IceTop: IceCube Surface Air-Shower Array

• IceTop is a surface air-shower array at the South Pole that uses ice-Cherenkov tanks to detect cosmic-ray events and low-energy muons.
• It employs detailed reconstruction methods—using lateral distribution functions and timing models—to derive key observables like S125 and muon density for precise energy and mass composition analyses.
• The detector also functions as a surface veto for neutrino events and is evolving into a hybrid array with scintillation and radio upgrades to mitigate snow effects and extend its energy sensitivity.

IceTop is the surface air-shower array of the IceCube Neutrino Observatory at the geographic South Pole. In its completed configuration it comprises 81 stations, each with two ice-Cherenkov tanks, deployed over about $1\,\mathrm{km}^2$ directly above the deep in-ice detector. IceTop measures the electromagnetic component of extensive air showers together with low-energy surface muons, while the deep IceCube array records the penetrating high-energy muon bundle; the combined observatory is therefore a three-dimensional cosmic-ray detector and also a surface veto for neutrino analyses. Across its standard and low-energy analyses, IceTop has been used for cosmic-ray measurements from $250\,\mathrm{TeV}$ to the EeV range, while the broader detector design and event classes cover roughly $100\,\mathrm{TeV}$ to $1\,\mathrm{EeV}$ depending on trigger, containment, and reconstruction strategy (Collaboration et al., 2012, Kolanoski, 2012, Soldin, 2022).

1. Detector architecture and operating principle

IceTop is installed at an altitude of 2835 m above sea level, corresponding to an atmospheric depth of about $680\,\mathrm{g/cm^2}$. The array consists of 81 stations on an approximately triangular grid with average spacing of about 125 m, with a denser central infill region at below $50\,\mathrm{m}$ spacing for lower-energy showers. Each station contains two cylindrical tanks separated by 10 m; each tank contains clear ice to a depth of 0.90 m and two Digital Optical Modules operated at different gains, one high-gain and one low-gain, in order to extend dynamic range from single-muon calibration signals to large shower-core signals (Collaboration et al., 2012, Soldin, 2022).

The tanks detect Cherenkov light emitted by charged shower particles traversing the ice. Signals are calibrated in units of vertical equivalent muons (VEM). In the technical design paper, 1 VEM is defined as 95% of the muon peak position, and a vertically through-going muon typically produces about 125 photoelectrons in a standard high-gain DOM (Collaboration et al., 2012). This calibration anchors the surface signal scale used throughout IceTop reconstruction and allows tank charges to be compared across stations and epochs.

Local coincidence is fundamental to IceTop operation. Hard Local Coincidence (HLC) denotes the case in which both neighboring tanks in a station trigger within $1\,\mu\mathrm{s}$; Soft Local Coincidence (SLC) denotes the case in which one tank triggers but its partner does not. HLC hits dominate standard shower reconstruction near the shower core, while SLC hits become important at large lateral distance, where signals are weaker and often muon dominated (Gonzalez, 2015). A common simplification is to treat IceTop as only a surface trigger layer, but the literature assigns it simultaneous roles as a standalone air-shower array, a coincident surface partner to the in-ice detector, and a veto detector for atmospheric backgrounds (Kolanoski, 2012).

2. Reconstruction framework and principal observables

Standard IceTop reconstruction uses cleaned tank signals and simultaneously fits the charge and timing distributions. The canonical lateral distribution function is written as

$$S(r)=S_{125}\cdot \left(\frac{r}{125\,\mathrm{m}}\right)^{-\beta-\kappa\cdot \log_{10}(r/125\,\mathrm{m})},$$

with $S_{125}$ the fitted signal strength at 125 m from the shower axis, $\beta$ the slope parameter, and $250\,\mathrm{TeV}$0 fixed from simulation (Soldin, 2022). $250\,\mathrm{TeV}$1 is the fundamental IceTop shower-size observable and the standard energy proxy for high-energy spectrum analyses.

The timing distribution is fit with a curved shower-front model. In instrumentation-oriented descriptions this is expressed as

$250\,\mathrm{TeV}$2

with the curvature term parameterized as

$250\,\mathrm{TeV}$3

where $250\,\mathrm{TeV}$4, $250\,\mathrm{TeV}$5, and $250\,\mathrm{TeV}$6 (Collaboration et al., 2012). The fit returns shower direction, core position, and size.

Snow accumulation over the tanks modifies the detector response, especially for the electromagnetic component. A standard correction writes the expected no-snow signal as

$250\,\mathrm{TeV}$7

with $250\,\mathrm{TeV}$8 the slant depth through snow and $250\,\mathrm{TeV}$9 an effective attenuation length. Recent summaries state that $100\,\mathrm{TeV}$0 is year dependent and lies between 2.1 and 2.3 m (Kang, 2023). Snow is repeatedly identified as one of the dominant IceTop systematics.

Beyond the single-component $100\,\mathrm{TeV}$1 framework, newer reconstructions explicitly separate electromagnetic and muonic contributions. A two-component lateral-distribution approach writes

$100\,\mathrm{TeV}$2

using a Double Logarithmic Parabola for the electromagnetic part and a Greisen-based function for the muon part, with event-by-event estimators $100\,\mathrm{TeV}$3 and $100\,\mathrm{TeV}$4 (Weyrauch et al., 2023). A later reconstruction development combines a two-component surface fit with in-ice information and extends reconstruction to showers with cores outside the IceTop array up to zenith angles of $100\,\mathrm{TeV}$5; for uncontained events in the range $100\,\mathrm{TeV}$6–$100\,\mathrm{TeV}$7 GeV, the reported energy resolution is below $100\,\mathrm{TeV}$8 in $100\,\mathrm{TeV}$9 (Pyras et al., 11 Jul 2025). This marks a shift from an exclusively contained, single-LDF analysis toward explicitly hybrid EM-plus-muon inference.

3. Cosmic-ray spectrum and mass composition

IceTop’s all-particle spectrum program combines an IceTop-only high-energy analysis with lower-energy extensions using the central infill. In overview form, the surface array measures the all-particle spectrum from $1\,\mathrm{EeV}$0 to the EeV range, with the low-energy extension covering $1\,\mathrm{EeV}$1 to $1\,\mathrm{EeV}$2 and the standard high-energy analysis covering above $1\,\mathrm{EeV}$3 (Soldin, 2022). A major IT73 analysis reported that the spectrum is not described by a single power law above the knee: the knee is measured at about $1\,\mathrm{EeV}$4, the spectrum hardens again above about $1\,\mathrm{EeV}$5, and then shows a sharp drop beyond $1\,\mathrm{EeV}$6 (Tamburro, 2013). Later summary papers describe the same broad structure as a knee around 5 PeV and a second knee around 100 PeV (Kang, 2023).

IceTop’s composition sensitivity is strongest in coincident events with the deep detector. The underlying scaling is often summarized by

$1\,\mathrm{EeV}$7

with $1\,\mathrm{EeV}$8: at fixed primary energy, heavier primaries produce more muons (Soldin, 2022). IceTop constrains shower size and geometry at the surface, while IceCube contributes the high-energy muon bundle via observables derived from the reconstructed energy-loss profile $1\,\mathrm{EeV}$9 along the shower axis.

In the coincident mass-composition analysis summarized in recent reviews, the inputs are $680\,\mathrm{g/cm^2}$0, zenith angle, the reconstructed energy loss at a reference slant depth of 1500 m, and two measures of the number of large stochastic losses. These variables are fed into an artificial neural network trained on H, He, O, and Fe simulations; the resulting mass-output distributions are converted to template PDFs and fitted statistically in each energy bin (Soldin, 2022). The all-particle spectrum from the coincident analysis agrees with the independent IceTop-only spectrum, and the inferred $680\,\mathrm{g/cm^2}$1 increases with energy. Recent summaries describe the composition as becoming heavier with increasing energy up to about $680\,\mathrm{g/cm^2}$2, with elemental-group knees shifting to higher energy with increasing charge (Kang, 2023).

This combined surface-plus-depth method is one of IceTop’s defining scientific roles. It makes the observatory sensitive not only to the all-particle flux but also to the charge-dependent evolution of the Galactic component through the knee region and toward the expected Galactic-to-extragalactic transition.

4. Muon measurements and tests of hadronic interaction models

A central IceTop capability is the measurement of the low-energy muon content of air showers from the surface detector alone. The 2015 muon analysis isolates muons at large lateral distance from the shower axis, typically beyond about 300 m, where the electromagnetic component has thinned out and tank signals around $680\,\mathrm{g/cm^2}$3 become identifiable as one or more muons. In this regime SLC signals become especially important (Gonzalez, 2015). IceTop is sensitive here to low-energy muons with threshold roughly $680\,\mathrm{g/cm^2}$4 MeV, complementing IceCube’s in-ice sensitivity to muon bundles above about 300 GeV (Gonzalez, 2015).

The extracted muon densities are fit with a Greisen-like lateral distribution,

$680\,\mathrm{g/cm^2}$5

with $680\,\mathrm{g/cm^2}$6 m. The principal summary observable is $680\,\mathrm{g/cm^2}$7 (Gonzalez, 2015). For vertical events, its energy dependence is reported to be consistent with the Akeno scaling index of 0.83 and the absolute scale is of the same order as HiRes-MIA (Gonzalez, 2015).

Later work extended this toward event-by-event mass-sensitive parameters. One method defines a muon-enriched observable $680\,\mathrm{g/cm^2}$8 from charges above $680\,\mathrm{g/cm^2}$9 VEM in the 350–450 m radial slice and combines it with $50\,\mathrm{m}$0 into a normalized mass parameter $50\,\mathrm{m}$1, with $50\,\mathrm{m}$2 for proton-like showers and $50\,\mathrm{m}$3 for iron-like showers by construction (Kang et al., 2021). A separate two-component likelihood-based LDF reconstruction yields $50\,\mathrm{m}$4 as an energy estimator with better than 10% resolution above 10 PeV and $50\,\mathrm{m}$5 as an event-by-event low-energy muon proxy with resolution improving to below 20% at larger muon numbers (Weyrauch et al., 2023).

Muon observables also expose limitations of present hadronic interaction models. The IceTop overview on cosmic-ray measurements reports that measured muon densities are bracketed by pure-proton and pure-iron simulations, but the post-LHC models EPOS-LHC and QGSJet-II.04 predict larger muon densities than favored by composition expectations from other measurements (Soldin, 2022). Recent IceTop summaries add that three composition-sensitive observables—surface muon density, the IceTop lateral slope $50\,\mathrm{m}$6, and the in-ice $50\,\mathrm{m}$7 observable—are not mutually consistent under any tested hadronic interaction model (Kang, 2023). This is an objective formulation of the broader muon-model tension: IceTop and IceCube jointly probe GeV and TeV muons in the same shower, and the present model set does not describe all of those observables simultaneously.

5. Veto operation, hybrid measurements, and broader astrophysical uses

IceTop functions as an active surface veto for down-going cosmic-ray air showers in neutrino analyses. A five-year data-driven study constructs an IceTop log-likelihood ratio from tank charge, residual time, and lateral distance relative to reconstructed in-ice tracks, using all 162 tanks and explicitly incorporating both positive evidence of a shower and negative evidence from the absence of expected hits. The reported atmospheric-background reduction for a down-going muon-neutrino sample with minimum neutrino energy around $50\,\mathrm{m}$8 TeV is $50\,\mathrm{m}$9 to $1\,\mu\mathrm{s}$0 (Tosi et al., 2019). This establishes IceTop as more than a threshold trigger: it is a pattern-recognition veto operating below the standard surface-trigger threshold.

IceTop has also entered explicitly hybrid surface configurations. The "IceAct" demonstrator program places compact imaging air-Cherenkov telescopes inside the IceTop array, one on the roof of the IceCube Laboratory and one on the ice about $1\,\mu\mathrm{s}$1 away. Each telescope has a $1\,\mu\mathrm{s}$2 field of view, an energy threshold of about $1\,\mu\mathrm{s}$3, and can be synchronized with the IceCube/IceTop data stream. In the Roof configuration, an IceAct trigger can cause the event builder to save all coincident IceCube and IceTop hits, making IceTop part of the event-building chain rather than a purely offline cross-check (Paul et al., 2021). The hybrid concept adds atmospheric Cherenkov image morphology to IceTop’s surface particle densities and IceCube’s deep muon bundle measurement.

Beyond spectrum and composition, IceTop has also been used for directional and neutral-particle searches. A four-year neutron search above 10 PeV performed an all-sky search from $1\,\mu\mathrm{s}$4 to approximately $1\,\mu\mathrm{s}$5 in declination and a targeted search above 100 PeV for several Galactic source classes; no significant excess was found in either case (Collaboration et al., 2016). Earlier reviews also describe large-scale anisotropy measurements at median energies around 400 TeV and 2 PeV and PeV gamma-ray searches using the absence of in-ice muon activity as a hadron veto (Tamburro, 2012). These applications show that IceTop is not confined to standard all-particle reconstruction but also operates as a directional and multi-messenger surface detector.

6. Snow-driven upgrade path and the evolving surface array

The central long-term hardware issue for IceTop is snow accumulation over the tanks. The 2024 trigger study states that some tanks are now buried under about 5 m of snow and that the standard IceTop trigger rate dropped from about 30.2 Hz in 2012 to 7.6 Hz in 2023 (Paudel, 2024). Snow both raises threshold and attenuates the measured electromagnetic signal, so it is simultaneously an efficiency problem, a calibration problem, and a reconstruction problem.

The current response is the IceCube Surface Array Enhancement, a hybrid upgrade based on elevated scintillation panels and radio antennas. A 2024 status report describes the enhancement as a hybrid surface array of scintillation detectors and radio antennas intended to lower the energy threshold for air-shower measurements, provide more efficient veto capabilities, enable more accurate cosmic-ray measurements, and improve detector calibration by compensating for snow accumulation. After commissioning, the prototype station at the South Pole has been recording air-shower data and has successfully observed coincident events with the existing IceTop array; production and calibration for the full planned array are ongoing, and one station each has been installed at the Pierre Auger Observatory and the Telescope Array for further R&D (Shefali et al., 2024).

The prototype station deployed in January 2020 consists of 8 scintillation detectors and 3 radio antennas. It has already demonstrated coincident air-shower measurements in scintillators, radio, and IceTop, and the lowest $1\,\mu\mathrm{s}$6 observed for a radio event was $1\,\mu\mathrm{s}$7, corresponding roughly to $1\,\mu\mathrm{s}$8 (Dujmovic et al., 2021). The enhancement is therefore not only a snow-mitigation project but also a transition from a single-technology surface array to a genuinely hybrid one.

Trigger logic is being adapted accordingly. The standard IceTop Simple Majority Trigger, IceTopSMT, requires 3 coincident station hits within $1\,\mu\mathrm{s}$9, where each station hit requires both tanks in a station. The new IceTop7HG trigger instead requires at least 7 tanks with high-gain DOM hits within a $$S(r)=S_{125}\cdot \left(\frac{r}{125\,\mathrm{m}}\right)^{-\beta-\kappa\cdot \log_{10}(r/125\,\mathrm{m})},$$0 time window and imposes no spatial condition beyond multiplicity and timing. For inclined showers in the zenith range $$S(r)=S_{125}\cdot \left(\frac{r}{125\,\mathrm{m}}\right)^{-\beta-\kappa\cdot \log_{10}(r/125\,\mathrm{m})},$$1–$$S(r)=S_{125}\cdot \left(\frac{r}{125\,\mathrm{m}}\right)^{-\beta-\kappa\cdot \log_{10}(r/125\,\mathrm{m})},$$2, the full-efficiency threshold decreases from about 30 PeV for IceTopSMT to about 10 PeV for IceTop7HG; South Pole test runs in 2023 showed stable operation at about 32 Hz (Paudel, 2024). This is specifically tuned to preserve useful coincidences with the radio component, which is most sensitive in the inclined, $$S(r)=S_{125}\cdot \left(\frac{r}{125\,\mathrm{m}}\right)^{-\beta-\kappa\cdot \log_{10}(r/125\,\mathrm{m})},$$3 PeV regime.

The enhancement is also a pathfinder for the future IceCube-Gen2 surface array. A recent surface-enhancement summary describes a planned array of 32 stations, each with 8 scintillators and 3 antennas, controlled by hybrid DAQ systems. The same paper reports a preliminary estimation of $$S(r)=S_{125}\cdot \left(\frac{r}{125\,\mathrm{m}}\right)^{-\beta-\kappa\cdot \log_{10}(r/125\,\mathrm{m})},$$4 with data from the 3 antennas of the prototype station and states that Gen2 is expected to have a sixfold increase in surface area and about a factor of 30 increase in aperture when data from all detectors are included (Venugopal, 23 Jun 2025). The cumulative implication is straightforward: IceTop remains the surface anchor of IceCube, but its future is explicitly hybrid, with scintillation, radio, and atmospheric Cherenkov measurements incorporated to lower thresholds, stabilize calibration, improve veto efficiency, and add new composition-sensitive observables.