between the twoyielded the radiation-induced energy intensity light spectrum.

In order to account for any possible wavelength-dependentvariation in the detective quantum efficiency of the spectrometerand the light attenuation in the optical fibre, the spectrometerand the fibre were calibrated using a calibrated tungsten halogenlight source (Ocean Optics, LS-1 CAL) with a known energy-intensity spectrum. During calibration the light source wascoupled to the fibre used in the experiments, and the ratio ofthe known energy-intensity (in mJ) of the light source to a singlereading of the spectrometer (integrated over 60 s, in counts) fora given wavelength resulted in a calibration curve (mJ/counts).To account for a possible non-linear response of the spectrometer

Fig. 6. Experimental setup to measure the light output spectra of radiation

induced light from a 15 MeV electron beam and a 100 kVp x-ray beam. The

optical fibre was coiled at the base of a water tank (flat on the bottom) and one

end of the fibre continues out of the tank towards a spectrometer. A depth of 4 cm

water was used for the high energy case while only about 3 mm of water was used

for the low energy case.

I. Silva, G. Pang / Radiation Physics and Chemistry 81 (2012) 599–608 603

to different light intensities, a non-linearity correction factor wasalso applied to the calibration curve using a non-linearity correc-tion feature in the software (SpectraSuite, Ocean Optics). Thiscorrection factor was found to be very small (o1.5%) for therange of counts being considered. In the spectrum measurement,the readings of the spectrometer (in counts) were multiplied bythe calibration curve to obtain the energy intensity spectrum ofthe radiation induced light in the optical fibre.

3.4. Temporal property of radiation induced light in optical fibres

In this experiment, an optical fibre was placed on a rotatingstage similar to the angular dependence measurement setup. Theoptical fibre was in line with an 18 MV x-ray beam from theLINAC, whose field size was 10�10 cm2. An MV x-ray beam wasused in this experiment because the timing and shape of the x-raypulses can be precisely monitored through the Beam-I connectorin the virtual machine interface box of the LINAC; this is not thecase for electron beams. The signal from Beam-I represents thecurrent of electrons incident on the x-ray target to generate xrays, and thus the Beam-I pulse is proportional to the x-ray beampulse (there is no Beam-I signal for the electron beam mode dueto the absence of the x-ray target). The optical fibre was rotated toapproximately 471 to maximise the collection of Cherenkovradiation in the fibre (see Section 4.2). The optical fibre wasextended outside the treatment room into the PMT for whichoutput as a function of time was measured on an oscilloscope.The Beam-I pulse was also measured on a different channel ofthe oscilloscope, such that the Beam-I pulse duration and thePMT output duration could be compared in time. A backgroundmeasurement was also taken with the primary x-ray beamblocked by closing the jaws on the LINAC, and subtracted fromthe primary measurement.

It is important to note that the Beam-I signal from the x-raytarget was transmitted down a 30 m long co-axial cable (furtherreferred to as the Beam-I cable) leading outside the treatmentroom to the virtual machine interface box (and then to theoscilloscope). The long cable caused a time delay of the signaldue to signal’s propagation down the cable. To assess the timedelay caused by these co-axial cables, additional co-axial cables ofdifferent lengths were added between the virtual machine inter-face box and the oscilloscope, and the time difference betweenthe PMT output and the Beam-I output were measured for eachco-axial cable length. The first cable was a short cable (3 m co-axial cable from the virtual machine interface box to the oscillo-scope) while the second was a long cable (30 m long cable), bothused in addition to the 30 m long Beam-I cable from inside thetreatment room to outside.

During the measurements with the oscilloscope, the Beam-Isignal was used as the reference time frame, i.e., each measure-ment was triggered to the Beam-I output channel. Any timedifference between the PMT output and the Beam-I outputappeared on the oscilloscope in the form of PMT output shiftsrelative to the Beam-I output.

3.5. Theoretical model

This section theoretically models the variables involved duringthe angular dependence measurements for the high energy case.Based on our experimental result (see Section 4.1), about 2.35% ofthe light that is generated in the fibre at 6 MeV and detectedby the PMT is from fluorescence while about 97.6% is fromCherenkov radiation. Thus, fluorescence is not significant in thefibre and can be ignored in the modelling. When a MV electronbeam is incident on the fibre, Cherenkov light is produced alongthe electron tracks inside the fibre. The factors that affect the

angular dependence of the fibre light output during irradiationinclude the electron path length (i.e., the continuous slowingdown approximation range or CSDA range) inside the opticalfibre, the fraction of light internally reflected at each incidentangle, and the electron beam spread outside the optical fibre. Wenote that a detailed theoretical model of radiation induced light inoptical fibres has previously been developed by Law et al. (2007,2006) without including any angular spread of the incidentelectron beams. We found that in our case it was necessary toinclude the angular spread of the incident electron beam in themodel in order to obtain a better agreement with ourexperimental data.

The intensity of Cherenkov radiation created in the irradiatedfibre is proportional to the electron path length inside the opticalfibre (Zrelov, 1970). The electron path length changes as afunction of the optical fibre’s angle relative to the electron beam.In this model, we assume that all electrons crossing the 600 mm-diameter fibre core move straight inside the fibre core, which isan approximation. The electron path length is limited by themaximum electron range in the fibre core at, or near, the extremeangles of rotation (i.e., 01 and 1801), which is about 3.16 cm for a15 MeV electron beam in silica. At other angles, the path length(D(a)) is related to the diameter of the fibre core (Dcore) and angleof rotation (a) by

DðaÞ ¼ Dcore

sinðaÞ: ð4Þ

The amount of light output from the fibre is also stronglyrelated to the fraction of created light that is internally reflecteddown the core of the optical fibre. The angle of internal reflectionfor the fibre used is approximately 121, which was verifiedexperimentally by directing a laser into one end of the fibre andmeasuring the spread of light output from the other end of thefibre. Thus, all light emitted at an angle less than or equal to 121about the centre of the fibre core will be transmitted down theoptical fibre via total internal reflection (Crisp and Elliott, 2005).In the present case only a fraction of the total Cherenkov lightcreated within the fibre core overlaps with the region of totalinternal reflection for light transmission for a given angle ofincidence, and a detailed calculation for this fraction is given inAppendix.

The final factor influencing the angular dependence of the lightoutput from the fibre is the possible angular spread of theelectron beam about the primary angle of beam incidence. Thisspread may be due to the scattering caused by several compo-nents in the LINAC head, particularly the scattering foil, beforereaching the fibre. In order to determine the actual magnitude ofangular-spread of the electron beam, a sheet of film was used tomeasure the size of the electron beam at different source-to-filmdistances. The film was set perpendicular to the electron beamand irradiated with the 15 MeV electron beam. The spread of theelectron beam at 10 cm away from the lead collimators wasmeasured to be approximately 1.5 cm outside the optical field(about 0.6 cm diameter), meaning that the angle of electronspread was approximately 7.51. The angular spread of the incidentelectron beam is modelled using a Gaussian function given by

Gðx,sgÞ ¼1ffiffiffiffiffiffiffiffiffiffiffiffi

2ps2g

q e�ðx�mÞ2=2ps2

g , ð5Þ

where x is the actual electron incident angle, m equals the primaryangle of beam incidence a and sg (¼31) is the standard deviationof the Gaussian function.

The final light output from the fibre is given by

Iða,sgÞ ¼

Z 1800

00DðtÞIRðtÞGða�t,sgÞdt, ð6Þ

I. Silva, G. Pang / Radiation Physics and Chemistry 81 (2012) 599–608604

where I represents the light output at each primary beam incidentangle a, D(a) is the electron path length factor given in Eq. (4),IR(a) is the fraction of light internally reflected for a given a (seeAppendix), and G(a,sg) is given in Eq. (5). Based on Eq. (6), we cancalculate the relative light output from the fibre as a function ofthe electron beam incident angle and compare the result withthat of experiment.

Fig. 7. Relative light output of the irradiated fibre as a function of beam incident

angle obtained from the angular dependence measurement at 15 MeV. Three

curves are shown: one obtained before and one after the fibre was pulled back by

2 cm and the difference between these two curves representing the light output

from a 2 cm segment of optical fibre. The light output was normalised to the peak

light output of the 2 cm segment curve.

Fig. 8. Results for 100 kV x-ray source angular dependence measurement. Light

output was measured by a PMT at various angles between 301 and 1501, and the

relative light output was plotted versus angle of rotation of the fibre. The curve

represents the mean of four trials, where each trial was normalised to the

301 value.

4. Results

4.1. Magnitude of radiation induced light in optical fibres

Table 1 shows the measured PMT light output under differentsources of irradiation. Each measurement was the mean of 512captures, and each case outlined in Table 1 was the mean of 20measurements. The results of Table 1 show the fibre light outputat each energy for the same dose rate.

As shown in Table 1, for the same dose rate at the fibre, thelight output of the fibre for the low energy kV irradiation was only2.35% and 2.28% of that for 6 MeV electron irradiation and 15 MeVelectron irradiation, respectively. Thus, combined with the resultsgiven below this indicates that the Cherenkov radiation is thepredominant light source in the optical fibre irradiated withMV beams.

4.2. Angular dependence of radiation induced light

Fig. 7 shows the two measured angular dependence curves:one obtained before (open circles) and one after the fibre waspulled back by 2 cm. A considerable reduction in light output wasvisible when the fibre was pulled back outside the primary field ofthe beam, as the fibre was mostly irradiated by scattered radia-tion at that point. The curves represent the mean of three samples(obtained over different days) each with 128 measurements, andthe error bars are the standard error of the mean.

Subtracting the mean light output of the pulled back fibre fromthe mean light output of the fibre measurement before it waspulled back yields the net light output from a small 2-cm segmentof optical fibre (stars in Fig. 7) whose volume irradiated wasconstant at all angles of fibre rotation. Since the net light outputcurve is the difference of two curves, the error bars are calculatedby adding the error bars from each curve. The net light outputcurve for the 2-cm segment of optical fibre shows a clear peak atabout 471 and a relative light output that approaches 0 at theextreme angles. We note that the measured angular dependencecurve before the subtraction (open circles) is strongly asymmetric(around the maximum) as compared to that obtained after thesubtraction (solid stars) mainly due to the variation of irradiatedvolume of the fibre with angle before the subtraction.

The angular dependence curve was also measured for a fibreirradiated by 100 kVp x rays, and is shown in Fig. 8. The curve wasthe mean of four different trials performed on different days. Eachtrial measured the light output between 301 and 1501 where thevalue at each angle was the mean of 70 measurements, and eachmeasurement was the mean of 256 samples. The error bars arethe standard error of the mean of the four trials. Unlike the high

Table 1Measured light output from an optical fibre under different sources of irradiation.

Irradiation source Normalised PMT light

output (mV min/cGy)

100 kVp x-ray beam 1.80

6 MeV electron beam 76.57

15 MeV electron beam 78.91

energy case, this curve shows no clear peak and is relativelyconstant over all angles suggesting no angular dependence oflight output for the low energy case. This, combined with theresult of the spectrum measurement given below, indicates thatthe radiation induced light for the low energy (kV) case is notCherenkov radiation.

4.3. Spectrum of radiation induced light in optical fibres

The measured energy spectrum of light photons emitted in thesingle fibre when irradiated by 15 MeV electrons is shown in Fig. 9,where the energy spectrum in photon energy (mJ) per wavelength iscompared with the theoretical curve 1/l3. The theoretical curve wasfound by adjusting the coefficient k in k/l3 to fit the experimentaldata with the least-squares error. To minimise the effect of outlierdata on the theoretical fitting, the least squares fit was found for a

Fig. 9. Measured energy spectrum of radiation-induced light in the optical fibre

irradiated with 15 MeV electrons in comparison with the least-square fitted,

theoretical 1/l3 curve for Cherenkov radiation.

Fig. 10. Measured energy spectrum of radiation-induced light in the optical fibre

irradiated with 100 kVp x rays.

Fig. 11. Measured Beam-I and PMT outputs as a function of time after corrections

for the propagation delays in the cable and fibre (see text).

Fig. 12. Theoretical model (solid line) of angular dependence curve in comparison

with experimental data measured with 15 MeV electrons.

I. Silva, G. Pang / Radiation Physics and Chemistry 81 (2012) 599–608 605

low-pass filtered energy spectrum (whose high frequency compo-nents are filtered out with a cut off frequency of 0.3 nm�1). The bestleast-squares fit was then plotted as the theoretical curve versus theoriginal energy spectrum measured in Fig. 9. It can be seen fromFig. 9 that the light output energy spectrum shows a definitewavelength dependence and an excellent fit with the theoreticalcurve 1/l3 except for a small deviation for the peak at around650 nm. The peak at 650 nm also causes an upward shift in they-intercept of the theoretical fit, resulting in a small deviation ataround 550 nm. Fig. 10 shows the light output energy spectrumfrom the same fibre irradiated by 100 kVp x rays, which has acompletely different shape with a clear peak at approximately650 nm due to light sources other than Cherenkov radiation. Thelocation of this non-Cherenkov light peak matches the small peakshown in Fig. 9, and is likely the cause for the small deviationbetween the theoretical curve for Cherenkov light and the measuredlight spectrum at 15 MeV. We note that the 650 nm peak was notobserved in de Boer et al. (1993) and thus is fibre specific.

4.4. Temporal property of radiation induced light in optical fibres

The timing of the Beam-I and PMT outputs was measured withdifferent lengths of additional co-axial cable between the virtualmachine interface box and the oscilloscope. Both a 3 m and 30 m

co-axial cable were used to assess the time delays caused by theBeam-I cables from the experimental setup. The addition of 27 mof cable caused a signal delay of 0.14 ms. Based on this result, a30 m long segment of Beam-I cable would cause a delay of0.155 ms. Additionally, it takes approximately 0.05 ms for theCherenkov light to propagate down the 10 m long optical fibre.Thus, to account for the propagation delay in the Beam-I cableand the fibre, the measured Beam-I signal and PMT signal (with-out the presence of the additional cable between the virtualmachine interface box and the oscilloscope) were shifted to theleft on the time axis by 0.155 ms and 0.05 ms, respectively, toobtain the final result shown in Fig. 11.

Fig. 11 shows the timing of the measured Beam-I and PMToutputs after corrections for the propagation delays in the cable andfibre. It can be seen from Fig. 11 that once corrected, the Beam-Isignal (or the x-ray pulse signal) and the PMT output signal (or theCherenkov light output signal) are almost simultaneous in time. Theresults shown in Fig. 11 are the average of 29 samples where eachsample from the oscilloscope was an average of 512 measurements.Fig. 11 is plotted with no error bars for a clear comparison of thetime difference between the Beam-I and PMT output.

4.5. Comparison between experiment and theory

Fig. 12 shows the final model (i.e., Eq. (6)) plotted against theexperimental 15 MeV angular dependence result and shows a

I. Silva, G. Pang / Radiation Physics and Chemistry 81 (2012) 599–608606

strong match in shape between each curve, and a slight deviationin their locations of the peak.

5. Discussion

We have measured the angular dependences and spectrum-and-temporal properties of the radiation induced light in opticalfibres. Although angular dependence light-output curves andradiation induced light-output spectra were measured previouslyusing high energy electron beams for some optical fibres (Beddaret al., 1992; de Boer et al., 1993; Law et al., 2007; Lambert et al.,2009), the optical fibre used in the present study is different fromthose used in previous studies (Beddar et al., 1992; de Boer et al.,1993; Law et al., 2007) as they were different types or fromdifferent manufacturers. In the previous studies, some authors(Beddar et al., 1992; Law et al., 2007) have shown that the angulardependence curve for an electron beam is asymmetric, i.e., thecurve is broader in the small angle side than the larger angle sideof the maximum, while others (de Boer et al., 1993) demonstratethat the angular dependence curve is more or less symmetric. Wehave used a new method to measure the angular dependencecurve (see Section 3.2) and discovered that a variation of irradiatedvolume with angle can result in a strongly asymmetric curve. Thereis also confusion in the literature (de Boer et al., 1993; Lambert et al.,2009) regarding whether the Cherenkov light intensity spectrum(the number of light photons as a function of wavelength) or theCherenkov energy intensity spectrum (the total energy of emittedlight photons as function of wavelength) from an irradiated opticalfibre follows the l�3 law, where l is the optical wavelength. Webelieve it is the energy intensity spectrum that is proportional tol�3, where the energy intensity for a given wavelength is expressedby the product of the number of light photons for the wavelengthand the energy of each photon. This has been verified in ourexperiment (see Section 4.3).

Our experimental results have demonstrated that the predo-minant light source in the irradiated fibre with MV beams isindeed Cherenkov radiation. Other light sources in the fibre (witha wavelength of approximately 650 nm) only contribute about 2%of the total signal. Thus it is not difficult to understand the resultsof the spectrum and temporal-property measurements as shownin Figs. 9 and 11. However, the result of the angular dependencemeasurement (Fig. 7) needs further discussion. If we calculate theCherenkov angle yC (see Eq. (3)) in silica, we find that yC equals471 for 15 MeV electrons, which matches the peak location of theangular dependence curve shown in Fig. 7 (for the 2-cm segmentof optical fibre). This is understandable since the maximumintensity of Cherenkov light will be generated in the directionparallel to the central axis of the fibre when the fibre is at an angleof yC with respect to the incident radiation beam. Thus, thelocation of the peak depends on the Cherenkov angle yC. We notethat the full width at half maximum (FWHM) of the angulardependence curve is non-zero. This is because (1) any Cherenkovlight that is emitted within the acceptance angle of the fibre (121in the present case) will be transmitted down the optical fibre viatotal internal reflection; and (2) there is an angular spread of theincident electron beam (see Eq. (5)) and thus not all incomingelectrons have the same incident angle. The FWHM of the angulardependence curve would be increased with an increase ofnumerical aperture of the fibres or an increase in angular spreadof the incident electron beam.

We also note that the angular dependence curve of the opticalfibre was measured previously in Mei et al. (2006) for a 6 MVx-ray beam. Comparing the previously measured at 6 MV to thepresent one using a 15 MeV electron beam, we can see that theangular dependence curve obtained in Mei et al. (2006) shows a

peak at a smaller incident angle (about 401) with a larger FWHM.This is due to the facts that (1) fast electrons generated from theinteractions between the incident photons and the fibre have amuch broader angular distribution than the 15 MeV electronbeam (and thus resulting in a larger FWHM); and (2) the meanenergy of fast electrons generated by a 6 MV x-ray beam in thefibre is much lower than that of the 15 MeV electrons, resulting ina smaller Cherenkov angle yC and thus a shift of the peak locationto a lower incident angle.

The results of this work can be used to optimise the design ofthe Cherenkov detector. Our spectrum measurement shows thatthe energy intensity spectrum of radiation induced light in theoptical fibre follows the l�3 law. Thus, the emission is signifi-cantly stronger in the blue than red. This information is useful indetermining the type of AMFPI needed for the Cherenkov detec-tor. Since the detective quantum efficiency of an amorphous-selenium (a-Se) is much higher for the blue light than the red, ana-Se based AMFPI operated in the avalanche mode (Tanioka et al.,1988; Rowlands and Yorkston, 2000) is a natural choice forCherenkov detector in Fig. 1. In addition, Table 1 shows themagnitude of light output per dose rate obtained with the PMTgain of 6�105. This information will be useful to determine theavalanche gain needed in the AMFPI under different operatingconditions with different dose rates.

Our angular dependence measurement indicates that the lightoutput in the irradiated fibre with MV electron beams reaches itsmaximum at about 471 and zero at 01. Thus, to maximise the lightoutput in the optical fibres, the fibres should be placed with anangle of approximately 471 compared to the incident radiationbeam. However, the fibres in the Cherenkov detector in Fig. 1must be aligned with the incident x rays to improve the spatialresolution. Fortunately, for MV photon beams, the light output inthe irradiated fibre at 01 is only 35% less than that of themaximum (Mei et al., 2006). Thus, the signal output in the fibre at01 is significantly higher for MV photon beams than that for MVelectron beams. Furthermore, our angular dependence measure-ment shows that there is little light output when the incident angleof the radiation beam is larger than 601. Thus, scattered radiationthrough large angles would have less effect on the detector signal.This, combined with the facts that (1) scattered x rays usually havelower energies than the primary and (2) there is a threshold energyrequired for Cherenkov radiation, indicates that the Cherenkovdetector could be used as an anti-scatter detector.

The result of the temporal property measurement indicates thatthere is virtually no delay between the onset of the interactionbetween the optical fibre and the radiation beam and the creation ofCherenkov light output signal. This information would be useful ifwe want to use the detector to do time-of-flight measurements or todevelop a new time-of-flight positron-emission-tomography (PET)detector for other applications (Moses et al., 2010; Rowlands, 2009).In the case of PET, a PET detector based on Cherenkov radiationwould have a much better temporal resolution with an equal orbetter QE and spatial resolution, as compared to current scintillatorsused for PET, e.g., LSO (Lu2(SiO4)O:Ce) (Weber et al., 2003). However,the light yield of the Cherenkov detector would be much less (about2–3 orders of magnitude less) than that of LSO. Further investigationis needed on the feasibility of using a Cherenkov detector for PETapplication.

To further demonstrate the feasibility of using Cherenkov radia-tion for portal imaging, we have built a prototype array detector.It consists of 100 optical fibres with a 30 cm end-segment of thefibres enclosed as a single array in a solid buildup material with adensity similar to that of glass (Fig. 13). The remaining (approxi-mately) 10 m segment of optical fibres is mapped to a double arraysteel ferrule connected to a HARPICON camera (high field avalancherushing photoconductor, COHU 5300) and placed outside the LINAC

Fig. 13. Picture of a prototype Cherenkov detector array.

Fig. 14. Open field image taken with the Cherenkov detector array at 6 MV with a

dose of one LINAC pulse (about 0.026 cGy at the detector surface).

I. Silva, G. Pang / Radiation Physics and Chemistry 81 (2012) 599–608 607

room for image readout (Tanioka et al., 1988). The HARPICON camerauses a-Se as the camera target and operates in avalanche mode toreduce camera noise (Tanioka et al., 1988). Light attenuation in the10 m segment of optical fibres is less than 35% for any wavelengthbetween 400 nm and 900 nm. We note that the length of the opticalfibres in the proposed area detector in Fig. 1 is only �30 cm,resulting in little light attenuation in the proposed area detector.Fig. 14 is an open field image taken with the prototype Cherenkovdetector array at 6 MV using one LINAC pulse (about 0.026 cGy at thedetector surface). These white regions represent the light outputfrom the optical fibres irradiated with MV x rays. This imagedemonstrated the feasibility of using Cherenkov radiation for MVx-ray imaging. As a comparison, a similar open field image was takenusing 100 kVp x rays (and a dose of 0.15 cGy at the detector surface)but nothing was visible on the image (the signal was too weak).

6. Conclusion

Experiments have been performed using a single silica opticalfibre to evaluate both the energy and angular dependences, andspectrum and temporal properties of the radiation induced light inoptical fibres. Our results have demonstrated that Cherenkov radia-tion is indeed the predominant light source in optical fibresirradiated with MV beams. A prototype array detector has beenbuilt and low-dose MV x-ray open-field images have been obtained.

Fig. A1. Schematic diagram showing the relation among different angles (y, o, c, j).

Acknowledgements

This work was supported by the Natural Sciences and Engi-neering Research Council of Canada (NSERC) and the Canadian

Institutes of Health Research (CIHR). We also thank Dr. J.A.Rowlands and Dr. A. Vitkin for helpful discussions andG. DeCrescenzo and P. Au for assistance with experiments.

Appendix. Fraction of light internally reflected

The purpose of this appendix is to find the fraction of allCherenkov light (emitted on a surface cone) that is within theacceptance angle (121) of the optical fibre. The light is internallyreflected when the angle of the light relative to the fibre axis (o)is less than the acceptance angle (oo121). The angle between theoptical fibre and the electron beam (j) and the angle betweenelectron beam and light emission (y) are known. This appendixaims to derive an equation for o in terms of j, y and c. Thisequation can then be used to find all values of c (between 0 and2p) that yield oo121, and the fraction of all these values thatsatisfy this condition is interpreted as the fraction of lightinternally reflected.

Suppose the electron beam is incident along the z axis, and thefibre in the x–z plane, as shown in Fig. A1, with the symbols:

�

j—The angle between the optical fibre and the electronbeam axis. � c—The angle between the x-axis and the projection of OB (the

light emission direction) in the x–y plane.

� o—The angle between the optical fibre and the light emission

direction.

� y—The angle between the electron beam axis and the light

emission direction.

Considering triangles ABC and AOB, we have

AB2¼AO2

þOB2–2AO �OB � coso, (A1)

AB2¼AC2

þCB2–2AC � CB � cosc, (A2)

where lines AC and CB are perpendicular to the z axis. CombiningEqs. (A1) and (A2), we have:

AO2þOB2–2AO �OB � cos�o¼AC2

þCB2–2AC �CB � cosc. (A3)

I. Silva, G. Pang / Radiation Physics and Chemistry 81 (2012) 599–608608

On the other hand, using triangles COB and AOC, we further have

CB¼OB � sin y, (A4)

AC¼AO sin j, (A5)

OB cos y¼AO cos j. (A6)

Substituting Eqs. (A4)–(A6) into (A3), we obtain an equationthat relates o, c, y and j, i.e.,

cos o¼cos y cos jþsin j � sin y � cos c. (A7)

The fraction of light internally reflected (IR) can then be calcu-lated for a given beam incident angle of j using the followingformula:

IRðfÞ ¼Z 2p

0dcYð121�oÞ, ðA8Þ

where Y (x) equals 1 if xZ0 and 0 otherwise, and o satisfiesEq. (A7). Note the parameter y in Eq. (A7) equals yC, which is givenin Eq. (3).

References

Antonuk, L.E., 2002. Electronic portal imaging devices: a review and historicalperspective of contemporary technologies and research. Phys. Med. Biol. 47,R31–R65.

Beddar, A.S., Mackie, T.R., Attix, F.H., 1992. Cerenkov light generated in opticalfibres and other light pipes irradiated by electron beams. Phys. Med. Biol. 37,925–935.

Crisp, J.E., Elliott, B., 2005. Introduction to Fibre Optics. Elsevier Ltd., Burlington.de Boer, S.F., Beddar, A.S., Rawlinson, J.A., 1993. Optical filtering in spectral

measurements of radiation induced light in plastic scintillation dosimetry.Phys. Med. Biol. 38, 945–958.

Herman, M.G., et al., 2001. Clinical use of electronic portal imaging: report ofAAPM radiation therapy committee task group 58. Med. Phys 28, 712–737.

Jelley, J.V., 1958. Cerenkov Radiation and its Applications. Pergammon, London.Law, S.H., et al., 2007. Transmission of Cerenkov radiation in optical fibres. Opt.

Lett. 32, 1205–1207.Lambert, J., et al., 2009. Cerenkov light spectrum in an optical fibre exposed to a

photon or electron radiation therapy beam. Appl. Opt. 48, 3362–3367.Letourneau, D., Pouliot, J., Roy, R., 1999. Miniature scintillating detector for small

field radiation therapy. Med. Phys. 26, 2555–2561.Law, S.H., et al., 2006. Optical fibre design and the trapping of Cerenkov radiation.

Appl. Opt. 45, 9151–9159.Mei, X., Rowlands, J.A., Pang, G., 2006. Electronic portal imaging based on Cerenkov

radiation: a new approach and its feasibility. Med. Phys. 33, 4258–4270.Moses, W.W., et al., 2010. Optimization of a LSO-based detector module for time-

of-flight PET. IEEE Trans. Nucl. Sci 57, 1570–1576.Pang, G., Rowlands, J.A., 2004. Development of high quantum efficiency, flat panel,

thick detectors for megavoltage x-ray imaging: a novel direct-conversiondesign and its feasibility. Med. Phys. 31, 3004–3016.

Pistorius, S., McCurdy, B.M., 2002. Exit Dosimetry: Quo Vadis. Can. Med. Phys.Newsl. 48, 120–125.

Prasad, P.N., 2004. Nanophotonics. Wiley and Sons Inc, New Jersey.Rowlands, J.A., Yorkston, J., 2000. Flat panel detectors. Handbook of Medical

Imaging, Vol. 1. SPIE, Bellingham., pp. 223–328.Rowlands, J.A., 2009. Personal communication.Seppi, E.J., et al., 2003. Megavoltage cone-beam computed tomography using a

high-efficiency image receptor. Int. J. Radiat. Oncol. Biol. Phys. 55, 793–803.Samant, S.S., Gopal, A., 2006a. Study of a prototype high quantum efficiency thick

scintillation crystal video-electronic portal imaging device. Med. Phys. 33,2783–2791.

Samant, S.S., Gopal, A., 2006b. Analysis of the kinestatic charge detection system asa high detective quantum efficiency electronic portal imaging device. Med.Phys. 33, 3557–3567.

Sawant, A., et al., 2006. Segmented crystalline scintillators: empirical andtheoretical investigation of a high quantum efficiency EPID based on an initialengineering prototype CsI(TI) detector. Med. Phys. 33, 1053–1066.

Tanioka, K., et al., 1988. Avalanche-mode amorphous selenium photoconductivetarget for camera tube. Adv. Electron. Electron Phys. 74, 379–387.

Warkentin, B., et al., 2003. Dosimetric IMRT verification with a flat-panel EPID.Med. Phys. 30, 3143–3155.

Weber, S., et al., 2003. Comparison of LuYAP, LSO, and BGO as scintillators for highresolution PET detectors. IEEE Trans. Nucl. Sci. 50, 1370–1372.

Zrelov, V.P., 1970. Cherenkov Radiation in High Energy Physics. Atomic EnergyCommission, US.