Performance of two Askaryan Radio Array stations and first results in the search for ultrahigh energy neutrinos

Ultrahigh energy neutrinos are interesting messenger particles since, if detected, they can transmit exclusive information about ultrahigh energy processes in the Universe. These particles, with energies above ${10}^{16}\mathrm{eV}$, interact very rarely. Therefore, detectors that instrument several gigatons of matter are needed to discover them. The ARA detector is currently being constructed at the South Pole. It is designed to use the Askaryan effect, the emission of radio waves from neutrino-induced cascades in the South Pole ice, to detect neutrino interactions at very high energies. With antennas distributed among 37 widely separated stations in the ice, such interactions can be observed in a volume of several hundred cubic kilometers. Currently three deep ARA stations are deployed in the ice, of which two have been taking data since the beginning of 2013. In this article, the ARA detector “as built” and calibrations are described. Data reduction methods used to distinguish the rare radio signals from overwhelming backgrounds of thermal and anthropogenic origin are presented. Using data from only two stations over a short exposure time of 10 months, a neutrino flux limit of $1.5\times {10}^{-6}\mathrm{GeV}/{\mathrm{cm}}^2/\mathrm{s}/\mathrm{sr}$ is calculated for a particle energy of ${10}^{18}\mathrm{eV}$, which offers promise for the full ARA detector.

Figure 1

An area map of the planned ARA detector at the South Pole. The stations are indicated by the black circles. Red filled circles denote the currently deployed stations.

Figure 2

The baseline design of an ARA station with a zoom illustrating the string details and a view of the deployed antennas of both polarization.

Figure 3

(Left) The components of the down-hole signal chain on each string in the ARA stations. (Right) The surface data acquisition system of the ARA stations, showing the most important components. Components framed in yellow are common to all strings.

Figure 4

The cumulative live time of stations A2 and A3 in 2013. Horizontal line segments indicate extended down times of the detectors.

Figure 5

Signals emitted by a Vpol (a) and an Hpol (b) pulser, as recorded by station A3 string D3. As is evident, the polarization separation is very clean and only the antennas of the emitted polarization show a response.

Figure 6

The rms of events recorded by four selected measurement channels from A2 in 2013, plotted in mV versus the day of 2013. The large variation observed in channel D4BH is indicative of failure of that particular channel.

Figure 7

Results of the fits for geometrical and timing calibration of the strings in station A3. Note that string D3 has been used as the reference and is fixed in the ${\chi }^2$-minimization of this calibration.

Figure 8

The residual of the rooftop pulser reconstruction with A3 (a) before and (b) after the geometrical calibration. The bimodal distributions contain noise events with a residual of roughly $10^{-1.5}$ and signal with lower residuals which migrate to smaller values after calibration.

Figure 9

The directional rooftop pulser reconstruction with A3 (a) before and (b) after the geometrical calibration. The axes show the difference between the reconstructed and the true azimuth (x-axis) and zenith (y-axis) angle. Reconstruction quality criteria are only applied loosely. It should be noted that the data from the rooftop pulser are not part of the calibration data sample.

Figure 10

Power plotted versus frequency for a recorded Vpol signal from an external source (blue solid line), with a flat input spectrum between 150 and 1000 MHz. The cumulative power distribution shows that most of the signal power is recorded below 500 MHz. The green dashed line shows the recorded spectrum for the natural ambient noise without additional signal for comparison.

Figure 11

The relative geometry of the Vpol antennas to the D6 pulser in A3.

Figure 12

Sample measurements of the antenna’s directional gain at the available angles (see Fig. 11) for the two ARA stations and polarizations: (a) A3 Vpol, (b) A2 Hpol. The data are plotted on a linear scale versus frequency. For comparison the current status of the simulation is shown as a black solid line. String D4 on A3 was not operating during the time of the measurement. The black dashed line represents the lower limit on the signal gain used to derive the systematic error on the detector sensitivity.

Figure 13

The directional gain results versus reception angle from A2 and A3 for all Vpol antennas compared to the current simulation of the bottom antennas (green line) at different frequencies. All data are normalized to an isotropic directionality pattern. The three data points per antenna originate from the three available calibration sources used in this calibration. Gains around $-25°$ angles can only be measured in the A2 geometry, gains around $+25°$ only in A3.

Figure 14

The effective area of the two ARA stations as a function of neutrino energy.

Figure 15

(a) The hit pattern for a simulated event containing a signal waveform. The hits are indicated by the colored squares. Hits which correspond to the signal wavefront are marked by black dashed ellipses to underline their visible regular pattern. (b) A typical hit pattern for pure thermal noise.

Figure 16

(a) The hit time differences in histograms for each of the five geometrical groups for the event shown in Fig. 15. The pair groups of the histograms are schematically shown as inlay in the bottom plot. The quality parameter here $(QP)=1.6$. (b) The hit time difference histograms for the noise event shown in Fig. 15. In this second case the quality parameter equals 0.5.

Figure 17

The quality parameter $QP$ (solid line) compared to a simple count of all hits in a pattern as they appear in the example in Fig. 15 (dashed line) for simulated neutrinos with energies between ${10}^{16}$ and ${10}^{21}\mathrm{eV}$ (blue) and thermal noise events (red). All values are scaled to cumulatively reach 99% in both noise distributions at the same X value. The distributions are normalized to the total event count.

Figure 18

(a) The cross-correlation graph for a calibration pulser signal using waveforms measured in two different antennas. (b) The two recorded waveforms after a shift by the time of maximum correlation $dt=-78.4\mathrm{ns}$.

Figure 19

The dependence of the azimuthal reconstruction on the residual [Eq. (14)]. The y axis indicates the difference between the reconstructed angle and the true angle for 165000 simulated neutrino events with energies between ${10}^{16}$ and ${10}^{21}\mathrm{eV}$. Events with a high residual triggered the detector but do not contain strong enough signal to be properly reconstructed and are likely thermal in origin.

Figure 20

Simulated neutrino vertices reconstructed in the azimuth and zenith angle with all quality criteria applied. (a) The difference between the reconstructed and true azimuth. (b) The reconstructed zenith angle plotted as a function of the true zenith angle of each event.

Figure 21

Reconstruction of pulsers deployed in IceCube holes in the deep ice (a) with A2 and (b) with A3. The plots include data for the expected positions and the reconstructed positions with their standard deviation. The influence of systematic differences of a few degrees between the true and reconstructed angles on neutrino identification is negligible.

Figure 22

(a) The distribution of recorded data in the two main cut parameters. All data correlated to known radio source locations (calibration pulser or surface) by reconstruction are removed. Hence, the remaining events can be thermal noise, misreconstructed radio events or neutrinos. (b) The distribution of simulated signal (see Sec. 3) in the two main noise parameters.

Figure 23

Reconstructed events that passed the thermal noise and reconstruction quality cuts for (a) A2 and (b) A3. The black boxes indicate the angular cut regions around the calibration pulser positions and the black line indicates the surface cut. Events inside the squares and above the surface line are rejected.

Figure 24

Reconstructed events that passed the thermal noise and reconstruction quality cuts for A3 (a) in the austral winter and (b) in the austral summer when station activity is maximal. Note that due to improper tagging of calibration pulser events in the first months of winter, many such events entered the final data sample. This is however accounted for in the final angular cuts of the analysis.

Figure 25

Neutrino limits and sensitivities from various experiments including the 7.5 month data analysis of the two ARA stations described herein. Systematic errors, as derived in Sec. 6, have been accounted for in the ARA2 limit. The ARA37 (3 yr) sensitivity is projected from the ARA2 (7.5 m) trigger level without accounting for systematic errors ($K(E)=2.44$). Data for other experiments are taken from [7, 26, 45, 46, 47, 48]. The neutrino fluxes are derived in [37] and [49]. See Appendix pp2 for $EF(E)$ scaling on the y axis.

Figure 26

A South Pole map with the two ARA stations and the XY reconstruction of an A2/A3 coincident event series via the parallax method (green dots). The events in this series are distributed over a time of 50 s starting from the rightmost point in the map.

Figure 27

Comparison between the expected zenith from the A2/A3 XY parallax reconstruction (green dashed line) and the reconstructed zenith angle by A2 only (green solid line). Errors from the XY reconstruction, calculated from the errors on the reconstruction of a surface pulser with known position, are shown as a green band.

Figure 28

The relative difference in effective area at analysis level, caused by various systematic error sources.

Figure 29

The residual of reconstructed events as a function of the SNR for examples of (a) neutrino simulations in A3 and (b) calibration pulser signals recorded by A3.

Figure 30

Top panel: The analysis efficiency of the ARA stations for simulated events (green) and for calibration pulser events on A3 (blue). Bottom panel: The relative difference in efficiency between simulation and pulser data (red) and the normalized event count $N_i$ per SNR bin (orange). The shown simulated events are produced with a primary energy of ${10}^{18}\mathrm{eV}$.

Figure 31

Numbers of neutrinos versus energy in half-decade bins, as expected to be seen with the ARA37 detector within 3 years. These numbers have been extrapolated from the presented trigger and analysis efficiency for the first two stations. The numbers are calculated for three different flux predictions which are partly shown in Fig. 25 and estimated in [37, 49, 51]. The total expected numbers of neutrinos for the predictions are 9.4 (Ahlers 2010) and 5.5 (Kotera 2010).

Figure 32

A typical calibration waveform separated by even (blue) and odd (green) samples with a fit waveform (red). The horizontal lines indicate the range for individual sample correction in timing.

Figure 33

The timing corrections for all delay elements (x axis) of one channel in station A3, separated by an even (a) and odd (b) delay line after several iterations of calibration.

Figure 34

Calibration data for the calculation of the ADC-to-voltage conversion for a single storage sample. Collected data (red dots), averaged data (black points) with errors, and a broken third order polynomial fit to the average data (blue line).

Figure 35

Arrival time difference for a calibration pulser signal on two channels of station A2 after timing calibration (red) and the full calibration (blue).

Figure 36

Comparison of Fourier spectra for a simulated waveform (undisturbed in blue) at various error levels in the digitizer calibration. (a) Different magnitudes of timing jitter applied to each sample. The rms represents the width of the Gaussian distribution used to generate the random jitter for each sample. (b) Different levels of a nonlinear addition in the ADC-to-voltage conversion gain as presented in Eq. (a1). The rms represents the width of the distribution used to generate $k$.

Figure 37

The neutrino limits and fluxes of Fig. 25 with an alternative y-axis scaling.

Figure 38

Recorded waveforms for an A2 event, which was part of the event sequence presented in Fig. 26. The event shown corresponds specifically to the source at the point indicated by the red X (bottom right plot), as it moves above the array. The reconstructed zenith angle in A2 is 51.6°. Note that channel D4BH on A2 is not operational.

Figure 39

Recorded waveforms for an A2 event, which was part of the event sequence presented in Fig. 26. The event shown corresponds specifically to the source at the point indicated by the red X (bottom right plot), as it moves above the array. The reconstructed zenith angle in A2 is 87.0°.

Figure 40

Recorded waveforms for an A2 event, which was part of the event sequence presented in Fig. 26. The event shown corresponds specifically to the source at the point indicated by the red X (bottom right plot), as it moves above the array. The reconstructed zenith angle in A2 is 57.5°.