Microwave detection of air showers with the MIDAS experiment

Paolo Priviteraa ∗, I. Alekotteb, J. Alvarez-Muniz c, A. Berlina, X. Bertoub, M. Bogdana,M. Bohacovaa, C. Bonifazid , W. R. Carvalhoc, J. R. T. de Mello Netod, P. Facal San Luisa,J. F. Genata, N. Hollona, E. Millsa, M. Monasora, L. C. Reyesa, B. Rouille d’Orfeuila, E. M. Santosd,S. Waynea, C. Williamsa, E. Zasc

aUniversity of Chicago, Enrico Fermi Institute and Kavli Institute for Cosmological Physics,5640 South Ellis Avenue, Chicago, IL 60637, USA

bCentro Atomico Bariloche and Instituto Balseiro (CNEA-UNCuyo-CONICET),8400 San Carlos de Bariloche, Rıo Negro, Argentina

cUniversidad de Santiago de Compostela, Departamento de Fısica de Partıculas,Campus Sur, E-15782 Santiago de Compostela, Spain

dUniv. Federal do Rio de Janeiro (UFRJ), Instituto de Fısica,Cidade Universitaria, Caixa Postal 68528, 21945- 970 Rio de Janeiro, RJ, Brazil

Microwave emission from Extensive Air Showers could provide a novel technique for ultra-high energy cosmicrays detection over large area and with 100% duty cycle. We describe the design, performance and first resultsof the MIDAS (MIcrowave Detection of Air Showers) detector, a 4.5 m parabolic dish with 53 feeds in its focalplane, currently installed at the University of Chicago.

1. Introduction

Current experiments, like the Pierre Auger Ob-servatory [1] and the Telescope Array [2], employtwo well-established and complementary tech-niques to detect Ultra-High Energy Cosmic Rays(UHECR, energy ≥ 1018 eV): the Extensive AirShower (EAS) particles are sampled on their ar-rival at ground by a sparse array of detectors; atthe same time, the excitation of atmospheric ni-trogen by the EAS particles results in UV lightisotropically emitted along the shower path inthe atmosphere, which is imaged by large mir-rors onto an array of photomultiplier tubes. Verylarge areas, of thousands of km2, must be covereddue to the tiny rate of these extremely energeticparticles.

Fluorescence detectors have the unique capa-bility of performing a calorimetric measurementof the EAS energy and of determining the max-imum of the shower development in the atmo-

∗corresponding author

sphere, Xmax, a parameter sensitive to the massof the primary cosmic ray [3]. Also, a single tele-scope covers a large detection area, simplifyingthe installation and maintenance when comparedto the surface array. On the other hand, opera-tion of fluorescence detectors is limited to clearmoonless night, with an effective duty cycle of≈ 10%, and atmospheric attenuation of UV lightmust be continuously monitored by several ancil-lary instruments.

Recently, it has been suggested [4] that radi-ation in the microwave band of the spectrum islikely to be emitted as a by-product of the en-ergy released through ionization by the EAS. Apossible emission mechanism is bremsstrahlungradiation from interaction of the free electronswith the neutral molecules of the air plasma pro-duced by the EAS in the atmosphere. This ra-diation should be essentially isotropic and unpo-larized, allowing for a fluorescence-like detectionwith 100% duty cycle. Also, atmospheric atten-uation, one of the main sources of uncertainty

Nuclear Physics B (Proc. Suppl.) 212–213 (2011) 329–335

0920-5632/$ – see front matter © 2011 Published by Elsevier B.V.

www.elsevier.com/locate/npbps

doi:10.1016/j.nuclphysbps.2011.03.044

Figure 1. The MIDAS prototype installed on the roof of the Kersten Physics Teaching Center at theUniversity of Chicago: the 4.5 m parabolic antenna dish (left) with the 53 pixels camera at its focus(right).

in the fluorescence technique, is negligible in theGHz range. A first attempt of UHECR detectionby the AMBER prototype with a 1.8 m diameterparabolic dish antenna was reported in [4].

The development of microwave detection ofUHECR can have a major impact on the per-formance of existing experiments, by providing ameasurement of Xmax with 100% duty cycle, andcan offer the opportunity of a low-cost coverageof very large area. We have built the MIDAS(MIcrowave Detection of Air Showers) detectorwith the goal of confirming and characterizingisotropic GHz microwave emission (as the molecu-lar bremsstrahlung radiation) from the air plasmacolumn produced by EAS. Provided that the mi-crowave signal is as strong as measured by [4], the4.5 m diameter antenna of the MIDAS prototypehas sufficient sensitivity for standalone detectionof UHECR with a sizable aperture. MIDAS iscurrently installed at the University of Chicago

Campus (Figure 1).

2. The MIDAS prototype

The MIDAS design follows closely the detectionconcept of UV fluorescence detectors [5]: a largepatch of the sky is imaged by a reflector dish ontoa pixelized camera, with individual pixel field ofview of few deg2. Showers distant more than afew km from the detector will be seen in a pixelas a pulse of several hundreds of ns width, digi-tized by a Flash ADC every 50 ns. Shower candi-dates are then discriminated over accidental coin-cidences by a trigger system optimized for tran-sient events.

The MIDAS reflector is a 4.5 m diameterparabolic dish (f/D = 0.34) which focuses the in-coming microwave radiation onto a 53 pixel cam-era with 7 rows of 7 or 8 pixels (Figure 1). Pix-els are arranged in a staggered layout to maxi-

P. Privitera et al. / Nuclear Physics B (Proc. Suppl.) 212–213 (2011) 329–335330

Figure 2. Analog electronics enclosure with 8channels and the low-voltage distribution system.

mize the coverage. The pixel’s field of view de-pends on its position on the camera (the centralpixel having ≈ 1.3◦ × 1.3◦) due to aberrations,for a total camera field of view of ≈ 20◦ × 10◦.Pixels are instrumented with commercial satellitetelevision LNBF (LNB with feedhorn), which in-tegrates the antenna feedhorn with a low noiseblock converter (LNB). The feed operates in theC-band (3.4-4.2 GHz) and the LNB includes a lownoise amplifier (13 K noise level and 70 dB ampli-fication figure) and a frequency down-converter.The down-converted RF signal (950–1750 MHz)from the LNBF is brought by 30 meters of RG-6U coaxial cable into the counting room, whereis converted to a voltage proportional to the RFpower by a power detector. The power detectorprovides a voltage level between 2 V and 0.6 Vinversely proportional to the logarithm of the RFinput power, with a sensitivity as low as -60 dBm.With a time response of ≈ 10 ns, the power detec-tor is well suited for an accurate measurement ofthe μs pulse corresponding to the typical transitof the shower in the pixel’s field of view. The com-plete analog electronics chain includes a power in-serter (’bias-tee’) for the LNBF, a 75 Ω to 50 Ωimpedance adapter and the power detector. The

Figure 3. Illustration of the First Level Trigger:in black, digitized trace of a relative calibrationpulse (see section 3.1); superimposed, in blue, theADC running sum of 10 consecutive time bins;when the running sum is smaller than a thresh-old (horizontal line in red) a FLT for the pixel isissued.

53 analog channels are organized in groups ofeight in rack-mount electronics enclosures (Fig-ure 2) which include low voltage distribution forthe power detectors and the LNBFs.

The output signal of the power detector is dig-itized by a 20 MHz Flash-ADC with 14 bit reso-lution. The FADC board, developed by the Elec-tronics Design Group at the Enrico Fermi Insti-tute of the University of Chicago, hosts 16 chan-nels and has a VME interface. An on-board FieldProgrammable Gate Array (FPGA) is used fordigital signal processing and trigger.

2.1. Trigger and DAQThe MIDAS trigger, implemented in the FP-

GAs of the Flash-ADC boards, selects candidateevents by pixels topology and time coincidencerequirements. For each pixel, the signal is con-tinuously digitized and the running ADC sum of10 consecutive time bins is calculated. Wheneverthis sum falls below a preset threshold (notice

P. Privitera et al. / Nuclear Physics B (Proc. Suppl.) 212–213 (2011) 329–335 331

Figure 4. Distribution of one pixel FLT rate fora 6-hour run. The high-rate tail extending above400 Hz is due to noise bursts (see Section 2.2)

that due to the inverting characteristics of thepower detectors, lower ADC values correspond tohigher RF power intensities), a First Level Trig-ger (FLT) is issued for the pixel (Figure 3) anda 10 μs gate is opened. To account for changingRF background, the threshold is adjusted everysecond to keep the pixel FLT rate close to 100Hz (Figure 4). The Second Level Trigger (SLT)performs a search for pre-defined patterns of FLTtriggers whose gates overlap in time. Valid pat-terns correspond to the expected topology of acosmic ray shower (straight tracks across the cam-era). When a SLT is issued, a stream of 100 μsof ADC data (including 500 pre-trigger samples)is stored in memory for each of the 53 channels.

The SLT is implemented independently on theeach FADC board, and currently selects track-likepatterns of 3 pixels. In a future upgrade, the FLTtriggers will be multiplexed and sent during eachdigitizer cycle to a Master Trigger Board, wherea dedicated FPGA will implement a global SLTtrigger looking for track-like patterns of 4 pixels.

Once an SLT has been issued and the dataare available on memory, they are readout by the

VME master and written into disk for further pro-cessing. The trigger signal is also sent to a GPSunit that tags the precise timing of the event; thetime-stamp is read through VME and stored withthe data.

A second stream of data is recorded each sec-ond for monitoring purposes: the ADC baseline(averaged over 10 ms), the FLT rate and the FLTthreshold of each pixel.

The data acquisition software steers both theevent and the monitoring data readout. Dataruns are restarted periodically and data is backed-up on a server for offline analysis. Programs havebeen developed for the control and monitoringof the antenna, of the power supplies and VMEcrate, allowing fully automatic and remote oper-ation of the system.

2.2. BackgroundSince the expected rate of EAS is very low, the

data taking with the MIDAS prototype is dom-inated by background events. A fraction of thebackground consists of accidental coincidences ofFLT triggers due to thermal noise fluctuations.These events are characterized by small signalsbarely below the FLT threshold and by timingrandomly distributed within the FLT trigger win-dow, and thus not consistent with an EAS movingin the field of view of the camera. The thermalnoise rate is ≈0.2 Hz with the current 3-pixel SLTtrigger, and will become much smaller when a 4-pixel SLT trigger will be implemented.

Several sources of GHz RF interference arepresent in the urban environment of the Univer-sity of Chicago campus, including cellular phonetowers, motors of various kind, and most notablythe navigation system of airplanes overflying theMIDAS antenna on their route to Midway Air-port. The RF interference level may increase sud-denly, generating bursts of events during severalseconds. In order to protect the DAQ and to re-duce the impact of the noise bursts in the pixelFLT threshold regulation, the trigger system isinhibited for the duration of the bursts. Due tothe variability of the RF interference, the totaltrigger rate varies greatly, but in general is below1 Hz with a typical value of 0.5 Hz.

Background events from RF interference

P. Privitera et al. / Nuclear Physics B (Proc. Suppl.) 212–213 (2011) 329–335332

Figure 5. Top: ADC baseline of a pixel as a func-tion of time for a particularly noisy run. Bottom:another pixel of the same run with a band-passRF filter installed, which eliminates the bursts ofnoise.

sources are characterized by very short, simulta-neous pulses usually illuminating the whole cam-era, and are thus easily rejected in the offline anal-ysis of the data. We expect a significant improve-ment of the background conditions with the un-dergoing installation of band-pass RF filters de-signed to cut out the airplane transmission fre-quency (Figure 5).

3. Calibration and sensitivity

Several measurements have been performed tocalibrate the MIDAS prototype, and to establishits sensitivity, both during the detector commis-sioning, and periodically during data taking.

3.1. Relative calibrationA log-periodic antenna positioned at center of

the reflector has been used to measure the pixelsrelative calibration. The antenna, excited by anRF pulse generator, illuminates the whole camerawith a 4 GHz RF pulse of a few μs pulse width,and pulse power varied in 5 dB steps between −60and 0 dBm. An example of detected calibration

Figure 6. Relative calibration measurement:measured pixel ADC counts as a function of thepulse power from the calibration antenna. Thecurve is the result of the fit discussed in Sec-tion 3.1. Above -40 dBm, the LNBF saturatesand these data points are not included in the fit.

pulse is shown in Figure 3.The response of the pixel to a pulse of intensity

Ppulse applied to the external antenna, can beparametrized as follows:

nADC = nsys − k · log(1 + f · Ppulse

Psys)

where nsys and Psys are the ADC counts andpower without external signal (i.e. the powercorresponding only to the system+sky temper-ature); f relates the power of the pulse appliedto the antenna to the effective power received atthe feed (accounting for signal losses, distance,polarization, etc.); and k is the calibration con-stant of the channel in ADC/dB. Figure 6 showsthe data for one channel, together with a fit ofnsys, k and f/Psys in the expression above. Thedistribution of the calibration constants of the 53

P. Privitera et al. / Nuclear Physics B (Proc. Suppl.) 212–213 (2011) 329–335 333

Figure 7. Distribution of the calibration con-stants measured for the 53 pixels.

pixels is shown in Figure 7: the pixels have thesame calibration constant within a few %, quiteremarkable for commercial units. Since the RFcalibration pulse illuminates all the camera pix-els at the same time, we could also check thatpixels are synchronous to better than the 50 nssampling time.

A new antenna has been mounted in the centerof the dish in order to provide a fixed calibra-tion source. Regular relative calibration runs areforeseen. Also, the antenna will be used to moni-tor the stability of the system during data takingby firing a set of 10 pulses with fixed power andduration every 15 minutes.

3.2. Absolute calibrationThe sun is a powerful source of GHz radiation,

measured daily by several solar radio observato-ries around the world. The signal measured bythe pixel at the center of MIDAS camera dur-ing a transit of the sun in its field of view isshown in Figure 8. The flux of the sun at 4 GHz,

Figure 8. Absolute calibration measurement: thebaseline (10 ms average) of the central pixel as afunction of time, during a transit of the sun. Thesun crossed the field of view of the pixel around12000 seconds after the start of the data takingrun.

measured on the same day of the MIDAS cal-ibration by the Nobeyama observatory [6], wasFsun = 88 · 10−22W/m2/Hz. The flux of the sunprovides an absolute calibration of the ADC peakvalue measured by MIDAS. The sensitivity of thedetector, given as system temperature, can thenbe derived. In fact, the difference Δn in ADCcounts between the sun peak and the baseline be-fore the sun transit in Figure 8 can be expressedas:

Δn = k · log(1 + Fsun/Fsys) ,

where k has been defined in Section 3.1. The sys-tem noise flux density Fsys can be derived, sincek, Δn , and Fsun are known. The system tem-perature Tsys is then obtained from:

12Fsys · Aeff = kb · Tsys ,

where kb is the Boltzmann constant and the fac-tor 1/2 takes into account that MIDAS feeds mea-sure only one polarization component. Assumingan effective area Aeff � 10 m2, Tsys is found tobe � 120 K. The sun calibration has been per-formed routinely, with compatible Tsys results.Also, we performed additional calibration mea-surements with the moon (≈ 1/100 Fsun) and

P. Privitera et al. / Nuclear Physics B (Proc. Suppl.) 212–213 (2011) 329–335334

Figure 9. Event display of a Monte Carlo simu-lated event. Top left: FLT pixels with color in-dicating increasing trigger time from red to blue.Top right: FADC traces of selected pixels. Bot-tom left: FLT trigger gate integrated over all pix-els in a row as a function of time. Bottom right:ADC running sum for the selected pixels.

the Crab Nebula (≈ 1/1000 Fsun) which yieldedconsistent estimates.

3.3. Simulation of UHECR eventsThe measurements reported in [4] suggest that

the emitted GHz power scales quadratically withthe EAS energy, but a linear scaling is not ex-cluded. With a quadratic scaling and a systemtemperature of 120 K, a MIDAS pixel viewing theshower maximum of a 5 · 1018eV shower at 10 kmdistance would measure ≈ 2000 ADC counts overthe baseline. With a linear scaling, a 1019eVshower at the same distance would yield ≈ 200counts. These estimates can be compared withthe measured pixels baseline fluctuation of about70 ADC counts. Thus, the MIDAS prototype hasa good sensitivity for UHECR detection.

We have developed an end to end Monte Carlosimulation of the MIDAS prototype, including thecamera beam patterns and the absolute calibra-tion, which is being used for a realistic estimateof the event rate, and for the characterization ofthe expected events in the same format as the

data. An example of a simulated event is shownin Figure 9.

4. Outlook

Establishing microwave emission from EAS as aviable technique for UHECR detection is of greatinterest for the field. For this purpose, we havebuilt a large field of view MIDAS detector, whichhas been successfully installed and commissioned.During the initial months of its operation, wehave concentrated on the calibration of the de-tector, and on trigger and data taking stability.Several upgrades are being implemented to im-prove the quality of the data and the operationof MIDAS: band-pass RF filters in each channelto eliminate noise bursts, a global camera triggerand a fixed calibration antenna for system moni-toring.

Once stable operation and data taking isachieved, we plan to run for several months at theUniversity of Chicago. MIDAS has the sensitivityto detect cosmic ray showers or to place a limit onthe scaling of the laboratory measurement in [4]to atmospheric air showers. Installation of theprototype at the Southern site of the Auger Ob-servatory for coincident detection of UHECR isalso foreseen.

REFERENCES

1. J. Abraham et al. [Pierre Auger Collabora-tion], Nucl. Instrum. Meth. A 523 (2004) 50.

2. H. Tokuno et al., AIP Conf. Proc. 1238(2010) 365.

3. R. U. Abbasi et al. [The High Resolu-tion Fly’s Eye Collaboration], Phys. Rev.Lett. 104, 161101 (2010); J. Abraham etal. [Pierre Auger Observatory Collaboration],Phys. Rev. Lett. 104 (2010) 091101

4. P. W. Gorham et al., Phys. Rev. D 78 (2008)032007

5. R. M. Baltrusaitis et al., Nucl. Instrum. Meth.A 240 (1985) 410; J. A. Abraham et al. [ThePierre Auger Collaboration], Nucl. Instrum.Meth. A 620, 227 (2010)

6. Nakajima et al. PASJ 37 (1985) 163.

P. Privitera et al. / Nuclear Physics B (Proc. Suppl.) 212–213 (2011) 329–335 335