Cosmic Rays: Origin and Spectrum

Cosmic rays are the most energetic particles known in nature: a single proton at $10^{20}$ eV carries about $16$ J — the kinetic energy of a well-struck tennis ball packed into one subatomic particle. They arrive across twelve decades of energy, from below $10^9$ eV to above $10^{20}$ eV, yet because they are charged they are bent by Galactic and intergalactic fields and reach us nearly isotropically. Their sources are hidden from ordinary imaging, and the whole science rests on three surrogate observables: the shape of the spectrum, the elemental and isotopic composition, and the secondary particles stamped on the beam as it propagates. The spectrum is a nearly featureless power law, $\mathrm{d}N/\mathrm{d}E \propto E^{-\alpha}$ , broken only by two soft kinks — the "knee" near $3\times10^{15}$ eV where it steepens, and the "ankle" near $4\times10^{18}$ eV where it flattens — before a sharp suppression above $\sim 5\times10^{19}$ eV. Reading those features turns on three ideas. First, a strong collisionless shock accelerates particles to a universal $E^{-2}$ spectrum through repeated crossings — first-order Fermi acceleration — with an index fixed by the compression ratio alone. Second, energy-dependent diffusive escape from the Galaxy steepens that source spectrum to the observed $\sim E^{-2.7}$ . Third, the finite size and field of any accelerator (the Hillas criterion) and losses on the cosmic microwave background (the GZK cutoff) sculpt the extreme high-energy end, where the sources must finally be extragalactic.

Key takeaways

Cosmic rays are charged particles — ~90% protons, ~9% helium, ~1% heavier nuclei and electrons — from below $10^9$ to above $10^{20}$ eV, arriving nearly isotropically as magnetic fields scramble their directions.
The all-particle spectrum is close to a power law $\mathrm{d}N/\mathrm{d}E \propto E^{-\alpha}$ , with $\alpha \approx 2.7$ below the knee at $\sim 3\times10^{15}$ eV, steepening to $\approx 3.1$ , then flattening at the ankle near $\sim 4\times10^{18}$ eV.
Galactic cosmic rays up to the knee are accelerated by diffusive shock acceleration at supernova-remnant shocks, giving a source spectrum $\mathrm{d}N/\mathrm{d}E \propto E^{-2}$ set only by the compression ratio $r\to 4$ .
Propagation steepens it: energy-dependent escape adds the diffusion index $\delta$ so the observed slope is $2+\delta$ , and boron/carbon ratios fix the confinement time at $\sim 10^7$ yr.
The highest-energy rays ( $\gtrsim 10^{18}$ eV) are extragalactic; above the GZK threshold $\sim 5\times10^{19}$ eV protons lose energy to the CMB by pion photoproduction, suppressing the flux.
Key idea: A strong collisionless shock accelerates particles to a universal $\propto E^{-2}$ power law via first-order Fermi acceleration; energy-dependent diffusive escape during propagation steepens this to the observed $\propto E^{-2.7}$ , while the finite size and field of any accelerator (Hillas criterion) and losses on the CMB (GZK) shape the spectrum's high-energy features.

Where this lives today

Cosmic-ray physics is now a mature multi-messenger enterprise. At the Galactic scale, gamma-ray telescopes — Fermi-LAT and the Cherenkov arrays H.E.S.S., MAGIC, VERITAS — image remnants such as RX J1713.7−3946 and IC 443; the "pion bump" in IC 443's low-energy spectrum proves protons are accelerated there, confirming DSA up to the knee. In space, AMS-02 and the Voyager probes (now beyond the heliopause) measure boron/carbon and other secondary ratios to percent precision, pinning $\delta$ and the confinement time. At the top end, the Pierre Auger Observatory (3000 km²) and Telescope Array measure the spectrum, composition, and arrival directions above $10^{18}$ eV; both see the GZK-like suppression near $4\times10^{19}$ eV, and Auger's large-scale dipole above $8\times10^{18}$ eV is the first solid evidence these particles are extragalactic. The frontier is multi-messenger: IceCube's neutrinos and the flare from blazar TXS 0506+056 tie hadronic acceleration to relativistic jets.

The physical picture

Picture a ping-pong ball between two walls slowly closing in: every bounce off an approaching wall speeds it up. A shock front is exactly this pair of converging walls, in momentum space. Turbulent magnetic clouds frozen into the upstream and downstream flows move toward each other across the front, so a particle scattering back and forth always meets an approaching scatterer and gains energy. Since it can also be swept downstream and lost each cycle, only a fraction survive to be boosted again — and the competition between a small gain and a small loss per cycle makes a power law, not a peaked bump.

Figure

A collisionless shock compresses the upstream flow by r = u₁/u₂ → 4 for a strong shock. A particle scattering across the front gains ⟨ΔE/E⟩ = (4/3)(u₁ − u₂)/c per cycle, and gain balanced against downstream escape fixes the source spectrum dN/dE ∝ E⁻²

The near-universal $E^{-2}$ index is the deep surprise: it depends only on the compression ratio, which for any strong shock in a monatomic gas is $\approx 4$ — regardless of density, field, or shock speed. Messy plasma astrophysics hands us a clean, parameter-free prediction. What we measure is steeper, $\sim E^{-2.7}$ , because the Galaxy is a leaky box: rays injected with the $E^{-2}$ shape diffuse around, but faster ones escape sooner, steepening the confined spectrum by $\delta$ . The second figure is the observable itself — the broken power law whose kinks mark where Galactic accelerators run out and the extragalactic sky takes over.

Figure

The all-particle cosmic-ray spectrum from 10⁹ to 10²⁰ eV is a nearly straight power law, steepening at the knee near 3×10¹⁵ eV from a slope of about 2.7 to 3.1, flattening at the ankle near 4×10¹⁸ eV, then cut off by the GZK suppression above 5×10¹⁹ eV

The core equations

The central object is the differential energy spectrum: how many particles arrive per unit energy. Over most of the range it is a clean power law whose two soft breaks are its grammar.

\frac{\mathrm{d}N}{\mathrm{d}E} = K\,E^{-\alpha}, \qquad \alpha \approx \begin{cases} 2.7 & E \lesssim 3\times10^{15}\,\mathrm{eV}\\ 3.1 & 3\times10^{15} \lesssim E \lesssim 4\times10^{18}\,\mathrm{eV}\\ 2.7 & E \gtrsim 4\times10^{18}\,\mathrm{eV}\end{cases} \qquad (1)

where: $\mathrm{d}N/\mathrm{d}E$ — differential number flux per unit energy; $K$ — normalization; $\alpha$ — local spectral index; $E$ — particle energy. The knee is where $\alpha$ steepens (Galactic accelerators fading) and the ankle where it flattens (a harder extragalactic component emerging). The spectrum integrates to a modest but dynamically important energy density, comparable to starlight, the magnetic field, and turbulent gas — a full partner in the interstellar pressure budget.

u_{\rm CR} \approx 1\ \mathrm{eV\,cm^{-3}} \approx 1.6\times10^{-12}\ \mathrm{erg\,cm^{-3}} \qquad (2)

where: $u_{\rm CR}$ — cosmic-ray energy density near the Sun. Reading it: this is the target any acceleration theory must feed against escape — the power budget of Worked Example 2. The engine is first-order Fermi acceleration: averaging the Doppler change over isotropic shock crossings gives a net gain per cycle first order — not second order — in $u/c$ , why shocks accelerate efficiently.

\left\langle \frac{\Delta E}{E}\right\rangle = \frac{4}{3}\,\frac{u_1-u_2}{c} \qquad (3)

where: $\Delta E/E$ — fractional gain per cycle; $u_1,u_2$ — upstream and downstream flow speeds in the shock frame; $c$ — speed of light. Because the gain is $\propto u/c$ , a $10^3\,\mathrm{km\,s^{-1}}$ shock boosts a particle by $\sim1\%$ per cycle. Balancing gain against downstream escape yields the source spectrum, described by two equivalent indices: the differential energy index $s$ and the phase-space index $q$ , related by $q=s+2$ since $N(p)\propto p^2 f(p)$ .

s = \frac{r+2}{r-1}, \qquad q = \frac{3r}{r-1}=s+2, \qquad \frac{\mathrm{d}N}{\mathrm{d}E}\propto E^{-s} \qquad (4)

where: $s$ — differential energy index; $q$ — phase-space index of $f(p)\propto p^{-q}$ ; $r=u_1/u_2$ — compression ratio. For a strong shock $r\to4$ : $s\to2$ , $q\to4$ — the celebrated $\mathrm{d}N/\mathrm{d}E\propto E^{-2}$ , fixed by $r$ alone. Propagation obeys a transport equation; the leaky-box approximation collapses the diffusion term to one escape time that falls with energy as faster particles escape sooner.

\frac{\partial N}{\partial t} = \nabla\!\cdot\!\left(D\,\nabla N\right) + Q - \frac{N}{\tau_{\rm esc}}, \qquad D(E) = D_0\left(\frac{E}{E_0}\right)^{\delta} \qquad (5)

where: $N$ — density per unit energy; $D(E)$ — diffusion coefficient; $Q$ — injection; $\tau_{\rm esc}\propto E^{-\delta}$ — escape time; $\delta\approx0.3$ – $0.6$ . In steady state $N\propto Q\,\tau_{\rm esc}\propto E^{-(s+\delta)}$ , turning the $E^{-2}$ source into the observed $E^{-2.7}$ . No accelerator can hold a particle whose Larmor radius exceeds its size. Equating gyroradius to source scale gives a hard ceiling — the Hillas criterion — deciding which objects can reach $10^{20}$ eV.

E_{\max} \sim Z\,e\,B\,R\,\beta_{\rm sh}\,c \qquad (6)

where: $E_{\max}$ — maximum confinable energy; $Z$ — charge number; $B$ — accelerator field; $R$ — region size; $\beta_{\rm sh}=v_{\rm sh}/c$ — flow speed. The ceiling scales as $Z\,B\,R$ , so only large, strongly magnetized sources (AGN lobes, jets) qualify — Worked Example 3. Finally, the Universe becomes opaque at the top: a proton above threshold sees CMB photons blueshifted past the pion-production threshold, bleeding energy every few tens of Mpc.

E_{\rm th} = \frac{(m_p+m_\pi)^2 - m_p^2}{4\,\varepsilon_\gamma}\,c^4 = \frac{m_\pi c^2\,(2m_p c^2+m_\pi c^2)}{4\,\varepsilon_\gamma} \approx 5\times10^{19}\ \mathrm{eV} \qquad (7)

where: $E_{\rm th}$ — GZK threshold for $p+\gamma_{\rm CMB}\to\Delta^+\to N+\pi$ ; $m_p,m_\pi$ — proton and pion rest masses; $\varepsilon_\gamma$ — typical CMB photon energy $\sim6\times10^{-4}$ eV. Above this the mean free path collapses to $\sim$ tens of Mpc — the GZK horizon. The same rigidity that sets $E_{\max}$ also sets how far a cosmic ray is bent, and whether it can point back to its source.

r_L = \frac{E}{Z\,e\,B\,c} \approx 1.1\ \mathrm{pc}\left(\frac{E}{10^{15}\,\mathrm{eV}}\right)\left(\frac{1\,\mu\mathrm{G}}{B}\right)Z^{-1} \qquad (8)

where: $r_L$ — gyroradius; $B$ — field ( $\sim3\,\mu$ G in the Galaxy). When $r_L$ approaches the disk half-thickness ( $\sim$ few hundred pc) the particle escapes — this happens near the knee for protons, tying confinement geometry to the spectral break.

Derivation ADiffusive shock acceleration and the universal E to the minus two spectrum

The goal is to combine the per-cycle energy gain of equation (3) with the per-cycle escape probability at a strong shock, and show the spectrum is a power law whose index depends only on the compression ratio $r=u_1/u_2$ . We track two geometric sequences — energy up, number down — until $k$ cancels.

Step 1. A particle isotropized on each side always sees the opposite flow approaching; averaging the Doppler change over the isotropic distribution of crossings gives $\Delta E/E = \tfrac{4}{3}(u_1-u_2)/c$ per cycle. Why: head-on encounters outnumber overtaking ones, so the angle average survives at first order in $u/c$ .

Step 2. Downstream advection carries particles away. The fraction lost per cycle is the flux advected across the downstream boundary over the flux recrossing the front, $P_{\rm esc} = 4u_2/c$ . Why: this loss makes the distribution a power law rather than a pile-up.

Step 3. After $k$ cycles the energy grows geometrically while the surviving number decays: $E_k = E_0\,(1+\Delta E/E)^k$ and $\mathcal{N}(>E_k)=N_0\,(1-P_{\rm esc})^k$ , where $\mathcal{N}(>E)$ counts particles boosted to at least $E$ . Why: both sequences share the integer $k$ , the bridge between energy and survival.

Step 4. Take logs and divide to eliminate $k$ : $\ln[\mathcal{N}/N_0]/\ln[E/E_0] = \ln(1-P_{\rm esc})/\ln(1+\Delta E/E)$ . In the small-increment limit this is $-P_{\rm esc}/(\Delta E/E) = -3u_2/(u_1-u_2) = -3/(r-1)$ , so the integral spectrum is $\mathcal{N}(>E)\propto E^{-3/(r-1)}$ . Why: expanding both logs turns a ratio of small numbers into a clean slope.

Step 5. Differentiate to get the observable differential spectrum, $\mathrm{d}N/\mathrm{d}E \propto E^{-(1+3/(r-1))} = E^{-(r+2)/(r-1)} \equiv E^{-s}$ ; equivalently $f(p)\propto p^{-q}$ with $q=s+2=3r/(r-1)$ . Why: the integral spectrum hides one power of $E$ ; differentiating restores it to the measured index.

Step 6. For a strong shock the Rankine–Hugoniot conditions give $r=(\gamma+1)/(\gamma-1)\to4$ for a monatomic gas ( $\gamma=5/3$ ), so $s\to2$ and $q\to4$ : $\mathrm{d}N/\mathrm{d}E\propto E^{-2}$ . Why: field, density, and shock speed have dropped out — only the compression ratio, capped at 4, survives. *For $r=4$ : integral index $3/(r-1)=1$ , differential $s=2$ , phase-space $q=4$ — the result of Bell (1978) and Blandford–Ostriker (1978).*

Go deeper · Derivation B · 1 minFrom source spectrum to observed spectrum via the leaky box

Derivation BFrom source spectrum to observed spectrum via the leaky box

The goal is to explain the gap between the $E^{-2}$ predicted at the source and the $E^{-2.7}$ measured at Earth, using the transport equation (5). We solve the leaky box in steady state and show the one bridging ingredient is the energy dependence of the escape time — the same quantity the boron/carbon ratio measures independently, so it is a cross-check, not a knob.

Step 1. In the leaky box the diffusion term becomes a homogeneous loss $N/\tau_{\rm esc}$ ; the steady state $\partial N/\partial t = 0$ reads $0 = Q(E) - N(E)/\tau_{\rm esc}(E)$ . Why: over $\sim10^7$ yr injection and escape balance, so the observed shape is set by their ratio, not by transient history.

Step 2. Solve: $N(E) = Q(E)\,\tau_{\rm esc}(E)$ — equilibrium density is injection rate times residence time. Why: the leaky-bucket picture — steady level equals inflow times drain time.

Step 3. Insert the power laws: the source $Q(E)\propto E^{-s}$ with $s\approx2$ , and since $D(E)\propto E^{\delta}$ the residence time falls as $\tau_{\rm esc}\propto 1/D \propto E^{-\delta}$ . Why: higher-energy particles leak out sooner, so they are under-represented at equilibrium.

Step 4. Multiply: $N(E)\propto E^{-(s+\delta)}$ . With $s=2$ , $\delta\approx0.6$ this is $E^{-2.6}$ , close to the observed $E^{-2.7}$ . Why: the observed slope is a sum of two meaningful numbers — acceleration ( $s$ ) and propagation ( $\delta$ ) — not one fitted index. *The same $\delta$ that steepens the primary spectrum also governs the measured decline of the B/C ratio with energy — an independent observational constraint on the diffusion index.*

Go deeper · Derivation C · 1 minThe GZK threshold from relativistic invariants

Derivation CThe GZK threshold from relativistic invariants

The goal is to compute the proton energy above which the CMB becomes an opaque wall via pion photoproduction, $p+\gamma\to\Delta^+\to N+\pi$ . The clean route is the frame-independent Mandelstam invariant $s=(P_p+P_\gamma)^2$ : equate its lab-frame value (fast proton, soft photon) to its threshold value (proton plus pion at rest) and solve.

Step 1. Write the invariant mass squared, neglecting the photon rest mass: $s = m_p^2 c^4 + 2\,E\,\varepsilon_\gamma\,(1-\cos\theta)$ , with $\theta$ the collision angle and $E$ the proton energy. Why: $s$ is frame-independent, the natural bookkeeping for a threshold that cannot depend on the observer.

Step 2. Threshold to make the $\Delta^+$ resonance (mass $\approx m_p+m_\pi$ ) requires $\sqrt{s}\ge (m_p+m_\pi)c^2$ , i.e. all the centre-of-mass energy goes into rest mass. Why: the cheapest way to make the new particle is to produce it at rest in the CoM frame.

Step 3. Take the most efficient geometry, a head-on collision with $1-\cos\theta=2$ , so $m_p^2 c^4 + 4E\varepsilon_\gamma \ge (m_p+m_\pi)^2 c^4$ . Why: a head-on hit maximizes the invariant mass, giving the lowest threshold.

Step 4. Solve: $E_{\rm th} = \dfrac{(m_p+m_\pi)^2c^4 - m_p^2c^4}{4\varepsilon_\gamma} = \dfrac{m_\pi c^2(2m_p c^2+m_\pi c^2)}{4\varepsilon_\gamma}$ . Why: expanding the square isolates the pion mass as the driver of the cost. *With $m_p c^2=938.3$ MeV, $m_\pi c^2=135$ MeV, and mean photon $\varepsilon_\gamma\approx6.3\times10^{-4}$ eV, the head-on threshold evaluates to $E_{\rm th}\approx1.1\times10^{20}$ eV; spectrum- and angle-averaging lowers the onset to $\sim5\times10^{19}$ eV.*

Limiting cases

The physics collapses to clean limits in every regime:

Case	Result	Meaning
Strong shock (high Mach)	$r = \frac{\gamma+1}{\gamma-1}\xrightarrow{\gamma=5/3}4,\quad s=\frac{r+2}{r-1}=2$	Compression saturates at 4; the differential index is 2, giving $\mathrm{d}N/\mathrm{d}E\propto E^{-2}$ independent of microphysics.
Weak shock (Mach $\to1$ )	$r\to1,\quad s=\frac{r+2}{r-1}\to\infty$	A barely-compressing shock gives a very steep, inefficient spectrum — strong shocks dominate.
Propagation (diffusive escape)	$\mathrm{d}N/\mathrm{d}E\propto E^{-(s+\delta)},\quad \delta\approx0.3\text{–}0.6$	Observed $\approx2.7$ = source index $2$ plus $\delta$ ; the same $\delta$ fixes the B/C decline.
GZK regime (above threshold)	$E_{\rm th}=\frac{m_\pi(2m_p+m_\pi)}{4\varepsilon_\gamma}\sim5\times10^{19}\,\mathrm{eV}$	Pion photoproduction on the CMB drains energy in tens of Mpc, so the flux is suppressed and sources must be nearby.

Common trapThe source spectrum is not the spectrum you measure

The most common error is to identify the observed $E^{-2.7}$ with what the accelerator produces. DSA predicts $\approx E^{-2}$ at the source; the extra steepening is entirely energy-dependent escape during propagation, so the two indices differ by $\delta\approx0.6$ . Two related traps follow. First, the knee and ankle are not sharp cutoffs of single sources — they are statistical transitions between populations, smeared over the Galactic ensemble. Second, accelerator limits scale with rigidity $E/Z$ , not energy: an iron nucleus ( $Z=26$ ) reaches $26$ times a proton's energy from the same source, so the knee should be composition-dependent, becoming heavier above the proton cutoff.

Worked examples

1 · The GZK threshold, computed (fully worked)

Find the proton energy at which pion photoproduction on the CMB switches on for a head-on hit with a typical photon, and reconcile with $\sim5\times10^{19}$ eV.

Step 1. Fix the photon energy. The CMB is a $2.725$ K blackbody, so $k_BT = (8.617\times10^{-5})(2.725) = 2.35\times10^{-4}$ eV and the mean photon energy is $\langle\varepsilon_\gamma\rangle = 2.70\,k_BT = \mathbf{6.3\times10^{-4}\ eV}$ .

Step 2. Evaluate the numerator of (7), $m_\pi c^2(2m_pc^2+m_\pi c^2)$ , with $m_\pi c^2=1.35\times10^8$ eV, $m_pc^2=9.383\times10^8$ eV: $(1.35\times10^8)(2.012\times10^9)=\mathbf{2.72\times10^{17}\ eV^2}$ .

Step 3. Divide by $4\varepsilon_\gamma$ : $E_{\rm th} = 2.72\times10^{17}/(2.52\times10^{-3}) = \mathbf{1.1\times10^{20}\ eV}$ for a head-on hit on a mean photon.

Step 4. Reconcile. The onset is softer than this single-photon, single-angle estimate: the CMB is broadband and all angles contribute, so averaging over the Planck spectrum lowers the effective threshold by about two. The GZK wall sits at $E_{\rm th}\approx5\times10^{19}$ eV, with a head-on single-photon threshold of $\approx1.1\times10^{20}$ eV.

2 · Can supernovae power the cosmic rays? (energetics budget)

Compare the power needed to sustain $u_{\rm CR}\approx1$ eV cm $^{-3}$ against escape with the mechanical power of Galactic supernovae.

Step 1. Required cosmic-ray luminosity is stored energy over residence time, $L_{\rm CR} = u_{\rm CR}\,V/\tau_{\rm esc}$ . With $V\sim10^{67}$ cm $^3$ (disk plus modest halo), $u_{\rm CR}=1.6\times10^{-12}$ erg cm $^{-3}$ , and $\tau_{\rm esc}\sim10^7$ yr $=3.2\times10^{14}$ s: $L_{\rm CR} = (1.6\times10^{-12})(10^{67})/(3.2\times10^{14}) \approx \mathbf{5\times10^{40}\ erg\,s^{-1}}$ .

Step 2. Supernova power: each explosion releases $\sim10^{51}$ erg at about one per $30$ – $50$ yr. At one per $30$ yr, $L_{\rm SN} = 10^{51}/(30\times3.16\times10^7) \approx \mathbf{1.1\times10^{42}\ erg\,s^{-1}}$ (one per $50$ yr gives $\approx6\times10^{41}$ ).

Step 3. Required efficiency: $\eta = L_{\rm CR}/L_{\rm SN} = 5\times10^{40}/1.1\times10^{42} \approx 0.05$ . Supernovae need convert only a few percent — of order $5$ – $10\%$ — of their mechanical energy into cosmic rays to power the whole Galactic population. That modest efficiency is the decisive argument for the SNR paradigm, and DSA reaches it naturally.

3 · Which sources can reach $10^{20}$ eV? (Larmor radius and Hillas)

Use rigidity to test candidate accelerators, and check what $10^{19}$ eV means for pointing.

Step 1. Recast the Hillas ceiling (6) via equation (8): confinement $r_L\le R$ gives $E_{\max}\approx0.9\times10^{15}\,Z\,(B/\mu\mathrm{G})(R/\mathrm{pc})$ eV. A $10^{20}$ eV proton ( $Z=1$ , $\beta_{\rm sh}\sim1$ ) demands $(B/\mu\mathrm{G})(R/\mathrm{pc}) \approx \mathbf{10^{5}\ \mu G\,pc}$ .

Step 2. A supernova remnant: amplified $B\sim100\,\mu$ G over $R\sim5$ pc gives $B\!\cdot\!R\sim500\,\mu$ G pc, so $E_{\max}\sim \mathbf{5\times10^{17}\ eV}$ — near the ankle at best, short of $10^{20}$ . SNRs run out around the knee.

Step 3. A giant AGN radio lobe: $B\sim10\,\mu$ G over $R\sim100$ kpc $=10^5$ pc gives $B\!\cdot\!R\sim10^6\,\mu$ G pc and $E_{\max}\sim \mathbf{9\times10^{20}\ eV}$ , clearing $10^{20}$ . Only large, magnetized extragalactic structures — AGN lobes and jets — survive the Hillas cut at the top.

Step 4. Pointing: a $10^{19}$ eV proton in a $1\,\mu$ G field has $r_L\approx1.1\,\mathrm{pc}\times10^{4} \approx 10$ kpc — a galaxy's size. Deflections stay large below the ankle, so cosmic-ray astronomy only becomes possible near $10^{20}$ eV, where gyroradii rival source distances.

Physics of the portrait

Cosmic-ray physics has no single founder, but its organizing idea — moving magnetic structures grinding a thermal population into a power law — carries one name.

Origins · 1 minEnrico Fermi and how this was discovered

Enrico Fermi

1901–1954 · the mechanism of cosmic-ray acceleration

Enrico Fermi was the rare physicist equally decisive in theory and experiment — he built the first nuclear reactor and reshaped particle physics, statistical mechanics, and astrophysics. His cosmic-ray contribution came in a 1949 paper proposing that charged particles gain energy stochastically by scattering off moving interstellar magnetic clouds. In that picture the net gain per encounter is second order in the cloud speed, $\langle\Delta E/E\rangle\propto(u/c)^2$ , because head-on and overtaking collisions nearly cancel — elegant, but far too slow and lacking a universal slope. The resolution came around 1977–1978, when Krymskii, Axford, Bell, and Blandford and Ostriker realized that at a shock the geometry breaks the symmetry: a particle always re-enters an approaching flow, so the gain becomes first order, $\langle\Delta E/E\rangle=\tfrac43(u_1-u_2)/c$ , and competition with escape fixes the index at $s=(r+2)/(r-1)\to2$ — regardless of speed, field, or gas. This diffusive shock acceleration is now the backbone of high-energy astrophysics, invoked wherever a supersonic flow meets a magnetized plasma: supernova remnants, relativistic jets, gamma-ray-burst blast waves, cluster shocks. Fermi's insight — that motion plus scattering makes a power law — outlived the mechanism he proposed and became the template for how the Universe makes its most energetic particles.

History

1912

— Hess flies electroscopes by balloon and finds ionization rising with altitude, revealing radiation from above

1938

— Auger detects coincident air showers, showing primary particles reach enormous energies

1949

— Fermi proposes stochastic (second-order) acceleration by moving magnetic clouds

1966

— Greisen, Zatsepin and Kuzmin predict the CMB energy cutoff, the GZK limit

1978

— Bell and Blandford and Ostriker formulate first-order diffusive shock acceleration

1991

— the Fly's Eye detector records the Oh-My-God particle at about 3×10²⁰ eV

Going further

Texts. Longair, High Energy Astrophysics, 3rd ed. — the definitive treatment of acceleration and propagation · Carroll and Ostlie, An Introduction to Modern Astrophysics, 2nd ed., interstellar-medium and cosmic-ray sections · Binney and Tremaine, Galactic Dynamics, 2nd ed., Galactic magnetic fields and diffuse components · Dodelson, Modern Cosmology, CMB spectrum and photon number density · Ryden, Introduction to Cosmology, 2nd ed., CMB photon fields. Spectral indices ( $\alpha\approx2.7,3.1$ ), knee ( $\sim3\times10^{15}$ eV), ankle ( $\sim4\times10^{18}$ eV), energy density ( $\sim1$ eV cm $^{-3}$ ), and confinement time ( $\sim10^7$ yr) are standard observational benchmarks. The DSA relations $s=(r+2)/(r-1)$ , $q=3r/(r-1)$ , gain $\tfrac43 u/c$ , escape $4u_2/c$ give integral index $1$ , $s=2$ , $q=4$ for $r=4$ . The GZK threshold $E_{\rm th}=m_\pi c^2(2m_pc^2+m_\pi c^2)/(4\varepsilon_\gamma)$ evaluates with $m_pc^2=938.3$ MeV, $m_\pi c^2=135$ MeV, $\varepsilon_\gamma\approx6.3\times10^{-4}$ eV ( $\langle\varepsilon\rangle=2.70\,k_BT$ , $T=2.725$ K) to $1.1\times10^{20}$ eV head-on and $\sim5\times10^{19}$ eV spectrum-averaged. The Galactic field $B\sim3\,\mu$ G sets the Larmor radius scale; the energetics ratio $u_{\rm CR}V/\tau_{\rm esc}\approx5\times10^{40}$ erg s $^{-1}$ vs $L_{\rm SN}\approx10^{42}$ erg s $^{-1}$ yields $\sim5$ – $10\%$ efficiency.

Prerequisites. Shock Waves and the Rankine–Hugoniot Conditions · Magnetohydrodynamics and Interstellar Magnetic Fields · Special Relativity and Four-Momentum · Cosmic Microwave Background. Leads to. Relativistic Jets and Launching Mechanisms · Active Galactic Nuclei: Unified Model · Blazars and Relativistic Beaming.

Ask the tutor: Ask me to walk through why a strong shock forces the spectral index to exactly 2, or to work the GZK threshold from Mandelstam invariants step by step.

Why are they called cosmic "rays" if they are particles?

The name is a historical fossil. When ionizing radiation from above was found in the 1910s, the fashionable guess was some penetrating gamma "ray". Only in the 1930s did deflection by Earth's field — the east–west effect — prove the primaries are mostly positively charged particles, chiefly protons and nuclei. By then "cosmic rays" was entrenched, and it survives even though the objects are baryons; the secondary "rays" reaching the ground are cascades of muons, electrons, and photons from a primary striking the upper atmosphere.

If sources are isotropic, how do we know supernova remnants are the accelerators?

Two arguments, neither relying on pointing. First, energetics: sustaining the Galactic energy density against escape needs $\sim5\times10^{40}$ erg s $^{-1}$ , and supernovae deliver $\sim10^{42}$ erg s $^{-1}$ , so a few-percent efficiency — exactly what shock acceleration achieves — closes the budget, and no other Galactic source class comes close. Second, imaging: in IC 443 the "pion bump" from proton–proton collisions is unambiguous proof that protons are accelerated in situ. Budget and imaging make the case without tracing a single ray home.

What physically distinguishes the knee from the ankle?

Opposite kinds of transition. The knee, near $3\times10^{15}$ eV, is where the spectrum steepens: Galactic accelerators reach their rigidity ceiling, so lighter nuclei drop out first and the flux falls faster — the top of the Galactic engine's reach. The ankle, near $4\times10^{18}$ eV, is where it flattens: a soft, declining Galactic component is overtaken by a harder extragalactic one, so a steep curve is refreshed by a flatter one underneath. The knee is one population running out of steam; the ankle is a tougher one taking over.

Why does the source spectrum come out E to the minus 2 for essentially any strong shock?

Because the index depends on a single pinned quantity. Gain-versus-escape gives $s=(r+2)/(r-1)$ , a function of the compression ratio $r$ alone — field, density, and shock speed cancel. And $r$ is not free: the Rankine–Hugoniot conditions cap it at $(\gamma+1)/(\gamma-1)=4$ for a strong shock in a monatomic gas, so $r=4$ gives $s=2$ every time. That is why a supernova shock and a relativistic jet produce nearly the same slope: they share the same limiting compression.

Is the flux suppression above 5×10¹⁹ eV definitely the GZK effect?

Strongly consistent with GZK, but not proven to be it alone. The suppression could instead reflect the maximum energy of the sources: if accelerators cannot push protons past a few $\times10^{19}$ eV, the spectrum ends there regardless of the CMB. The two scenarios predict different composition and different "GZK horizons," and Auger's evidence for a heavier composition at the top keeps both alive. Distinguishing a propagation cutoff from a source cutoff is a central open question — which is why composition measurements matter so much.

Will we ever do real cosmic-ray "astronomy" and point back to sources?

Only at the very highest energies, and even then imperfectly. Below the ankle, gyroradii in the microgauss field are far smaller than a galaxy, so directions are scrambled and the sky is isotropic. Near $10^{20}$ eV the Larmor radius approaches tens of Mpc, deflections shrink to a few degrees, and correlations with nearby matter emerge — Auger's dipole is the first hint. But the fields are poorly mapped, so pointing stays fuzzy. The practical path is multi-messenger: neutrinos, which travel undeflected, may localize the accelerators the cosmic rays cannot.

Cosmic Rays: Origin and Spectrum

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · The GZK threshold, computed (fully worked)

2 · Can supernovae power the cosmic rays? (energetics budget)

3 · Which sources can reach $10^{20}$ eV? (Larmor radius and Hillas)

Physics of the portrait

Going further

Before this

Where this leads

Cosmic Rays: Origin and Spectrum

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · The GZK threshold, computed (fully worked)

2 · Can supernovae power the cosmic rays? (energetics budget)

3 · Which sources can reach 102010^{20}1020 eV? (Larmor radius and Hillas)

Physics of the portrait

Going further

Before this

Where this leads

3 · Which sources can reach $10^{20}$ eV? (Larmor radius and Hillas)