# Cosmic-ray spectrum and spectral features

The cosmic-ray energy spectrum describes how the flux of high-energy particles arriving at Earth falls with energy. Over roughly a dozen decades of energy it follows, to first approximation, a single power law, \(dN/dE \propto E^{-\gamma}\) with \(\gamma \approx 2.7\), onto which a small number of named deviations are superimposed: the knee, the second knee, the ankle, the instep, and the flux suppression.<sup>[1](https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf)</sup><sup> • </sup><sup>[2](https://arxiv.org/html/2508.21692)</sup> This article covers the measured all-particle spectrum and these features; the interpretation in terms of sources and propagation is treated in the sibling articles on acceleration and propagation.

| Key fact | Value |
|---|---|
| Spectral index below the knee | γ ≈ 2.7<sup>[1](https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf)</sup> |
| Knee | \(3\text{–}5 \times 10^{15}\) eV; index steepens from ~2.7 to ~3.1<sup>[2](https://arxiv.org/html/2508.21692)</sup> |
| Second knee | ~\(10^{17}\)–\(10^{17.7}\) eV (location contested); index to ~3.3<sup>[1](https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf)</sup><sup> • </sup><sup>[3](https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8)</sup> |
| Ankle | ~\(4\text{–}5 \cdot 10^{18}\) eV; index flattens to ~2.5<sup>[1](https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf)</sup><sup> • </sup><sup>[3](https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8)</sup> |
| Flux suppression | ~4.7–5 × \(10^{19}\) eV; index steepens to ~5.1<sup>[3](https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8)</sup><sup> • </sup><sup>[4](https://link.aps.org/doi/10.1103/PhysRevLett.125.121106)</sup> |
| Arrival rates | ~1 per m² per s at \(10^{11}\) eV; 1 per m² per yr at \(10^{15}\) eV; ~1 per km² per yr at \(10^{18}\) eV<sup>[5](https://link.springer.com/article/10.1140/epjs/s11734-025-01501-6)</sup> |
| Direct-measurement limit | Balloons and satellites reach only ~\(10^{14}\) eV<sup>[2](https://arxiv.org/html/2508.21692)</sup> |

## The spectrum at a glance

The flux follows a power law over a wide energy range, a signature of non-thermal acceleration processes.<sup>[6](https://ar5iv.labs.arxiv.org/html/0904.0725)</sup> Because the index is about 2.7, the flux falls steeply with energy, and the spectrum spans many decades in energy.<sup>[2](https://arxiv.org/html/2508.21692)</sup> In practical terms, about one particle per second crosses a square meter at \(10^{11}\) eV (100 GeV), the rate falls to one per square meter per year at 1 PeV, and to roughly one per square kilometer per year at 1 EeV.<sup>[5](https://link.springer.com/article/10.1140/epjs/s11734-025-01501-6)</sup>

Four named features interrupt the power law. The knee is a steepening at a few PeV, where the index changes from about 2.7 to about 3 (about 3.1 in the 2025 review literature).<sup>[1](https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf)</sup><sup> • </sup><sup>[2](https://arxiv.org/html/2508.21692)</sup> The second knee is a further steepening, to an index of about 3.3, at around 100 PeV according to the PDG compilation, though its exact location and even its sharpness are debated (see below).<sup>[1](https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf)</sup><sup> • </sup><sup>[2](https://arxiv.org/html/2508.21692)</sup> The ankle is a hardening at a few EeV, where the index flattens to about 2.5.<sup>[1](https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf)</sup> Finally, above roughly \(5 \times 10^{19}\) eV the spectrum suppresses sharply.<sup>[3](https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8)</sup>

## How the spectrum is measured

[Cosmic rays](https://www.edgepedia.org/cosmic-rays) up to about \(10^{14}\) eV (100 TeV) can be measured directly by balloon- or satellite-borne instruments, which are more accurate because the particle is detected before it interacts in the atmosphere.<sup>[2](https://arxiv.org/html/2508.21692)</sup><sup> • </sup><sup>[5](https://link.springer.com/article/10.1140/epjs/s11734-025-01501-6)</sup> Above that energy the flux is too low for practical direct detectors, so experiments measure extensive air showers, the particle cascades initiated when a cosmic ray hits the atmosphere, using ground-based arrays.<sup>[2](https://arxiv.org/html/2508.21692)</sup> The knee region, roughly \(10^{14}\)–\(10^{18}\) eV, is covered by arrays such as [KASCADE-Grande](https://www.edgepedia.org/kascade-grande), GRAPES-3, Tibet ASgamma, ARGO-YBJ, HAWC, IceCube/IceTop, and Tunka, with LHAASO, SWGO, TAIGA, Telescope Array, and the [Pierre Auger Observatory](https://www.edgepedia.org/pierre-auger-observatory) extending the reach to \(10^{20}\) eV.<sup>[5](https://link.springer.com/article/10.1140/epjs/s11734-025-01501-6)</sup>

Because each experiment has its own energy scale, cross-calibration is essential. The Global Spline Fit (GSF) approach treats each experiment's energy-scale offset as a nuisance parameter: shifting the energy axis by a factor \(1 + z_E\) shifts the flux by \((1 + z_E)^{-1}\), allowing all data sets to be fitted simultaneously. Recent DAMPE, ISS-CREAM, and LHAASO data now constrain the proton spectrum from the lowest rigidities to the highest energies observed.<sup>[7](https://doi.org/10.22323/1.501.0248)</sup>

## The features in detail

**The knee.** The steepening from \(\gamma \cong 2.7\) to \(\gamma \cong 3.1\) at about \(3 \times 10^{15}\) eV was first deduced from shower-size-spectrum observations by Kulikov and Khristiansen et al. in 1956.<sup>[8](https://ar5iv.labs.arxiv.org/html/astro-ph/0611884)</sup> Modern reviews place it at \(3\text{–}5 \times 10^{15}\) eV; IceCube/IceTop three-year data and the GSF 2024 model favor a knee near 5 PeV, slightly above the ~4 PeV commonly quoted.<sup>[2](https://arxiv.org/html/2508.21692)</sup><sup> • </sup><sup>[9](https://doi.org/10.22323/1.484.0087)</sup>

**The second knee.** The PDG review puts a second steepening, to γ ≈ 3.3, at about 100 PeV.<sup>[1](https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf)</sup> Other determinations differ: Auger measurements place it at \(1.24 \times 10^{17}\,\mathrm{eV}\), a re-analysis of Yakutsk 1974–2004 data supports \((6 \pm 2) \times 10^{17}\,\mathrm{eV}\), and a 2026 A&A analysis uses about \(5 \times 10^{17}\,\mathrm{eV}\).<sup>[3](https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8)</sup><sup> • </sup><sup>[8](https://ar5iv.labs.arxiv.org/html/astro-ph/0611884)</sup><sup> • </sup><sup>[10](https://www.aanda.org/articles/aa/full_html/2026/02/aa50483-24/aa50483-24.html)</sup> Auger data reaching down to ~100 PeV suggest the second knee is not a sharp feature but is linked to a softening of the heavy-primary spectrum beginning around \(10^{17}\,\mathrm{eV}\) seen by KASCADE-Grande.<sup>[2](https://arxiv.org/html/2508.21692)</sup>

**The ankle.** The spectrum flattens at a few EeV, where γ changes to about 2.5.<sup>[1](https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf)</sup> The ankle at ~\(3 \cdot 10^{18}\) eV was first observed by Haverah Park, Akeno, and Yakutsk.<sup>[8](https://ar5iv.labs.arxiv.org/html/astro-ph/0611884)</sup> Auger's own fit gives \(4.9 \cdot 10^{18}\) eV,<sup>[3](https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8)</sup> while GSF 2024 locates the ankle around 8 EeV.<sup>[9](https://doi.org/10.22323/1.484.0087)</sup>

**Instep and suppression.** Auger's high-statistics measurement of 215,030 events above \(2.5 \times 10^{18}\) eV found that at about \(1.3 \times 10^{19}\) eV the index changes from \(2.51 \pm 0.03(\mathrm{stat}) \pm 0.05(\mathrm{syst})\) to \(3.05 \pm 0.05(\mathrm{stat}) \pm 0.10(\mathrm{syst})\), a feature called the instep, and steepens further to \(5.1 \pm 0.3(\mathrm{stat}) \pm 0.1(\mathrm{syst})\) beyond \(5 \times 10^{19}\) eV.<sup>[4](https://link.aps.org/doi/10.1103/PhysRevLett.125.121106)</sup> A 2025 review places the suppression at ~\(4.7 \times 10^{19}\) eV.<sup>[3](https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8)</sup>

## How experiments compare

Direct and indirect measurements have agreed fairly well over the last two decades when total statistical and systematic errors and energy-scale uncertainties are considered.<sup>[3](https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8)</sup> At the highest energies, the Pierre Auger Observatory and Telescope Array spectra agree within 4% below ~10 EeV, with larger disagreement near the end of the instep and at the suppression.<sup>[1](https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf)</sup> KASCADE-Grande reports an up to 20% lower absolute flux than other measurements, possibly because of its sea-level location and its energy normalization via hadronic interaction models; at ~\(10^{18}\) eV it is statistically consistent with Auger.<sup>[2](https://arxiv.org/html/2508.21692)</sup> The TALE measurement is systematically lower than the three-year IceTop measurement, probably due to energy-scale uncertainties of the differing detection techniques.<sup>[3](https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8)</sup>

Between the knee and the ankle, discrepancies among IceCube, Auger, and KASCADE-Grande persist largely unchanged since the 2017 GSF analysis; the GSF collaboration attributes them to the differing hadronic interaction models used to interpret air-shower data.<sup>[7](https://doi.org/10.22323/1.501.0248)</sup> These models, which describe how the first collisions in an air shower behave, are the source of these attributed discrepancies.

## Element-group structure and rigidity scaling

Direct measurements give fitted spectral indices of p: −2.66 ± 0.02, He: −2.58 ± 0.02, C: −2.61 ± 0.07, O: −2.67 ± 0.07, Ne: −2.72 ± 0.10, Mg: −2.66 ± 0.08, Si: −2.67 ± 0.08, and Fe: −2.63 ± 0.11, compatible within uncertainties for He through Fe, which suggests a common origin; the proton–helium difference is significant only at a modest level.<sup>[11](https://iopscience.iop.org/article/10.1088/1742-6596/2429/1/012001/pdf)</sup>

At the knee itself, the picture is that the feature arises from sequential breaks in the spectra of individual elements, starting with the light elements. A rigidity dependence of the cut-off energies (a cutoff at the same \(E/Z\) for each species) is likely but not yet established beyond doubt.<sup>[8](https://ar5iv.labs.arxiv.org/html/astro-ph/0611884)</sup> Supporting this, KASCADE-Grande observed a knee-like feature in heavy primaries at \(10^{16.92}\,\mathrm{eV}\), and an ankle-like hardening in the light primaries where the slope changes from \(-3.25\) to \(-2.79\) at \(10^{17.08}\,\mathrm{eV}\).<sup>[2](https://arxiv.org/html/2508.21692)</sup> IceCube/IceTop three-year data confirmed a knee around 5 PeV and a second knee around 100 PeV, with the heavy-component spectrum showing a knee-like structure around \(10^{16.7}\,\mathrm{eV}\) where the index changes from \(-2.7\) to \(-3.3\).<sup>[2](https://arxiv.org/html/2508.21692)</sup> A 2026 A&A study finds the primary knee at \(\sim 4 \times 10^{15}\,\mathrm{eV}\) consistent with a constant-rigidity cutoff according to KASCADE-Grande data processed with post-LHC hadronic models, but not according to other datasets, and concludes the second knee at \(\sim 5 \times 10^{17}\,\mathrm{eV}\) requires more complexity than the cutoff of a single source population.<sup>[10](https://www.aanda.org/articles/aa/full_html/2026/02/aa50483-24/aa50483-24.html)</sup> In the GSF fit, the oxygen and iron groups near knee energies remain the least certain components.<sup>[7](https://doi.org/10.22323/1.501.0248)</sup>

## Recent developments

Several measurements have refined the spectrum in the last few years. LHAASO reports a hardening at ~200 GV across most of the primary cosmic-ray spectra, a softening at ~10 TV in protons and helium, and a subsequent hardening at ~100 TeV in the combined proton and helium spectrum (Alemanno et al. 2024);<sup>[12](https://iopscience.iop.org/article/10.3847/1538-4357/ada426)</sup> the HAWC all-particle spectrum from 10 to 500 TeV shows a softening at approximately 45 TeV.<sup>[3](https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8)</sup> Ground-based work continues to validate the knee at roughly 4 PeV, the ankle around 5 EeV, and suppression at 40 EeV, with additional features including a low-energy ankle at \(2 \times 10^{16}\) eV and a second knee around \(10^{17}\) eV.<sup>[12](https://iopscience.iop.org/article/10.3847/1538-4357/ada426)</sup>

The GSF 2024 model, which lists a dip near 5 TeV, knee near 5 PeV, low-energy ankle around 10 PeV, second knee at about 100 PeV, ankle around 8 EeV, instep at roughly 20 EeV, and toe near 50 EeV, finds that above ~10 TeV the all-particle flux is higher than in GSF 2019, mainly because of newly included sub-PeV data; updated Auger data also lead to a steeper drop in proton flux below the ankle.<sup>[9](https://doi.org/10.22323/1.484.0087)</sup> The 2025 Auger spectrum above 2.5 EeV is parameterized with a fall-off energy E3 and four spectral indices.<sup>[13](https://arxiv.org/pdf/2507.08573)</sup>

## References

1. Cosmic Rays — Review of Particle Physics (2026). https://pdg.lbl.gov/2026/reviews/rpp2026-rev-cosmic-rays.pdf
2. The cosmic-ray spectrum in the PeV to EeV energy range (2025). https://arxiv.org/html/2508.21692
3. Cosmic ray energy and composition measurements with GRAPES-3 and other experiments, Eur. Phys. J. ST (2025). https://link.springer.com/article/10.1140/epjs/s11734-025-01707-8
4. Features of the Energy Spectrum of Cosmic Rays above 2.5 EeV Using the Pierre Auger Observatory, Phys. Rev. Lett. 125, 121106. https://link.aps.org/doi/10.1103/PhysRevLett.125.121106
5. Cosmic ray sources and detectors, Eur. Phys. J. Special Topics. https://link.springer.com/article/10.1140/epjs/s11734-025-01501-6
6. Blümer, Kampert & Wischczenski, Cosmic Rays from the Knee to the Highest Energies, Prog. Part. Nucl. Phys. https://ar5iv.labs.arxiv.org/html/0904.0725
7. Global Spline Fit GSF-2025, ICRC 2025 proceedings. https://doi.org/10.22323/1.501.0248
8. Cosmic Rays from the Knee to the Ankle – Status and Prospects. https://ar5iv.labs.arxiv.org/html/astro-ph/0611884
9. Global Spline Fit (GSF) 2024, ICRC 2024 proceedings. https://doi.org/10.22323/1.484.0087
10. What can cosmic-ray knees reveal about source populations?, A&A (2026). https://www.aanda.org/articles/aa/full_html/2026/02/aa50483-24/aa50483-24.html
11. Galactic cosmic rays: direct measurements, status, J. Phys. Conf. Ser. 2429, 012001. https://iopscience.iop.org/article/10.1088/1742-6596/2429/1/012001/pdf
12. Precise Measurement of the Cosmic-Ray Spectrum and 〈ln A〉 by LHAASO, ApJ. https://iopscience.iop.org/article/10.3847/1538-4357/ada426
13. Pierre Auger Observatory spectrum above 2.5 EeV (2025). https://arxiv.org/pdf/2507.08573

---
*Topic: Encyclopedia › Physical world and mathematics › Physics › Particles and nuclei › Astroparticle physics › Cosmic rays › Cosmic ray overview and phenomenology › Cosmic-ray spectrum and spectral features*

*Initially written Sep 17, 2026 · Reviewed: Sep 30, 2026 · Edited: Sep 30, 2026; Oct 11, 2026; Oct 15, 2026; Oct 21, 2026; Oct 25, 2026 · Last review: Sep 30, 2026*

*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*

License: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license