Radio Detection of Ultra-High Energy Cosmic Rays · RADIO DETECTION OF UHECR discovery of CR...

30TH INTERNATIONAL COSMIC RAY CONFERENCE

Radio Detection of Ultra-High Energy Cosmic RaysHEINO FALCKE1,2

1Department of Astrophysics, Institute for Mathematics, Astrophysics and Particle Physics, Radboud Uni-versity, P.O. Box 9010, 6500 GL Nijmegen, The Netherlands2ASTRON, Oude Hoogeveensedijk 4, 7991 PD Dwingeloo, The [email protected]

Abstract. The radio technique for the detection of cosmic particles has seen a major revival in recentyears. New and planned experiments in the lab and the field, such as GLUE, Anita, LUNASKA, Co-dalema, LOPES as well as sophisticated Monte Carlo experiments have produced a wealth of new in-formation and I review here briefly some of the main results with the main focus on air showers. Radioemission of ultra-high energy cosmic particles offers a number of interesting advantages. Since ra-dio waves suffer no attenuation, radio measurements allow the detection ofvery distant or highly inclinedshowers, can be used day and night, and provide a bolometric measureof the leptonic shower component.The LOPES experiment has detected the radio emission from cosmic rays, confirmed the geosynchrotroneffect for extensive air showers, and provided a good calibration fomula to convert the radio signal intoprimary particle energy. Moreover, Monte Carlo simulations suggest that also the shower maximum andthe particle composition can be measured. Future steps will be the installation of radio antennas at theAuger experiment to measure the composition of ultra-high energy cosmic rays and the usage ofthe LOFAR radio telescope (and later the SKA) as a cosmic ray detector. Here an intriguing additionalapplication is the search for low-frequency radio emission from neutrinos and cosmic rays interactingwith the lunar regolith. This promises the best detection limits for particles above 10

21 eV and allowsone to go significantly beyond current ground-based detectors.

Introduction

Radio astronomy has always been closely con-nected to cosmic ray physics. Already the veryfirst cosmic radio emission detected by Carl Janskyin 1932 originated from cosmic ray constituents inthe Milky Way. We now know that at low ra-dio frequencies the diffuse Galactic radio emis-sion is mainly produced through synchrotron radia-tion of relativistic electrons. They are propagatingthrough the interstellar medium and the Galacticmagnetic field and were most likely accelerated insupernova explosions.

Also, the brightest radio sources discoveredthereafter, like quasars and radio galaxies (activegalactic nuclei) and supernova remnants, are todaythe main suspects for the origin of cosmic rays. So,without the advent of radio telescopes, we proba-bly would not know much about the non-thermaluniverse today.

It is therefore no surprise that radio antennaswere early on also considered for directly detect-ing cosmic ray air showers. In fact the huge Lovellradio telescope in Jodrell Bank was initially built inorder to detect cosmic ray radar reflection [14, 27]— it did not succeed but detected the radar reflec-tion of Sputnik instead and made history.

Radio detection of air showers has a number ofadvantages: the detector material itself, a simplewire, is cheap, radio emission is not absorbed in theatmosphere and can thus see the entire shower, andinterferometric techniques should allow relativelyprecise localization. But does this work in practiceand how does the radio signal actually look like?

I will here mainly summarize some of the mainrecent results and not recall the entire history ofthis field. Here one can point to the well-known re-view by Allan from 1971 [1] and a brief summaryof the early results given by Falcke & Gorham[20]. Main points were the prediction of radioCherenkov radiation by Askaryan [6, 7] and the

79

Proceedings of the 30th International Cosmic Ray ConferenceRogelio Caballero, Juan Carlos D'Olivo,Gustavo Medina-Tanco, José F. Valdés-Galicia (eds.)Universidad Nacional Autonóma de México,Mexico City, Mexico, 2009Vol. 6, pages 79-93

RADIO DETECTION OFUHECR

discovery of CR related radio pulses through Jel-ley et al. in 1965 [39] (a nice historical recount ofthe discovery was given by Trevor Weekes [67]).

Despite many experimental problems at thetime, quite a number of basic properties of airshower were established within a decade culminat-ing in the empirical “Allan formula” [1].

A long hiatus of this field began in the 1970’s,as witnessed by a quote from Alan Watson in his1975 ICRC rapporteur talk in Munich, where heobserved that “Apart from work at 2 MHz whichis planned for Yakutsk, it is clear that experimen-tal work on radio signals has been terminated else-where.”

Occasional attempts with single or few radioantennas at EAS-TOP and CASA/MIA [26] didnot lead to further radio detections in the 1990’s,making some colleagues (unjustifiably) even doubtthe reality of the earlier results in private conversa-tions.

Only in recent years, the technique has seen anastounding revival. A good overview is probablyfound in references [59, 20, 21, 52].

Scientifically, this started with attempts to de-tect radio emission from neutrinos hitting the moonby Hankins, Ekers, & O’Sullivan [29] and Gorhamet al. [25]. For air showers, the realization thatthe emission can be understood as geosynchrotronemission by Falcke & Gorham [20] and Huege &Falcke [32] also inspired new efforts.

Technologically, the revival is certainly dueto high-dynamic range digital radio receivers andpost-processing capabilities that are now available.There is also a general revival in low-frequencyradio astronomy as seen in a number of projectssuch as LOFAR[22], MWA[15], LWA[44] andGMRT[64].

In the following we will give a summary ofsome of the results in this field, with particular em-phasis on radio emission from air showers and re-sults obtained with LOPES.

Theory

After the experimental realization in the late 1960’sthat the Earth magnetic field is a factor in radioair shower emission, early theoretical modeling byKahn & Lerche considered the Lorentz boosted lat-eral current induced by the geomagnetic field [41],

an approach that has been revisited very recentlyby Werner & Scholten [68, 63].

A different approach was presented by Falcke,Gorham, and Huege [20, 32] where the radio emis-sion was explained in terms of “geosynchrotronemission”. This approach takes an important ex-tra factor into account, namely the curvature ofthe trajectories of the individual electron/positronpairs in the geomagnetic field. The fact that theemission region is smaller than a wavelength (andoptically thin) allows a relatively simple coherentaddition of the radio waves. This “single-particleapproach” makes it straightforward to combinethe radio emission with Monte Carlo calculations[33, 36]. The overall level of the geosynchrotroncomponent seems to be sufficient to explain thebulk of the observed radio emission [34]. Afterall, geosynchrotron subsumes most of the “lateralcurrent interpretation”.

However, also in the geosynchrotron picturea couple of extra effects still need to be takeninto account, such as the current induced bythe change in charges through creation, annihi-lation and recombination[63, 49]. The recom-bination of electrons may also lead to an addi-tional Bremsstrahlung component [49]. Also, thenon-zero refractive index of air and the originalAskaryan effect through Cherenkov emission fromthe charge excess [18, 50] will play a role at somelevel. The static Coulomb contribution should alsobe looked at [50] as well as optical depth effectsthat could become relevant for air showers around1020 eV (proposed in the context of radio radar ex-periments [23]).

Hence, while major progress has been made,there is still room for improvement. Nonetheless,the predictive power of current Monte Carlo codes,if coupled with air shower simulations, is probablyalready quite significant. This requires knowledgeof the lepton evolution in the showers and adequateshower libraries [47].

Calculations with the REAS2 code and COR-SIKA code by Huege et al. have recently shownsome interesting results [35]: If measured at a char-acteristic radius of∼ 300 m from the shower corethe radio signal is tightly correlated with the pri-mary particle energy (Fig. 1). Shower-to-showerfluctuations and different elemental compositionof the primaries induce just 5% variations in the

80


1e+08

1e+09

1e+10

1e+11

1000 10000 100000 1e+06

Nm

ax

32 to 64 MHz amplitude at 275 m [µV m-1]

iron proton gamma

1018 eV 1019 eV 1020 eV

Fig.1. Integrated 32-64 MHz radio flux measured at 275m from REAS2 simulations as function of energy fordifferent primary particles and showers with 60◦ zentithangle. Shower-to-shower fluctuations are only 5%. Fig-ure taken from reference [35].

radio flux. This is due to the fact that the radial ra-dio distribution on the ground pivots around a fewhundred m for different Xmax, depending a bit onshower geometry. This also means that measur-ing the radial slope of the radio emission shouldgive clues for the location of the shower maximum(Fig. 2) and together with the absolute radio fluxallow one to separate primaries of different el-emental composition. This tantalizing predic-tion naturally requires experimental confirmationbut already shows how important the theoreticalwork is.

LOPES

Quite a few experiments have been built or havebeen discussed in the last couple of years that em-ploy the radio detection method. We will try tobriefly discuss them here in turn.

A very productive experiment has been theLOFAR PrototypE Station (LOPES), which madeuse of early prototype hardware developed forthe LOFAR radio telescope (see below) and helpedto bring about the current renaissance in radio de-tection techniques[19].

LOPES [30, 19] was a collaboration of radioastronomers involved in LOFAR and the groupsinvolved in the KASCADE[45] and KASCADEGrande array[53]. The idea was to put a signifi-

600

650

700

750

800

850

900

950

1000

1050

1100

3 4 5 6 7 8 9 10 11

Xm

ax [g

cm

-2]

E275 m / E725 m at 32 to 64 MHz

iron proton gamma

1018 eV 1019 eV 1020 eV

Fig. 2. Location of shower maximum, Xmax, vs. lateralslope of radio emission from REAS2 simulations. Thetight correlation suggests that perhaps Xmax should bemeasurable with radio antennas and showers with 60◦

zentith angle. Figure taken from reference [35].

cant amount of radio antennas – allowing for inter-ferometric measurements – near a well-developedair shower array. This facilitates a cross corre-lation between conventional air shower measure-ments and radio observations.

LOPES consisted initially of 10 single dipoleantennas (LOPES10) that were then expanded to30 antennas (LOPES30). In the last phase LOPESwas again rearranged to have 20 antennas of which10 are in a dual polarization mode (Fig. 3). Polar-ization investigations are now underway (see Isaret al. [37]).

The LOPES antennas digitize the incoming ra-dio waves with a 12 bit A/D converter operatingat 80 MHz. An analog filter restricts the observ-able frequency range to 40-80 MHz, i.e. the secondNyquist zone. All antennas share a joint clock dis-tribution and have a 6.7 second ring-buffer whichis triggered and read-out roughly twice a minute bythe KASCADE array (Fig. 3).

LOPES itself is restricted to the dimensions ofKASCADE (200 m), but with the help of KAS-CADE Grande events out to 500 m can be seen [3].

The energies of cosmic rays seen in the radioat KASCADE is typically a few times1017 eV. In-clusion of KASCADE Grande provides informa-tion up to1018 eV. Below 1017 eV the radio signalvanishes in the noise.

81


Clearly, the Forschungszentrum Karlsruhe,where LOPES is located, is not an ideal location ofradio observations due to an enormous man-maderadio background noise (Radio Frequency Interfer-ence, RFI). This can be overcome somewhat in thepost-processing through digital filtering methods.

Also, in addition to the LOPES antennas afew additional log-periodic antennas (“Christmastrees”) with new electronics have been added inorder to develop a self-triggering algorithm [5].Here the background noise is an even more se-vere problem. On the other hand, studying self-triggering under these conditions, will allow oneto self-trigger almost everywhere else in the worldas well.

CODALEMA

Parallel to LOPES the CODALEMA experimentwas set-up[4], which initially had a complemen-tary approach: set up particle detectors nearan existing radio astronomy telescope at Nancay(France) and try self-triggering. This had the ad-vantage of a radio quiet site — much better thanLOPES — and well-calibrated antennas, but thedisadvantage of having to calibrate a new particle-detector array and to trigger on a yet not under-stood radio signal. In the latest version the CO-DALEMA array is now completely independentand employs a set of 16 wide-band active dipolesaligned on two 600 meter long baselines in theNorth-South and East-West directions. The radioarray is triggered by a ground detector array of 240meters square containing 13 plastic scintillator sta-tions. The recording bandwidth is 1-200 MHz, butsignal detection is mainly done around 50 MHz,where also LOPES operates. Hence, the two ex-periments are now quite compatible.

LOFAR

A next big step in radio detection of air showerswill be the LOFAR array which was planned asa large radio astronomy experiment [22, 58]. Inthe summer 2007 the project had to be downsizeddue to financial shortfalls, however, it will still bea major step forward. According to the currentplans, LOFAR will consist of 36 antenna fields(“stations”) in the Netherlands plus a number of

stations across Europe (E-LOFAR: Germany, UK,France, Sweden, Italy, Poland, Ukraine). The first20 stations are expected to operate early 2009.

Most of the antennas will be in a central con-centration (“core”) of 2 km diameter, where 18 sta-tions are foreseen. Each station has two sets ofreceiver systems operating from 10-90 MHz (low-band antennas, LBA, Fig. 4) and 110-240 MHz(high-band antenna tiles, HBA). The LBA fieldconsist of 48 dual-polarization inverted-V anten-nas, while the HBA fields consist of two sub-fieldsof 24 dual-polarization tiles. Each tile consistsagain of 16 bowtie-shaped fat dipoles (Fig. 5).

This means that in the inner 2 km there will bemore than 800 dual-polarization low-frequency an-tennas and about∼ 14, 000 high-frequency anten-nas (> 800 tiles) that will be able to observe brightradio events and deliver unprecedented detailed in-formation about radio shower properties. LBAsand HBAs share one receiver, so each LBA/HBApair cannot observe at the same time, however, it iswell-possible to have one half of the receivers ob-serve at the low-frequencies, while the other halfobserves at higher frequencies.

For normal radio astronomical observations theradio data from one station is combined into onedata-stream (“digital beam-forming”) to look in apredetermined direction. However, every antennais also connected to a one second ring-buffer withFPGA-based processing and triggering capability.This allows one to trigger on the raw radio datastream in an intelligent way. Since 8 antennasshare one memory board (with 4 FPGAs), there isthe possibility to trade the number of antennas fortriggering power or buffer length.

SKA

As a next step in radio astronomy at thelow frequencies the Square Kilometre Ar-ray (SKA) project is planned for> 2015(www.skatelescope.org), which will provide evenmore opportunities for radio detection of cosmicrays and neutrinos [21]. The SKA will employ amix of receptor technologies (dipoles, tiles, smalldishes) depending on the frequency range (70MHz- 10 GHz) and have a phased roll-out. The

82


y c

oo

r din

ate

[m

]

x coordinate [m]

Grande stations

LOPES 10LOPES 30

LOPESSTAR

DA

frontendPC

RAMmodule

2 GByte

active antenna

PCDAQ

GB>100

dig.data optical

transmit.

80 MSPS

40−80 Mhz

RF1 Gbit/s

amp+filter

80 MHz sample clock

40 MHz digital clock

sync signaldigital data

on optical fiber

dig.data PCI

bus

optical

receiver

Memory Buffer (TIM−Module)

2 x 1 Gbit/s2nd input

40 MHz

80 MHz

Clock Card5 MHz input

1 Hz input

trigger inputveto

time−stamp

Master ClockModule

& distributionclock generation

sync signaldistribution

distributionsync signaldistribution

clock

Slave Clock Module

sync signal

from KASCADE

sync signal

2 m, 100 m,or 150 m

RML (Receiver Module LOPES)

ethernet

100 m or 180 m

coax cable

Fig.3. Left: LOPES and KASCADE Grande Layout. Right: Schematic of LOPES Electronics.

Fig.4. LOFAR Low-Band Antennas.

Fig. 5. Inside of a LOFAR High-Band Antenna (HBA)tile, showing one crossed bow-tie antenna.

first phases will concentrate below 1 GHz and willbe of high interest for radio particle detection asdiscussed here.

Auger Radio

In the spirit of the LOPES experiment, putting ra-dio antennas next to existing cosmic ray experi-ments, it makes sense to also place radio anten-nas at today’s largest cosmic ray array, the PierreAuger observatory [66, 8]. This has been at-tempted recently with a few prototype radio an-tennas (van den Berg et al. [65]), including someLOPES antennas. The medium-term goal is tocover a 20 km2 region with self-triggering radioantennas. In the same region an infill array andan upgrade to a fluorescence telescope will lowerthe energy threshold of Auger. This will nicelyconnect to the LOPES and CODALEMA measure-ments in energy and allow for the first time triplecoincidences between radio, particle detectors, andfluorescence and hopefully further dramatically in-crease the quality of the data.

Radio in Ice and from the Moon

Apart from the radio air shower experiments a highlevel of attention has also been attracted by the pos-sibility to detect showers generated in solid me-dia, such as ice, salt or the lunar regolith. Thisgoes back to the original suggestion by Askaryan[6, 7]. Further theoretical progress in the 1990’s

83


by Alvarez-Muniz and Zas et al. [2, 69, 12] andsuccessful accelerator experiments, validating thetheory, breathed new life into this field [60, 24].

One idea is to use the huge detector volume ofthe moon and observe it with sensitive ground-based radio telescopes in search for nanosecondpulses which are dispersed by the Earth iono-sphere. First such experiments were made with theParkes radio telescope [29] and the 64 m Kalyazinradio telescope [13]. An experiment (GLUE) usingthe NASA deep space network antenna at Gold-stone received wide attention and produced in-teresting limits on the ultra-high-energy neutrinosflux [25]. Currently extensive experiments, LU-NASKA and NuMoon, are progressing at the Aus-tralia Telescope Compact Array (ATCA, James etal. [38]) and the Westerbork Synthesis Radio Tele-scope (WSRT, Scholten et al. [62]). Later LOFARand the SKA could be used to further improve thecurrent upper limits to interesting levels or evendetections [61, 21]. Even a detection of neutrinosusing radioon the moon has been considered[40].

Instead of looking up, one can also look downon the Earth and use for example the large ice sheetof Antarctica or salt domes as detector targets.While salt domes [51] are currently not pushedvery strongly due to high drilling costs, experi-ments involving Antarctica are flourishing.

For quite some time already the RICE experi-ment [46] has radio antennas in the ice near the Ice-Cube array [42] and a major extension of IceCubewith radio antennas is actively discussed [42, 43].Alternatively radio antennas have been tested at theRoss ice shelve in Antarctica (ARIANNA) whichis logistically more conveniently located [9, 11].

An alternative approach to embedding a largenumber of radio antennas in the detector volumeis to just fly over it and use the long range capa-bility of radio detection. Gorham et al.[48] haveused a military satellite (FORTE) to search for neu-trino induced radio pulses from the ice. Recentlythe dedicated balloon experiment ANITA [10] waslaunched for the first time to circle Antarctica andto detect there distant radio pulses from up-goingneutrinos. Unfortunately the flight was cut short byunfortunate wind conditions, but nonetheless thedata analysis is proceeding.

This brief summary already shows that thenumber and breadth of radio experiments for cos-

Fig.6. Gain pattern of a LOPES antenna at 50 MHz (top)and 70 MHz (bottom) showing the entire sky. The colorscale goes from a gain of 0 (black)to 5 (white).

mic ray and neutrino detection is rather largealready.

Experimental Results & Calibration

In the following we will summarize some of theimportant experimental conclusions concerningthe air shower radio properties that have beenfound recently, here mainly focused on air show-ers and based on the LOPES results.

First of all one has to realize that the simplic-ity of the antenna comes at the cost of more com-plicated calibration. The sensitivity of a dipoledepends on frequency, direction, and polarization.For LOPES the absolute calibration has been per-formed as part of the PhD thesis of S. Nehls [54]using an elevated calibrated reference antenna. Onthe other hand, given the simple structure of the an-tenna, it is also possible to calculate the expectedbeam pattern on the sky using standard antennasimulation packages.

84


An example is shown on Fig. 6, which showsthat the beam shape is elongated and even notpeaking towards the zenith for frequencies abovethe resonance frequency of the LOPES dipoles(∼ 60 MHz). The elongated structure is related tothe orientation of the dipoles (EW) and would berotated by 90◦ for the other (NS) polarization. Inturn this also means that the crossed-dipole, if un-calibrated, will always produce highly polarizedsignals.

Energy Calibration

In addition to the antenna dependencies, the radioemission will also depend on the shower geome-try. The main factors that have been identified arethe particle energy,Ep, the angle between showeraxis and geomagnetic field (“geomagnetic angle”),α, the zenith angle,θ′, and the distance from theshower core,r. For an east-west polarized an-tenna one finds that the radio emission for showersat the same energy and distance is proportional to1 − cos α (Fig. 7a), the signal drops exponentiallywith radius (Fig. 7b), and increases linearly withprimary particle energy (Fig. 7c).

Altogether this has been nicely parametrized inHorneffer’s formula [31] (at the moment valid onlyfor the EW polarization):

εest = (11 ± 1.) [(1.16 ± 0.025) − cos α] cos θ

exp

(

−r

(236 ± 81)m

)(

Ep

1017eV

)(0.95±0.04)

[

µV

mMHz

]

(1)

We note that the exponential decay of the ra-dio signal is also seen by the CODALEMA exper-iment [4].

This prescription can now be inverted to predictthe energy of the incoming particle. Comparisonbetween the energy predicted from radio with theenergy estimated from KASCADE Grande, showsa scatter of 27% between the two methods forEp > 1017 eV. This is very encouraging, given thatshower-to-shower fluctuations in the KASCADEGrande estimate alone should produce a 25% scat-ter. Hence, the scatter in the radio measurementsshould be much less.

Fig. 7. Calibration results of LOPES showing the nor-malized radio voltage vs. air shower parameters wherethe other parameters have been divided out. Each pointrepresents the average and spread of all events in thatbin. Panels a-c are from top to bottom.

This would support the claims from MonteCarlo simulations [35] that radio is a good tracerof the energy. The main reason why one suspectslower scatter in the radio measurements with re-spect to particle detection on the ground is the factthat radio emission in not absorbed in the atmo-sphere. Hence, radiation from every particle is vis-ible on the ground — it is in that sense a bolomet-ric measurement. Variations in the location of theshower maximum will be less dramatic comparedto measurements of particles on the ground whichare just a fractional tail of a quickly declining func-tion, whose values are quite sensitive to Xmax.

85


Fig.8. Example spectrum of a LOPES event for two dif-ferent methods of beam-forming. The measured spectralslope isε ∝ ν

−1. The lines represent two spectra fromMonte Carlo simulations. Also shown is the noise spec-trum that has been corrected.

Spectrum

One topic that had been difficult to tackle in thepast has been the spectral shape of the radio sig-nal, i.e. how much power is emitted at which fre-quency? Historic experiments were relatively nar-row band and non-simultaneous data had to becombined. Modern broad-band receivers allow oneto study the instantaneous spectral index, but re-quire careful bandpass calibration.

Nigl et al. [56] employed two methods to getthe spectral shape (Fig. 8): Fourier transform ofthe (not squared) electric field around the pulse po-sition and measurements of the pulse heights afterapplying narrow-band digital filters to the signal.Both methods give consistent results.

The spectra can be represented by a power-lawfunction or an exponential decay. For the narrowfrequency range of LOPES, extending only a factorof two, we cannot distinguish between the two pre-scriptions. For a power-law function the averagespectral index isν−1±0.3. This would mean thatthe power of the signal falls of withν−2. This isconsistent with the simple expectations of coherentgeosynchrotron [20] and only slightly steeper thanthe Monte Carlo simulations suggest [34]. Also,the Codalema experiment finds powerlaw spectrawith spectral indices in the range -1.5 to 0. LOPESsees spectral indices in the range -1.5 to -0.4, sothere is some agreement, but more detailed inves-tigations have to be performed in the future.

Fig. 9. Image of a bright radio flash at the time of max-imum. The dark dots mark the fitted LOPES positionand the direction found by KASCADE. The size of theellipse is determined by the expected image resolution.

Direction & Imaging

The next question then is, how well can we localizethe radio emission? This has become of particularimportance given the finding of anisotropies andcorrelations between cosmic ray arrival directionsand nearby extragalactic objects [8].

Using radio astronomical imaging techniques,we can actually image the radio flash from the airshower. For LOPES L. Bahren has developed aspecial tool (“skymapper”) which can actually dothis on the tens of nanosecond (i.e., the samplingrate) level (see Fig. (9) and in three dimensions(Fig. 10).

This issue of the positional accuracy has thenbeen further investigated by Nigl et al. [55] us-ing LOPES data. Conventional interferometry isvery sensitive to positional changes. For pointsources one expects an angular error∆αmin =±

12

1SNR

λD

in the azimuthal direction, where SNRis the signal-to-noise ratio,D the separation of theantennas, andλ the observing wavelength. Forhigh SNR images the positional accuracy for pointsources is always better than the image resolution.Hence, for an SNR of 10 for an antenna separa-tion of 100 m and observing frequencies around 60

86


Fig. 10. Three dimensional view of the radio emissionfrom a cosmic ray air shower detected with LOPES. 3Dsidelobes have been “cleaned” by only displaying high-brightness regions. The peak at 3 km height appearsto be real, but the extended structure along the axis isstrongly affected by remaining sidelobes.

MHz, as in LOPES, one expects an error of only0.15◦, while the point-spread function of the inter-ferometer has a much larger width of about 2-3◦ inazimuth.

Comparisons of the shower direction (assum-ing a fixed shower core) between KASCADE andLOPES, actually shows an average offset of1.3◦

(Fig. 11), which does decrease with increasing sig-nal level (and increasing SNR). This is not badcompared to the imaging resolution of LOPES, butwe would have expected better results still. So,what is the dominating source of error?

Some insight can be obtained from Fig. 12which shows that the location of the radio centroidon the sky is also a function of the distance or ra-dius of curvature. One has to remember that theshower maximum of the showers that LOPES seesis just a few km high. This is still in the near fieldof the interferometer. This means that radio wavesemitted in the shower maximum will not appearas a plane wave, but will show a curvature with aradius corresponding to the distance from the ob-server. In the real world, the wave front will beeven more complicated, since the emission is notconstrained to a small region but extends along the

Fig. 11. Offsets on the sky between the direction ofthe air shower axis as measured with KASCADE andLOPES vs the radio pulse amplitude. Events with cir-cles were taken during thunderstorm conditions.

shower axis. In principle one has then a wavefrontthat is the superposition of many spherical wavesemitted at different positions.

In a proper 3D imaging process one would tryto deconvolve the data and put together a 3D imagecube. We are not yet able to do this. So, all we cando at present is to try different radii of curvatureand search for the maximum in the emission. Thisis what Fig. 12 shows: different cross sections of aradio image focused at different distances in stepsof 250m. The maximum is found at a distance ofabout 3 km. This is 30 times farther than the typi-cal baselines on the ground and only possible dueto the good SNR. What is clear from the figure isthat not only the emission level changes but alsothe position of the maximum. The problem is thatfor small radii of curvature a small change in ra-dius is similar to an inclination of a plane wave.Inclining the (virtual) receptor plane implies a po-sitional shift on the sky. As seen in this example,the shift can be up to 3◦. Hence, an error in the ra-dius of curvature determination or – perhaps moreimportant – a non-spherical wavefront will propa-gate into an positional error!

This requires that radio shower parameters arereally derived from a 4 (or 6) dimensional datacube, consisting of time, 3 spatial coordinates, andpotentially 2 shower core location parameters [3].

So, any experiment will improve its spatial ac-curacy not only with greater baselines but also withincreasing the number of antennas. In addition we

87


Fig. 12. Cross section along the elevation axis in an im-age of a cosmic ray radio flash for different radii of cur-vature. The maximum field strength is found around aradius of 3km.

need to understand the exact geometry of the radioshower front. Here, further simulations and the de-tailed observations with the many antennas of LO-FAR should clarify that issue. This may make fur-ther dramatic improvements in the astrometry ofradio air showers possible.

Electric Fields and Lightning

One other important factor that has been looked atwith some worries, is the influence of the atmo-spheric electric field on the radio emission. Whilethe Earth magnetic field is very stable, the electricfield can change significantly from 1-10 V/cm dur-ing fair weather, to 100 V/cm in heavy rain clouds(Nimbostratus), and up to 1000 V/cm in severethunderstorms. If the radio emission is affected bythe electric field one would not be able to inter-pret the radio signal quantitatively, since measur-ing precisely the instantaneous electric field struc-ture is almost impossible.

To get a first idea of the importance of theelectric field, we could simply look at the Lorentzforce,F = e( ~E + ~v/c × ~B) in cgs, which is driv-ing the geosynchrotron emission. For a relativisticparticle (v/c ' 1) the electric field will then dom-inate if E > B ' 150V cm−1(B/0.5G). Hence,from this simple approximation one would expect

Fig. 13. Radio pulse height (normalized by energy) ver-sus the geomagnetic angle. Blue dots are taken duringclear weather, purple during nimbostratus, and red pointsduring thunderstorm conditions.

only for severe weather a modification of the radiosignal.

Buitink et al. [16] investigated radio pulseheights with LOPES under different weather con-ditions. They selected time slots where the localweather station recorded clear weather, heavy rain,or thunderstorms and compared the three (Fig. 13).Within the errors, the radio pulses followed thepreviously found correlation with the geomagneticangle (if normalized to the same energy and ra-dius). However, significant outliers are found inthe thunderstorm data set – and only there. Theamplification of the radio signal can be a factor ten.

This seems to confirm the simple estimate wemade above and the fact that the Lorentz forceis dominating. Further verification comes fromMonte Carlo simulations. Buitink et al. [17] havenow included electric fields in the CORSICA andREAS2 codes, allowing one to model the E-fieldinfluence in detail. The simulations have calcu-lated radio pulses for different values of the elec-tric field. Again a significant amplification is seenas soon as the E-field reaches 1000 V/cm, whileat 100 V/cm the radio emission shows only littledifferences.

The CORSIKA simulations also show a modi-fication of the electron/positron energy distributionand the shower structure. Many pairs are deflectedsignificantly from the shower axis, which might bedetectable with particle detectors.

88


Finally, we note that besides the radio pulseheight also other parameters are impacted by thun-derstorm electric fields. For example, in the inves-tigation of the emitted radio spectrum one brightradio event stood out with a much steeper radiospectral index. It was found to be a thunderstormevent.

Moreover, also in the positional offsets be-tween LOPES and KASCADE thunderstormevents stand out. In Fig. 11 all outliers are eventsmeasured during thunderstorms. Whether this isdue to an actual deflection or an asymmetry in theradio emission is not yet clear.

In summary, we can state that air showerspassing through the strong fields of thunderstormclouds are brighter, further offset from the show-er core measured on the ground, and have asteeper radio spectrum. For the measurement ofcosmic rays this means that radio –at least at fre-quencies above 40 MHz– remains a reliable tech-nique as along as data taken during thunderstormis discarded.

On the other hand, the current results stronglysuggest that in high E-fields not just the radio emis-sion is altered, but the entire shower (at least theelectronic part). This point may warrant furtherinvestigation for its own sake. Moreover, it hasbeen speculated that cosmic ray air showers couldplay a role in initiating lightning through a run-away breakdown effect [28].

Radio methods could help to investigate thisconnection experimentally, since both –the light-ning strike and the air shower– would be detectableby the same instrument. Also, further MonteCarlo simulations will investigate whether there isenough energy gain through the electric field orionization through the air shower to actually startthe runaway breakdown process. The LOFARproject will try to address some of these issues.

Self-triggering

Overall, the current results have provided a verycomprehensive picture of the radio properties.However, whether the radio detection will matureinto a standard technique depends on whether itis possible to actually trigger on the radio sig-nal. This has been attempted but the final break-through still stands out. Recent attempts were

made for example by the CODALEMA experi-ment [57]. Within LOPES, LOPESSTAR [5] hasbeen designed specifically to investigate this issuein more detail.

We do now understand where the challengeslie. In many cases, disturbing radio interferencehas actually been generated by the devices andelectronics of the cosmic ray experiments them-selves. So, designing radio quiet electronics,power supplies and data communication is an im-portant first step. Also, radio contains an enormousamount of information (as witnessed by every FMreceiver) and many processes of the modern worldproduce wanted and unwanted radio signals. So,filtering out the correct information is more com-plicated then just looking for a peak in the electricfield.

In a broad-band receiver the biggest contribu-tion of the basic noise level typically comes fromnarrow-band RFI transmitters. Hence, a digital fil-ter to cut these signals out – which is currently onlyemployed in the post-processing, is a crucial stepin the triggering electronics. Moreover, the char-acteristics of the pulse itself need to be consideredas well. For a human eye it is quite simple to dis-tinguish a cosmic ray pulse from those generatedby a passing 1970 Chevrolet. Hence pulse shapeparameters need to be used in the triggering. Thiswill in any case require a bit more intelligence onthe trigger board, than with conventional experi-ments. Such a pulse-shape parameter search to im-plement self-triggering in hardware is currently un-der way at LOPESSTAR showing some interestingprogress recently. These techniques will eventu-ally be tested at Auger and one can be hopeful thatthis last and crucial step will be achieved in the nottoo distant future.

Radio Pulses from the Moon

The main focus of this article was on the detec-tion of radio emission of air showers. However, wewant to end with a few comments on the prospectsfor radio emission from ultra-high energy cosmicrays with upcoming radio telescopes, in particularwith LOFAR. Here the atmospheric detection willbe implemented in the project — with the “tran-sient buffer board” being at the heart of this newtechnique. It turns out that this buffer-board may

89


1017

1018

1019

1020

1021

1022

1023

1024

10−12

10−11

10−10

10−9

10−8

10−7

10−6

E [eV]

E2 d

N/ d

E [

GeV

/cm

2 / sr

/ s]

Incoherent beam of 62 quarter stations

30 days

90 days

1 year

Spectral index = −3.3 Spectral index = −3.0 Spectral index = −2.7

30 days

90 days

1 year

Coherent beam of 36 quarter stations 77 full stations

Fig.14. Expected LOFAR sensitivities for detecting ultra-high energy cosmic rays hitting the moon.“Full stations” hererelate to the originally planned 96 tiles. 36 quarter stations are the currently planned layout for the central LOFAR core.“Coherent” and “incoherent” relate to the two methods of combining the datafrom different stations.

also improve the detection of cosmic rays hittingthe moon.

Scholten et al. [61] have shown that the 100-200 MHz range is ideal for detecting cosmic raysabove1020 eV hitting the lunar surface. In LOFARany such event could be detected in a beam formedtowards the moon. This detection could be used totrigger the buffer boards and download the rawdata from all LOFAR antennas. With the raw dataof the individual antennas the exact nature and ori-gin of the pulse could be determined much moreprecisely, if one uses some of the offline process-ing steps known from the air shower detection.

The question is how sensitive is this technique?The originally planned LOFAR would have beenvery favorable for this [61], however, due to thedownsize of LOFAR the sensitivity has decreased.K. Singh has now recalculated the expected sensi-tivity with the latest available station layout. Theresult is shown in Fig. 14. With the lower sensitiv-ity, the minimum energy that can be detected hasmoved up above1021 eV, which will decrease theexpected count rate.

For comparison we also show the latest Augerspectrum with an extrapolation of the power lawwith three spectral indices through the GZK cut-off. For a powerlaw ofE−3 LOFAR could in prin-

ciple reach this extrapolation with a total observ-ing time of 30 days, if a large number of tied-arraybeams can be formed – a mode that still needs tobe tested.

While strong evidence for a GZK cut-off above1019.6 eV has been seen by Auger, it cannot be ex-cluded that there will be some recovery of the spec-trum from local sources or neutrinos. Hence, it isstill worth looking in this regime. Radio will prob-ably be the only technique that can deliver at leastvery meaningful upper limits in this energy range.

Future expansions of LOFAR, such as theSKA, can improve these limits even more, actuallyreaching down to the GZK cut-off with the clearexpectation of actual detections. This will thenprobably provide the ultimate detection experimentfor cosmic ray particles at the highest energies wecan ever measure.

Conclusions

The radio detection technique for cosmic particleshas seen quite some ups and downs in the lastdecades. Pronounced dead in the 1970’s it has nowrisen from the ashes. But is it here to stay? Thechances are good at least.

90


First of all, we have made major progress inunderstanding the emission mechanism. For radioin solid media (Cherenkov emission) codes havebeen developed and accelerator experiments havebeen performed, giving trust in the reality of the ef-fect. For radio emission from air showers (geosyn-chrotron) good Monte Carlo codes and solid ex-perimental verification are now available and moreand more details are being worked in.

The LOPES experiment, and in some areas alsoCODALEMA, has given us already detailed infor-mation about the radio air shower properties andperformed a very useful cross-calibration betweenparticle and radio detectors. Moreover, major ex-periments have embraced the technique. The largeLOFAR radio telescope has the cosmic ray detec-tion built in, the Auger collaboration is testing it,ANITA has flown, and also the IceCube collabora-tion is seriously preparing radio experiments.

A few issues still need to be solved: how doesan optimal trigger system look like for radio anten-nas? What is the optimal layout for a large radioarray? Nonetheless, there is a good chance that ra-dio detection will become common place over thenext few years.

It will be interesting to see in which directionthis will develop. The simulations for air show-ers indicate that the addition of radio may increaseenergy resolution and directional accuracy. Alsocomposition information (through Xmax) seems tobe encoded in the radio signal. After the break-through of the hybrid technique with Auger, maybewe will see “tri-brid” detectors in the future. Moreand better data is almost always better in physics.If cosmic ray air shower arrays are to continue inthe next decades, they will likely include radio an-tennas. For the highest energy events, the new gen-eration of low-frequency radio telescopes provideshope that even particles above1021 eV could inprinciple be detected in the future — if they exist.

References

[1] H. R. Allan, Prog. in Element. Part. and Cos.Ray Phys. 10 (1971) 171.

[2] J. Alvarez-Muniz, R. A. Vazquez, and E. Zas,Phys. Rev. D 62 (2000) 63001.

[3] W. D. Apel et al., Astroparticle Physics 26(2006) 332.

[4] D. Ardouin et al., Astroparticle Physics 26(2006) 341.

[5] T. Asch et al. (LOPES collaboration), Pro-ceedings of the 30th ICRC, Merida, Mexico,2007, Vol. 5, p. 1081.

[6] G. A. Askaryan, Soviet Phys. JETP 14 (1962)441.

[7] G. A. Askaryan, Soviet Journal of Exper-imental and Theoretical Physics 21 (1965)658.

[8] (Auger Collaboration), Science 318 (2007)938.

[9] S. W. Barwick, ARIANNA: A NewConcept for UHE Neutrino Detection,arXiv:0610631[astro-ph].

[10] K.J. Palladino, Proceedings of the 30thICRC, Merida, Mexico, 2007, Vol. 5, p. 1441.

[11] S.W. Barwick, Proceedings of the 30th ICRC,Merida, Mexico, 2007, Vol. 5, p. 1601.

[12] A. R. Beresnyak,Expected properties of ra-dio pulses from lunar EeV neutrino showers,arXiv:0310295[astro-ph].

[13] A.R. Beresnyak et al., Astronomy Reports 49(2005) 127.

[14] P. M. S. Blackett and A. C. B. Lovell, RoyalSociety of London Proceedings Series A,1941, Vol. 177, p. 183.

[15] J. D. Bowman et al., AJ 133 (2007) 1505.[16] S. Buitink et al., A&A 467 (2007) 385.[17] S. Buitink et al. (LOPES collaboration), Pro-

ceedings of the 30th ICRC, Merida, Mexico),2007, Vol. 4, p. 161.

[18] R. Engel et al., Proceedings of the 29th ICRC,Pune, India, 2005, Vol. 6, p. 9.

[19] H. Falcke et al., Nature 435 (2005) 313.[20] H. Falcke and P. Gorham, Astropart. Phys. 19

(2003) 477.[21] H. Falcke, P. Gorham, and R. J. Protheroe,

New Astronomy Review 48 (2004) 1487.[22] H. D. Falcke et al., Highlights of Astronomy

14 (2007) 386.[23] P. W. Gorham, Astroparticle Physics 15

(2001) 177.[24] P. W. Gorham et al., Physical Review Letters

99 (2007) 171101.[25] P. W. Gorham et al., Physical Review Letters

93(4) (2004) 041101.

91


[26] K. Green et al., Nuclear Instruments andMethods in Physics Research A 498 (2003)256.

[27] A. G. Gunn, Jodrell Bank and the pursuitof cosmic rays, (EAS Publications Series,2005), p. 15.

[28] A. V. Gurevich and K. p. Zybin, Physics Let-ters A 329 (2004) 341.

[29] T. H. Hankins, R. D. Ekers, and J. D.O’Sullivan, MNRAS 283 (1996) 1027.

[30] A. Horneffer et al.,LOPES: detecting ra-dio emission from cosmic ray air showers(InJ. Hough and G. H. Sanders, editors);Grav-itational Wave and Particle Astrophysics De-tectors, (Edited by Hough, James; Sanders,Gary H. Proceedings of the SPIE, 2004,Vol. 5500, p. 129); Presented at the Societyof Photo-Optical Instrumentation Engineers(SPIE) Conference), 2004, p. 129.

[31] A. Horneffer et al. (LOPES collaboration),Proceedings of the 30th ICRC, Merida,Mexico, 2007, Vol. 4, p. 83.

[32] T. Huege and H. Falcke, A&A 412 (2003) 19.[33] T. Huege and H. Falcke, A&A 430 (2005)

779.[34] T. Huege and H. Falcke, Astroparticle

Physics 24 (2005) 116.[35] T. Huege, R. Ulrich and R. Engel, Proceed-

ings of the 30th ICRC, Merida, Mexico,2007, Vol. 4, p. 617.; T. Huege, R. Ulrich andR. Engel,Energy and composition sensitivityof geosynchrotron radio emission from EAS,arXiv:0707.3761[astro-ph].

[36] T. Huege, R. Ulrich and R. Engel, Astroparti-cle Physics 27 (2007) 392.

[37] P.G. Isar et al. (LOPES collaboration),Proceedings of the 30th ICRC, Merida,Mexico),2007, Vol. 5, p. 1093.

[38] C. W. James et al., MNRAS 379 (2007) 1037.[39] J. V. Jelley et al., Nature 205 (1965) 327.[40] S. Jester and H. Falcke, in prep. (2008).[41] F. D. Kahn and I. Lerche, Proc. Roy. Soc.,

1966, Vol. A-289, p. 206.[42] A. Karle et al., Nuclear Physics B Proceed-

ings Supplements, 2003, Vol. 118, p. 388.[43] A. Karle et al., Proceedings of the 30th ICRC,

Merida, Mexico, 2007, Vol. 4, p. 835.[44] N. E. Kassim et al., Exploring the Last Elec-

tromagnetic Frontier with the Long Wave-

length, Array (LWA), Long Wavelength As-trophysics, 26th meeting of the IAU, JointDiscussion 12, Prague, Czech Republic,2006,Vol. 56, p. 12.

[45] H. O. Klages, W.-D. Apel, and K. Bekk, Nu-clear Physics B Proceedings Supplements 52(1997) 92.

[46] I. Kravchenko et al., Phys. Rev. D 73(8)(2006) 082002.

[47] S. Lafebre et al., Proceedings of the 30thICRC, Merida, Mexico, 2007, Vol. 4, p. 577.

[48] N. G. Lehtinen et al., Phys. Rev. D 69(1)(2004) 013008.

[49] Q. Luo, MNRAS 370 (2006) 2071.[50] N. Meyer-Vernet, A. Lecacheux, and D. Ar-

douin, Radio pulses from cosmic ray airshowers - Boosted Coulomb and Cherenkovfields, arXiv:712[astro-ph]

[51] R. Milincic et al., Journal of Physics Confer-ence Series 81 (2007) 2007.

[52] R. Nahnhauer and S. Boser,Acoustic and Ra-dio EeV Neutrino Detection Activities(WorldScientific Publisher, 2006).

[53] G. Navarra et al., Nuclear Instruments andMethods in Physics Research A 518 (2004)207.

[54] S. Nehls et al., Nucl. Instr. Methods A sub-mitted (2008).

[55] A. Nigl et al. (LOPES Collaboration), A&A487 (2008) 781.

[56] A. Nigl et al. (LOPES Collaboration), A&A488 (2008) 807.

[57] O. Ravel et al., Nuclear Instruments andMethods in Physics Research A 518 (2004)213.

[58] H. J. A. Rottgering et al.,LOFAR - Open-ing up a new window on the Universe,arXiv:0610596[astro-ph].

[59] D. Saltzberg and P. Gorham,Radio Detectionof High Energy Particles(American Instituteof Physics Conference Series, 2001).

[60] D. Saltzberg et al., Physical Review Letters86 (2001) 2802.

[61] O. Scholten et al., Astroparticle Physics 26(2006) 219.

[62] O. Scholten et al. (NuMoon collaboration),Proceedings of the 30th ICRC, Merida,Mexico, 2007, Vol.4 , p. 275.

[63] O. Scholten, K. Werner, and F. Rusydi,A

92


Macroscopic Description of Coherent Geo-Magnetic Radiation from Cosmic Ray AirShowers, arXiv:709:0709.2872[astro-ph].

[64] G. Swarup et al., Current Science 60 (1991)95.

[65] A. van den Berg, et al. (Auger collabora-tion), Proceedings of the 30th ICRC, Merida,Mexico, 2007, Vol. 5, p. 885.

[66] A.A. Watson, Highlights from the PierreAuger Obseervatory - the birth of the hybridera, arXiv:0801.2321[astro-ph].

[67] T. C. Weekes,Radio Pulses from Cosmic RayAir Showers( D. Saltzberg and P. Gorham,editors, Radio Detection of High Energy Par-ticles, volume 579, American Institute ofPhysics Conference Series, 2001), p. 3.

[68] K. Werner and O. Scholten,MacroscopicTreatment of Radio Emission from CosmicRay Air Showers based on Shower Simula-tions, arXiv:0712.2517[astro-ph].

[69] E. Zas, F. Halzen, and T. Stanev, Phys. Rev.D 45 (1992) 362.

93

Radio Detection of Ultra-High Energy Cosmic Rays · RADIO DETECTION OF UHECR discovery of CR...

Documents

Transcript of Radio Detection of Ultra-High Energy Cosmic Rays · RADIO DETECTION OF UHECR discovery of CR...