Cosmic ray interactions with matter and radiation

Pasquale Dario Serpico

Laboratoire d’Annecy-le-Vieux de Physique Théorique, USMB, CNRS, F-74940 Annecy, France

Summary. — The main goal of the present lectures is to outline the key particle
interactions and energy loss mechanisms in the Galactic medium that high-energy
particles are subject to. These interactions are an important ingredient entering
the cosmic ray propagation equation, contribute shaping cosmic ray spectra. They
also source the so-called secondary species, like gamma rays, neutrinos, “fragile”
nuclei not synthesised in stars and antiparticles, all routinely used as diagnostic
tools in a multi-messenger context. These lectures are complementary to Denise
Boncioli’s ones, focusing instead on processes happening at ultra-high energies in the
extragalactic environment. They include propædeutic material to Felix Aharonian’s
and, to some extent, Stefano Gabici’s and Carmelo Evoli’s lectures.

1. – Introduction

As illustrated in other courses of these lectures series, in cosmic ray (CR) (1 ) physics
we are in general interested in a statistical description of the species, typically encoded in
ensemble averages (and often angular-averages) of the single-particle distribution func-
tions, sufficient to describe such rarefied species. The corresponding set of coupled Boltz-

(1 ) Here I use the term CRs in a broad sense to mean non-thermal, high-energy astrophysical
particles, not necessarily “measured at the Earth”. While focusing on charged particles, I will
also comment on some aspects involving the secondary γ-rays produced in their interactions.

c Società Italiana di Fisica 1

2 P. D. Serpico

mann equations is the master tool to describe such systems. Compared to other settings,
however, most of the peculiarities of the CR dynamics arise from the highly non-trivial
nature of the transport or left-hand-side (LHS) part of the Boltzmann-Vlasov equation,
where collisionless phenomena (involving electromagnetic inhomogeneities in the magne-
tised environments where CRs evolve) are responsible for effects like spatial or momentum
diffusion. This does not mean, however, that collisional effects, i.e. the right-hand-side
(RHS) of the Boltzmann equation, are unimportant: For instance, the maximum energy
at which particles can be accelerated, Emax , as well as the spectra and the composition
of CRs are influenced by those terms. Also, charged CRs give rise collisionally to γ-rays
and ν’s, which are of great importance for astrophysical diagnostics. The structure of
this course is the following: In this section we introduce some basic notions needed for
the lectures. Sec. 2 deals with ionisation and Coulomb losses, while Sec. 3 deals very
briefly with Bremsstrahlung: These two sections include the essentials of the electromag-
netic CR interactions with interstellar matter. Then, in Sec. 4, we cover the topic of
interactions of CRs with photon and magnetic fields, which is of importance for electrons
and positrons. Sec. 5 deals with hadronic interactions, discussing their role not only as
energy loss mechanisms but above all as source terms for secondary byproducts. The
most important species produced as byproducts of both leptonic and hadronic interac-
tions are photons: Sec. 6 deals with some aspects of high-energy photon interactions, of
relevance to some extent for direct detection of gamma rays, but above all for extragalac-
tic very high-energy photons and links with very high-energy neutrinos. It provides thus
a natural link with topics covered in D. Boncioli’s lectures. Finally, Sec. 7 makes use of a
number of the notions outlined, incorporating them in some (simplified) CR propagation
problems to substantiate some statements made in this introductory paragraph.
Fundamental microscopic quantities like cross-sections, hopefully familiar from nu-
clear and particle physics courses, are the building blocks needed in the considerations
that follow. I will use natural units for microscopic scales (where c = kB = ~ = 1), so
that e.g. energy, mass, momenta, temperature, inverse length and inverse time have the
same unit, eV and multiples. Astrophysical units (parsecs and multiples) are instead
used for distances. Cross sections are in barns (1 b=10−24 cm2 ). In agreement with most
astrophysical literature, the Gaussian electromagnetic convention is used (with the 4π’s
in Maxwell equations, not in Coulomb or Biot-Savart laws). The charge of the positron
√ p
is e = α ' 1/137 ' 0.085. The magnetic field energy density is for instance B 2 /(8π).
The environmental density of targets and their energy distribution are crucial in
determining the mean-free-path (mfp) ` and the associated collision rate Γ of a particle.
If σ is the cross-section of the interaction process, β is the particle velocity (associated
to the Lorentz factor γ = (1 − β 2 )−1/2 ), its mean-free-path and interaction rate are

1 β
(1) `= , Γ = σβn = .
σn `

Over an infinitesimal distance interval ds, the dimensionless optical depth associated to
the process having mfp ` is dτ = ds/`.
Cosmic ray interactions with matter and radiation 3

Knowing the differential cross-section with respect to the energy transferred W ,

dσ/dW , one can define moments of the energy loss; for instance, the first moment,
known as stopping power is
Z
dE 1 dE d|p| dσ
(2) − =− =− ≡n dW W .
dx β dt dt dW

Similarly, the second moment can be used to define an energy straggling parameter in
terms of which to describe the variance of E losses (see [1, 2]). Loss rates are useful to
estimate the stopping time or range over which the particle of mass m loses its energy
via the process:

E E
dE 0 dE 0
Z Z
E E
(3) tloss ≡ ∼ , dloss ≡ ∼ .
m −dE 0 /dt −dE/dt m −dE 0 /dx −dE/dx

Collision are dubbed catastrophic when the particle of interest disappears or the fraction
of its initial energy E transferred to other particles, ∆E/E, is sizeable. In this case,
tloss ∼ Γ−1 and dloss ∼ `. If the primary particle retains its nature and most of its energy
in each elementary process, losses are dubbed continuous. In this case, tloss  Γ−1 and
dloss  `.
In a collision, the square of the center-of-momentum (CoM) energy is a relativistic
invariant:
!2
X
(4) s= pi = m2a + m2b + 2Ea Eb (1 − βa βb cos ϑ) ,
i

where the last equality is specific for a two-body collision a + b → X. If equated to the
square of the sum of masses of the final state, this allows one to estimate threshold ener-
gies. For instance, the minimum energy in the Lab frame to to produce one antiproton
in a collision of a CR proton with a medium proton at rest can be inferred applying s-
conservation to pp → p̄ppp (process of the type pp → p̄ X with the lightest X compatible
with charge and baryon number conservation), with the final state particles at rest in
the Lab frame; eq. (4) reduces to 2m2p + 2Ep mp = (4mp )2 , i.e. Ep > 7 mp ' 6.6 GeV.
Another useful invariant in a+b → a0 +b0 is the (square) of the momentum transferred

2 2
(5) q 2 = (pa − p0a ) = (pb − p0b ) .

For the case of a particle of mass m and momentum p scattering with another particle,
indicating with a prime the outgoing momentum and ϑ the angle between incoming and
outgoing momentum, one has

2 el
(6) q 2 = (p − p0 ) = 2m2 − 2EE 0 (1 − ββ 0 cos ϑ) = −4|p|2 sin2 ϑ/2 ,
4 P. D. Serpico

where the last step is valid for an elastic process (2 ). This variable helps us illustrate a
key difference between processes of interest for CR and collider studies. The reader is
familiar with the strong angular dependence of Rutherford scattering (if not, appendix A
provides a reminder) describing the elastic deflection of a particle of charge Zp e and mass
m with momentum |p| (velocity β) impinging on a heavy target at rest of charge Zt e.
Its differential cross section is

dσ Zp2 Zt2 α2 dσ 4Zp2 Zt2 α2 γ 2 m2

(7) = ⇐⇒ = ,
dΩ R 4|p|2 β 2 sin4 ϑ/2 dΩ R q4

where the RHS gives the manifestly Lorentz-invariant form. Note how the process is
dominated by small-angle scatterings (low q 2 ) (3 ). On the other hand, if one wants to
produce a particle of mass M in a collision, a rough criterion must be |q 2 | > M 2 . So,
the bulk of the cross-section (what is most relevant for CR physics in the atmosphere,
for instance) is dominated by the so-called forward (i.e. |ϑ| ' 0) physics; machines like
LHC largely (but not exclusively! See e.g. [3]) focus instead on large-angle scatterings,
which are associated to large exchanged momenta. Together with the relative rarity of
very energetic CR, which makes hard to study processes with large s, this is the main
reason why CRs and high-energy collider physics have largely complementary targets.

2. – Ionisation and Coulomb losses

Both CR hadrons and leptons can interact electrostatically with a medium containing
electrons with density ne . A typical benchmark value of ne in the interstellar medium
(ISM) of the Milky Way is 1 cm−3 , distributed in atomic, molecular and ionised hydrogen
and helium gas (with traces of heavier elements) as well as in dust grains. While this is
a very rarefied medium for terrestrial standards, the large distance and long timescales
involved nonetheless imply significant CR energy losses.
It is conceptually useful to classify the processes that CR experience depending on
the amount of energy W transferred to their target. If this is below some value WH , the
motion of the impinging particle of mass m and charge Ze is basically unaffected and
the loss can be considered as a continuous process. Technically, this regime is the one
where the cross-section is well modelled by Rutherford diffusion of a fast light projectile
in the frame where the CR is at rest (see Fig. 1). From eq. (7), changing variable to
q 2 (note that eq. (6) implies dq 2 = −2|p|2 sin ϑdϑ = −|p|2 dΩ/π) and accounting for the

(2 ) Note: Any real photon verifies q 2 = 0; if the scattering process is mediated by the electro-
magnetic interaction, since q 2 6= 0 it means that a virtual photon is being exchanged.
(3 ) The small-angle divergence of the Rutherford cross-section is unphysical since when passing
far away from the target, the projectile sees the whole electrically neutral atom or plasma.
A more physical setting consists in computing the scattering from a Yukawa potential V =
Zt Zp αe−µr /r where the Yukawa mass constant µ is the inverse of the screening length. The
corresponding result, equivalent to eq. (7) where q 2 is replaced by q 2 + µ2 , is dubbed screened
Rutherford cross section.
Cosmic ray interactions with matter and radiation 5

target e
target e

CR
Fig. 1. – Diagram for ionisation and Coulomb collision losses, whose elementary cross-section
can be derived as Rutherford scattering of target electrons in the frame where the CR is at rest.

fact that in the Lab frame, which is the electron rest frame, the energy W transferred to
the electron is related to q via q 2 = −2me W , one has

dσ 2πZ 2 α2
(8) = ,
dW R me β 2 W 2

valid between a low energy WL & O(B) , where B is the binding energy or related to the
plasma frequency (neglected above, since we considered the electrons as free), and WH .
Note that in principle the knocked-out electrons also contribute to the low-energy CR
population, but this contribution is typically negligible in CR applications [4, 5] .
A posteriori, eq. (8) justifies neglecting energy losses in interactions with nuclei (so-
called nuclear stopping power), since for a target of mass M , me is replaced by M  me
at the denominator of the RHS expression (4 ). Using eq. (2) one infers

dE 2πne Z 2 α2 WH
(9) − = LC , LC ' ln .
dx me β 2 WL

Eq. (9) provides the basic description of the continuous energy losses, with more and
more refined considerations replacing LC ' 20, the so-called Coulomb logarithm, with
more accurate expressions. In specific applications to CR physics, relevant formulae—
that can be found for instance in [9, 10, 11]—consist in replacing LC in eq. (9) with
functions appropriate for the medium conditions and primary particle considered. For
conditions of interest to us, note that the dE/dt is almost energy-independent.
In the following, rather than presenting fitting formulae, let me briefly describe the un-
derlying theoretical approaches through which one completes and improves upon eq. (9).

(4 ) Note that for angular deflection, instead, nuclei contribute a factor Z more than electrons
according to eq. (7). This is unimportant for CRs in the ISM, since spatial transport is colli-
sionless. However, it does matter if one attempts to describe the lateral distribution function of
CR showers in the atmosphere, as first successfully tackled by Molière [6, 7] and improved upon
by Bethe [8].
6 P. D. Serpico

For example, the stopping power integrated at low exchanged energies, W < WH , cannot
depend on the arbitrary parameter WL . This problem can be solved within a quantum-
mechanical treatment, the first and simplest version of which was provided by Bethe, in
terms of a mean excitation potential, I. It yields

2πne Z 2 α2 2me β 2 WH
 
dE 2
(10) − = ln − β .
dx W <WH me β 2 (1 − β 2 )I 2

At the opposite kinematical end when large energies are exchanged, W > WH , the
Rutherford cross-section becomes in general inadequate, because the description of the
projectile (spinless and without recoil in its initial rest-frame) and/or the target elec-
trons (spinless) are insufficiently accurate. For heavy projectiles, often the only cor-
rection taken into account is the fact that the electron is a Dirac fermion, not a scalar.
That amounts to replace the Rutherford cross-section by the so-called Mott cross-section.
This corrects eq. (7) by the multiplicative factor [1 − β 2 sin2 θ/2] → cos2 θ/2 (the lat-
ter in the relativistic limit). Kinematics imposes the maximum energy transfer to be
Wmax = 2γ 2 β 2 me /[1 + 2γme /m + m2e /m2 ] ' 2γ 2 β 2 me , see Appendix B for a derivation.
The appropriate modification to eq. (8) then amounts to the factor [1 − β 2 W/Wmax ].
Integrating over large energy exchanges, from WH to Wmax then yields

2πne Z 2 α2 2πne Z 2 α2 2γ 2 β 2 me
   
dE Wmax 2 2
(11) − = ln − β ' ln − β .
dx W >WH me β 2 WH me β 2 WH

Corrections to the Rutherford cross section also arise when accounting for the recoil of
the impinging particle and its spin. These lead to multiplicative factors of unity plus
factors suppressed by ratios of the energy to the target mass or powers of them, and
are thus negligible for hadronic CRs; further details can be found e.g. in [12]. A brief
review on where these “corrections” originate from at a fundamental level, as well as
the link between the classical Rutherford scattering and the computations of relevant
cross-sections in QED is provided in Appendix C.
Integrating over all transfers W , i.e. summing eq. (10) to eq. (11), finally gives

4πne Z 2 α2 2me β 2 γ 2
   
dE 2
(12) − ' ln −β .
dx Bethe me β 2 I

The microphysics entering the probably familiar Bethe stopping power formula, eq. (12),
(for a derivation and more extensive discussion see e.g. chapter 13 of [2]) is sufficient
to describe losses of hadronic CRs. More accurate descriptions used for laboratory ap-
plications often require a number of corrections in the description of the medium and
projectile [13]. For instance, at low-energies the atomic shell structure is important
(C-correction), and a projectile with Z > 1 may not be fully ionised, thus motivating
the notion of effective charge, Zeff . At high-energy, medium-polarisation effects are im-
portant (density correction, or δ-term). Z-dependent higher-order corrections are also
sometimes introduced, e.g. to account for Coulomb distortion of wavefunctions (eq. (12)
Cosmic ray interactions with matter and radiation 7
− − − −
e e e e

e− e− e− e−
Fig. 2. – Diagrams for Møller (or electron-electron) scattering: t-channel to the left, u-channel
to the right. There is a minus sign between the two amplitudes, due to the fermionic nature.

assumes Born approximation). The free parameters entering these improved descriptions
are often empirically-deduced; further details can be found e.g. in [13] or in the “Passage
of particles through matter” mini-review in the PDG [14].
For electrons and positrons, the cross-sections leading to eq. (11) are inadequate, at
least if one wants to describe processes at better than the 10% level. The recoil of the
particles is relevant, the identical nature of particles enters for e− ’s, spin does matter and
in general more than one Feynman diagram contribute even at tree level. The relevant
e− e− scattering in QED is described by the Møller cross section [15](5 ) , see Fig. 2

2πα2 W2 (γ − 1)2 W 2
 
dσ 2γ − 1 W
(13) = 1+ + − ,
dW Mø me β 2 W 2 (E − W )2 γ2E2 γ2 E − W

with the further condition that Wmax = E/2 due to the identical nature of particles. For
e+ e− scattering, Bhabha cross section [20] is the relevant one, see Fig. 3 for the governing

(5 ) The Danish physicist Christian Møller actually inferred this formula heuristically. Its deriva-
tion within QED is due to Bethe and Fermi [16]. This may be the appropriate occasion to remind
the readers about the often-forgotten role that Fermi played in the early days of QED. Albeit
less well-known than many other of his theoretical contributions (quantum statistics, weak the-
ory. . . ), it was crucial especially since he provided an appealing, accessible and coherent synthesis
of the advances that colleagues like Dirac, Heisenberg, Pauli and others had obtained. His 1932
pedagogically remarkable review [17] was eventually at the basis of the reference textbook by
Heitler [18]. Fermi’s approach trained in QED generations of physicists in the 30’s and 40’s,
influencing people like Feynman and Weisskopf. For a review, see [19].
8 P. D. Serpico
+ + + +
e e e e

e− e− e− e−
Fig. 3. – Diagrams for Bhabha (or electron-positron) scattering: t-channel to the left, s-channel
to the right. There is a minus sign between the two amplitudes, due to the fermionic nature.

diagrams at tree level:

( 2  2
2πα2 γ2 − 1 W

dσ 1 γ−1 W
(14) = 2 2
1− 2
+
dW Bh me β W γ E 2 γ E
" #
2 2
γ2 − 1 W
   
γ−1W γ+2 γ−1 E
− −2 +
γ+1 E γ γ2 E γ W
 2  2 "  2  #)
γ−1 W 1 1 3 γ−1 W W
+ + + 2− 1− .
γ+1 E 2 γ 2γ γ E E

Relevant stopping power formulae can be found e.g. in [12, 14], and are reported here:

e−
" 2 #
2πne α2

dE me (γ − 1) 2γ − 1 1 γ−1
(15) − = ln +1− log 2 + ,
dx W >WH me β 2 4WH γ2 8 γ

and
(16)
e+
2πne α2 me (γ − 1) β 2
  
dE 14 10 4
− = ln − 11 + + + .
dx W >WH me β 2 WH 12 γ + 1 (γ + 1)2 (γ + 1)3

Adding the low energy transfer contribution, one gets:

(17)
e−
" 2 #
2πne α2 m2e β 2 γ 2 (γ − 1)

dE 2γ − 1 1 γ − 1
− = ln + (1 − β 2 ) − log 2 + ,
dx me β 2 2I 2 γ2 8 γ

and
(18)
e+
2πne α2 2m2e β 2 γ 2 (γ − 1) β 2
  
dE 14 10 4
− = ln − 23 + + + .
dx me β 2 I2 12 γ + 1 (γ + 1)2 (γ + 1)3
Cosmic ray interactions with matter and radiation 9

In addition, positron annihilation should be taken into account at sufficiently low

energies; it is obviously responsible for a catastrophic loss. Its cross section is:
n
h
1/2 i
1/2 o
+
πα2 γ 2 + 4γ + 1 ln γ + γ 2 − 1 − (3 + γ) γ 2 − 1
e
(19) σann = .
m2e (γ + 1) (γ 2 − 1)

Finally, note that the processes described here from the CR perspective are important
also for the interstellar medium (ISM), since they ionise and heat it. Notably, CR are
thought to be the only ionising medium effective in dense molecular clouds, which are
stellar formation nurseries. This is another role played in astrophysics by CR collisional
effects, albeit its detailed treatment also require modeling of atomic and molecular physics
not discussed here. For interested readers, a seminal paper is [21], while a recent review
of the subject can be found in [22].

3. – Bremsstrahlung

Sufficiently energetic particles tend to radiate a photon under the braking action of
the nuclear and electronic electric fields, a process associated to significant losses for
CR leptons, see Fig. 4. The relevant differential bremsstrahlung cross section for an
electron of initial energy E to produce a photon of wavenumber k (ending up with
energy Ef = E − k) impinging onto the target i can be written as
" ! #
dσi α3 Ef2 (i) 2 Ef (i)
(20) = 2 1+ 2 φ1 − φ ,
dk me k E 3 E 2

a result due to Bethe and Heitler [23] which relies on Born perturbation theory (plane
waves for the incoming/outgoing electron), on whose general form we will comment
below. If the target is an unshielded charge Ze, one has
 
2 2 2EEf 1
(21) φ1 = φ2 = Z φu , φu = 4Z ln − ,
me k 2

while more complicated expressions hold e.g. for partially shielded nuclei. For a neutral
plasma of electrons and a single nucleus, summing eq. (20) over the two species yields
" #
4(Z 2 + Z)α3 Ef2

dσtot 2 Ef 2EEf 1
(22) = 1 + − ln − .
dk m2e k E2 3 E me k 2

For scattering on a target of density nt , we thus derive

Z E
dE 1 dσ
(23) − = nt EΦrad
t , Φrad
t ≡ dk k ,
dx Brems E 0 dk
10 P. D. Serpico

e e γ

target
target
Fig. 4. – One of the QED tree-level diagrams for bremsstrahlung.

with Φrad
t ∼ O(α3 ), almost (exactly, according to eq. (22)) independent on E. A few
comments worth making:

• Compared to eq. (8), eq. (20) is suppressed by a factor α, which finds an obvious
interpretation in QED since the elementary diagram for bremsstrahlung contains
an extra vertex associated to the radiated photon, compare Fig. 4 with Fig. 2.

• The bremsstrahlung stopping power, eq. (23), grows in importance with respect to
eq. (12) when the E grows, with the stopping timescale for bremsstrahlung defined
in eq. (3), nearly independent of E.

• The analogous of eq. (20) for CR nuclei of mass m = A mp and charge Z e would
lead to a cross section suppressed by Z 4 A−2 (me /mp )2 ' 10−7 (Z 4 /A2 ), which im-
mediately shows how much more inefficient electromagnetic radiative processes are
expected to be for nuclei; energy losses of CR nuclei at intermediate and high-E
are dominated by hadronic interactions, dealt with in Sec. 5.

This is but a very short introduction to the topic, which in fact has vast applica-
tions in astrophysics and where it is sometimes known also under free-free emission.
Although a quantum treatment is needed for a throughout analysis, several regimes are
also amenable to classical approximations. Besides the textbook [24], another useful and
modern reference adopting the classical electrodynamics approach, with several astro-
physical applications is [25]. A classical reference with a more particle physics approach
for the material covered here is [26]; a complete reference textbook covering clearly clas-
sical and quantum aspects, relativistic and non-relativistic limits is [27]. These are useful
references that cover not only the material introduced in this section, but also the one
outlined in the following one.

4. – Interactions with electromagnetic fields

Besides rarefied gas (and dust), the ISM of our Galaxy, as well as external ones, also
contains photons and is threaded with magnetic fields. These are further targets for
energy-loss mechanisms of CRs, described in the following.
Cosmic ray interactions with matter and radiation 11

That the ISM is magnetised is most clearly revealed by radio observations (Fara-
day rotation, synchrotron radiation) but also via polarized light emission (due to dust
grains) or, in some cases, via Zeeman splitting. A magnetic field coherent (at least) up
to scales comparable to several kpc-long Galactic structures, such as spiral arms, has
been detected. Radio observations of external Galaxies as well as our own, notably via
synchrotron emission (see below), indicate several kpc thick magnetized halos embed-
ding the stellar disks. The ISM magnetic field extends to small scales down to at least
the typical pc-scale distance between neighbouring stars, where the field orientation is
believed to fluctuate following the turbulence of the ISM. Inferred intensities of Galactic
fields (typically via radio synchrotron emission, its polarization and its Faraday rotation)
are in the 1-10µG range. For a relatively recent review and ample references on Galactic
and cosmic magnetism, including microphysics and simulations, see for instance [28].
The most important photon backgrounds for CR studies are:

• The CMB (blackbody of cosmological origin and temperature of about 2.7 K),
pervading the whole universe (Galaxy and extragalactic sky alike). While today
its energy density is ∼ 0.3 eV/cm3 , it scales with redshift as (1 + z)4 .

• The extragalactic background light (EBL), pervading the whole universe in the
IR to UV range, due to primary stellar emission and secondary radiation from
reprocessing, with a benchmark number density of roughly a photon /cm3 ; a recent
determination of its spectral energy density can be found in [29]. Backgrounds at
different wavelengths also exist (e.g. radio) but are less important for what follows.

• In the Galactic environment, in addition there are important UV, optical, and IR
backgrounds associated to stellar light emission and its reprocessing, with typically
larger energy density than the CMB and non-homogeneously distributed, notably
peaking towards the inner Galaxy. The energy densities of magnetic fields, radiation
and cosmic rays are equal within less than one order of magnitude in our Galaxy,
amounting to . 1eV/cm3 each.

In the ISM of our Galaxy, the benchmark density of matter ∼1 cm−3 is to be compared
with ∼ 410 cm−3 CMB photons, and two orders of magnitude lower densities for starlight.
Combined with typical Myr propagation timescales and the much larger hadronic cross
sections than electromagnetic ones, this makes obvious that interactions with radiation
are irrelevant for CR nuclei propagating in the Galaxy. By the way, a CR proton with
energy E = γmp propagating in the Milky Way sees a background photon of energy
 as having an energy Eγ ∼ γ in its rest-frame: At the typical GeV-TeV energies of
interest, and up to the PeV scale, the CR sees sub-MeV photons, a kinematical factor
preventing potentially interesting inelastic channels to be open. In what follows, we
will thus concentrate on interactions of CR electrons with radiation, which are instead
relevant.
Note that the situation is different for ultra-high energy cosmic rays, whose extra-
galactic origin implies:
12 P. D. Serpico

k k0 B γ

p p0 e e

Fig. 5. – Left: one of the two QED tree-level diagrams for (Inverse) Compton scattering, with
labelled momenta. Right: The analogous diagram for synchrotron radiation, with an interaction
with a virtual photon associated to an external B-field.

• Propagation times & Gyrs.

• A medium with ∼ 107 times lower densities of matter, making hadronic interactions
irrelevant.

• Larger Lorentz factor (& 109 ), thus allowing interesting kinematical threshold for
interactions with radiation to be open.

For a treatment of these processes, we refer to D. Boncioli’s lectures.

.
4 1. Inverse Compton. – The elementary QED process at the hearth of the interac-
tions of leptons with radiation is the two body scattering eγ → e0 γ 0 , dubbed Compton
scattering if considering the photon as projectile and the electron as target (at rest), or
Inverse Compton if the electron is the energetic particle transferring part of its energy
to a low energy photon (see Fig. 5, left).
First, let us focus on its kinematics. Imposing energy momentum balance in the
photon (whose momenta are labelled via k, k 0 )-electron (momentum p, p0 ) scattering:

(24) kµ + pµ = kµ0 + p0µ ,

the on-shell condition for the electron mass, m2e = p2 = p02 , and that k 2 = k 02 = 0 for
real photons, one has:

(25) m2e = (p + k − k 0 )µ (p + k − k 0 )µ =⇒ p · (k − k 0 ) = k · k 0 .

Let us evaluate eq. (25) in the electron rest-frame, p = (me , 0, 0, 0); let the x-axis be the
incoming direction of the photon, and the x − y plane the one spanned by the incoming-
outgoing direction. Then k = (1, 1, 0, 0), while k 0 = 0 (1, cos θ, sin θ, 0), so that


(26) me ( − 0 ) = 0 (1 − cos θ) =⇒ 0 =  .
1+ me (1 − cos θ)
Cosmic ray interactions with matter and radiation 13

Eq. (26) illustrates that the ratio /me controls the photon energy change, so that unless
the energy of the photon in the electron rest-frame is comparable to or larger than the
mass of the electron me , the photon energy is only slightly altered. If the electron has a
velocity β (Lorentz factor γ) in the Lab frame, eq. (26) is valid in the electron rest-frame,
denoted via a tilde, which is boosted by γ with respect to the lab:

˜
(27) ˜0 = ˜
,
1+ me (1 − cos θ̃)

with ˜ = γ(1 − β cos θ). If we want to express ˜0 in the Lab frame, we need to perform
the “reverse boost”, such that 0 = γ(1 + β cos(π − θ̃))˜ 0 ; hence

˜ (1 − β cos θ̃)(1 − β cos θ)

(28) 0 = γ(1 − β cos θ̃) ˜
= γ2 γ
.
1+ me (1 − cos θ̃) 1+ m e
(1 − β cos θ)(1 − cos θ̃)

In the limit where  γ  me , i.e.  Ee  m2e , known as Thomson regime, with the
appropriate aberration relation, cos θ = (cos θ̃ − β)/(1 − β cos θ̃), one has that

 2
4 2    Ee
(29) 0 = γ 2 (1 − β cos θ̃)2 ⇒ h0 i = γ '5 MeV ,
3 eV GeV

where we used the isotropy of the scattered photons in the electron rest-frame to perform
the average. For  Ee > m2e , known as Klein-Nishina regime, one has instead

(30) 0 ' γme ∼ Ee .

These two kinematical regimes also manifest differences in the cross-section of the
process. The general differential cross-section at tree level in QED was derived by Klein
and Nishina [30], constituting one of the first QED predictions (see the diagram in Fig. 5,
left); for unpolarised photons in the rest-frame of the electron, it reads

 0 2 
α2 0

dσ   2
(31) = + − sin θ̃ ,
dΩ 2m2e  0 

where  (0 ) is the initial (final) energy of the photon, and θ̃ its deflection angle. In terms
of σT ≡ 8πα2 /(3m2e ) (Thomson cross-section) the total cross section writes
Z π
dσ
σ = 2π sin θ̃dθ̃ =
0 dΩ
   
3 1 + x 2x(1 + x) 1 1 + 3x
(32) σT − ln(1 + 2x) + ln(1 + 2x) − ,
4 x3 1 + 2x 2x (1 + 2x)2
14 P. D. Serpico

where x ≡ /me , so that

(33) σ(x) ' σT (1 − 2x + . . .) for x  1 (Thomson limit)

 
3 1 1
(34) σ(x) ' σT ln 2x + for x  1 (Klein − Nishina limit) .
8 x 2

Combined with the previously derived kinematical relations eqs. (29,30), we can thus
conclude that in the Thomson limit, collisions are frequent, the background photon
energy gets boosted quadratically in the electron energy, but the electron only loses a
small amount of energy, so that the loss can be considered continuous. On the other
hand, in the Klein-Nishina limit the cross-section is suppressed, but once interacting the
electron transfers a significant fraction of its energy to the background photon. Knowing
W = 0 −  and the differential cross section, we can compute the energy loss via eq. (2),
after changing variable from cos θ̃ to W . It is however instructive to deduce the stopping
power in the Thomson-limit with classical considerations.
.
4 2. Classical treatment. – In classical electromagnetism, the infinitesimal power dP
per unit solid angle carried by waves across the infinitesimal surface dA ~ (its direction
being the normal n̂ to the surface) can be written as

(35) ~ = S · n̂R2 dΩ ,
dP = S · dA

R being the distance from the radiating source, in terms of the Poynting vector S defined
as

E×B |E|2
(36) S= = Ŝ .
4π 4π

The Poynting vector thus represents the power per unit solid angle carried by waves
across the surface normal to their propagation direction. In the non-relativistic limit,
the radiated power P (which is a −dE/dt) is described by the Larmor formula

dP q 2 a2 sin2 φ 2
(37) = ⇒ P = q 2 a2 ,
dΩ 4π 3

φ being the angle between the normal to the surface and the acceleration vector. In fact,
eq. (37) is also valid in the relativistic case, provided that a2 is now replaced by the
four-acceleration (or, four-force over mass) squared, a2 → aµ aµ , m aµ = dpµ /dτ . It is
important to remember that the radiated power is a Lorentz-invariant, as one can infer
from the fact that dE and dt are both the 0-th components of (parallel) four-vectors.
Consider the interaction of an electromagnetic wave with a charged particle at rest,
and let us compute the power that this particle radiates. The Lorentz force for v → 0
gives an acceleration qE/m. For a sinusoidal wave E = E0 sin(ωt + φ) along some
Cosmic ray interactions with matter and radiation 15

direction, the Larmor formula gives for the (time-averaged) emitted power

2 2 2 2 q 4 E02
(38) hP i = q ha i = .
3 3 m2 2
In scattering theory, the cross-section is the ratio of the radiated power to incident flux,

hP i 8π q 4
(39) σ= = = σT (if q = ±e, m = me ) ,
|hSi| 3 m2

where the Poynting vector is |hSi| = E02 /(8π) and we used the fact that the electric and
magnetic field of the wave have the same amplitude.
We can also write
 2
|B|2 E2

|E|
(40) hP i = σT ũ , ũ = + = 0,
8π 8π 8π

in terms of ũ, the energy density in the e.m. field in the electron rest frame. The power in
eq. (40) can be interpreted as equal to the scattering rate (itself scattering cross section
times number density of photons in the field) times the average energy of the photons.
Since the power in eq. (40) is a relativistic invariant, to get the expression in the Lab
frame (where the electron is moving at β) it is enough to express the energy density ũ in a
manifestly invariant form. Remembering that [u] = [ × n], and that the number density
n transforms as a time since both dtd3 x (four-volume) and nd3 x (number of particles)
are relativistic invariants, we find that u transforms as the square of the energy,

β2
 
2 2 2
(41) ũ = uγ (1 − β cos α) =⇒ hũi = uγ 1 + ,
3

where the last step holds for an isotropic radiation field. The energy lost by the electrons
per unit time is the difference of the scattered power minus incoming power σT u, hence

β2
   
dE 2 4 4
(42) − = σT u γ 1 + − 1 = γ 2 β 2 uσT ' γ 2 uσT .
dt 3 3 3
.
4 3. Synchrotron radiation. – If u is interpreted as the energy density of both photons
and magnetic fields, eq. (42) describes not only inverse-Compton losses (in the Thomson
regime), but also synchrotron losses, i.e. the power radiated when evolving in a magnetic
field. This is due to the similar fundamental nature of both processes, see the right
diagram of Fig. 5, describing synchrotron radiation as a scattering onto a virtual photon
associated to an external magnetic field B (6 ). To convince oneself of that, one can
compute the power radiated according to Larmor’s formula by a charge q moving with

(6 ) Also bremsstrahlung can be considered Compton scattering of the CR electrons off the
virtual photons of the electrostatic fields of the background charges: The right diagram of Fig. 5
16 P. D. Serpico

Lorentz factor γ with respect to the frame where a the static magnetic field is present.
In the CR frame, the induced electric field and associated acceleration write

q
(43) E0 = −γv × B =⇒ a = − γv × B ,
m

hence the radiated synchrotron power is

2q 4 γ 2 2 2 2 4
(44) Ps = 2
v B sin ζ ⇒ hPs i = σT γ 2 u2B ,
3m 3

with the angle-averaged expressed (assuming isotropy) is equivalent to eq. (42), once
using u = B 2 /(8π)! For the equality to hold, it means that the frequency of the virtual
photons associated to B should verify the Thomson approximation. In order to check
that, we now estimate the energy of synchrotron photons emitted in the Lab frame,
verifying that they are indeed much smaller than the electron energy.
As a preliminary result, we remind the concept of beaming for relativistic motion.
The time dilation and space contraction by the factor γ in a (primed) frame moving at
velocity β with respect to the Lab one write

(45) t0 = γ(t − βx) , x0 = γ(x − βt) .

From these transformations one derives

dx γ(dx0 + βdt0 ) u0x + β

(46) = = ,
dt γ(dt0 + βdx0 ) 1 + βu0x

and

dy dy 0 u0y
(47) = = ,
dt γ(dt0 + βdx0 ) γ(1 + βu0x )

where we introduced the velocity components u0x = dx0 /dt0 , uy = dy 0 /dt0 . This implies
the aberration formula

uy u0y sin θ0
(48) tan θ ≡ = = .
ux γ(u0x + β) [γ(β + cos θ0 )]

This means that a photon emitted at θ0 = 0 travels at θ = 0, while a photon emitted

at θ0 = π/2 travels at sin θ ' θ ' 1/γ, as illustrated in Fig. 6.

is formally identical to Fig. 4 when neglecting the recoil of the target. As detailed in [26], the
differential cross section dσ to emit a photon within dk—in the so-called Weizsäker-Williams
approximation valid for soft photons—is dσ ' σT dw, where dw ' αdk/k is the probability to
emit a photon within dk, as seen by inspecting eq. (20). For a pedagogical introduction, see [31].
Cosmic ray interactions with matter and radiation 17

Fig. 6. – Illustration of the beaming effect in the lab (unprimed frame) for an isotropic emission
in the source (primed).

To compute the spectrum of photons emitted in a time-varying emission process, one

takes the Fourier transform of the signal (if this sound unfamiliar, see e.g. chapter 2
in [24]). In the non-relativistic limit, the circular uniform motion implies emission at
ωg qB me

 B 
the single frequency (cyclotron line) νg = 2π ≡ 2πm = 2.8 Hz Z m µG . In the
relativistic case, one must take into account that: i) the emission is observed only during
the fraction of the orbit within the beaming angle; ii) the timescale of the observer is
further different from the timescale of the emission; iii) the gyroradius rg = v⊥ /ωg is
replaced by the Larmor radius rL = γrg . As shown in Fig. 7, the time difference between
passage at point B and point A is given by

AB 2 rL
(49) ∆t ≡ tB − tA = = .
β γβ

For an observer well beyond B along the direction AB, the signal duration is

AB AB (1 − β 2 ) AB 1 rL
(50) ∆tobs = tB +δtprop
B −(tA +δtprop
A )= (1 − β) = ' 2
' 3.
β β 1+β β 2γ γ

As a poor man’s proxy for a Fourier transform, we can estimate the typical frequency of
this synchrotron radiation as the inverse of the above, hence
2
γ3

ωg B Ee
(51) νs ' = γ2 ⇒ Es ' 500 µeV .
rL 2π µG GeV

One can check that indeed the Thomson limit is well-respected for all energies of interest.
A precise calculation could be based on the Compton-scattered spectrum of the virtual
photons of the B−field, obtained by going to the electron’s rest frame. This approach is
however made cumbersome due to the need to transform into a non-inertial rest-frame.
Hence, the spectrum is usually computed via the classical Liénard-Wiechert potentials
seen by a distant observer, as sketched in Appendix D, based on chapter 3 of [24]. Such
a calculation shows that the power emitted per unit frequency scales as ∼ ν 1/3 at low
frequencies and as ν 1/2 e−ν/νc at high frequencies, with the critical frequency νc of the
18 P. D. Serpico

Fig. 7. – Diagram to estimate the frequency of the synchrotron radiation, accounting for the
beaming.

same order as the νs in eq. (51), which does provide a reasonable proxy for the actual
result.
Based on eq. (44), the power emitted from a power-law e spectrum φ ∝ Ee−α is

Z
Ps ∝ dEe B 2 Ee2 δ(νs − κB Ee2 )Ee−α

dPs B 2 Ee2 −α/2 α/2 1+α 1−α

(52) =⇒ ∝ νs B ∝ B 2 νs 2 .
dνs dνs /dEe

Hence, from the slope of the synchrotron radiation power ((1 − α)/2) one can infer the
slope of the parent electron distribution (−α). This is a very useful tool to infer remotely
the properties of non-thermal electrons at the sources, as opposed to the time and space
integrated flux of CR electrons measured at Earth. Note that the same dependence holds
for the the IC spectrum from a monochromatic photon background, if the scattering
happens in the Thompson regime. Also pay attention to the (usually more than linear)
dependence of the emitted power on the B-field strength: In cases where non-thermal
emissions at shocks could be imaged in the X-rays, this has been used to argue for
magnetic field amplifications in the so-called rims at shocks, see for instance [32].

5. – Hadronic interactions

In principle, hadronic CR can lose energy both via elastic and inelastic (7 ) collisions
with target nuclei. In practice, elastic scattering is usually disregarded: Given the CR
kinematics, elastic cross-sections peak in the forward direction and the CR energy loss

(7 ) By inelastic we mean collisions where new particles are produced or the internal structure
of the target nucleus or the projectile is changed.
Cosmic ray interactions with matter and radiation 19

Fig. 8. – The program of the second edition of the Varenna course, in 1954 [35] .

is negligible. Inelastic losses are however the dominant E-loss mechanism for nuclear
kinetic energies above a few hundreds MeV/nuc, and will be our focus in what follows.
Major particle physics discoveries in the cosmic rays between ∼ 1930 and mid-1950’s
(positron, muon, pion, strange hadrons) were due to inelastic hadronic interactions of
20 P. D. Serpico

CRs in the atmosphere. This was the central topic at the first gathering of this series
in Varenna, in 1953, with lectures of Patrick M. S. Blackett and Cecil Powell among
others [33]. It is also worth stressing how, the same year, the international cosmic ray
conference taking place at Bagnères de Bigorre, France, marked a split between research
in CR astrophysics and fundamental particle interactions, henceforth investigated more
effectively at colliders (for an account of that conference, see e.g. [34]). This “shift”
was clearly in the spirit of the time: The second edition of the Varenna course [35], the
1954 one whose scientific program was inaugurated by Fermi a few months before his
death, already saw a prominence of the accelerator approach to particle physics. The CR
aspects were relegated to Session IV and to some extent Session III of Part I, in which
the dominant role was played by B. Rossi (see Fig. 8).
The current fundamental theory for strong interactions, QCD among quarks and glu-
ons, is only perturbatively applicable to processes involving large exchanged momenta (8 ),
such as those studied at the LHC. In the processes of interests to us, quarks and gluons
within nucleons are not resolved; QCD is in its strongly coupled regime, and perturba-
tion theory is inapplicable. A more useful way to think about nuclear interactions is via
the effective Yukawa theory, involving nucleons and nuclei exchanging rather strongly
coupled massive mediators (the pions), responsible for the intensity and relatively short
range of the strong nuclear interaction. Since nuclear sizes are comparable or larger
than the range of the Yukawa force, cross sections (dominated by the inelastic channels
above GeV/nucleon) scale as the geometric cross-section of the nuclei. Nuclei of mass
number A have a radius R ' 1.2 A1/3 fm (with the ∼fm size of the proton essentially
linked to Λ−1 inel
QCD ), hence the inelastic cross sections of a nucleus η with a target t, σηt , is
inel 2/3 2/3
expected to be σηt ∼ π(Rη2 + Rt2 ) ∼ 45 (Aη + At ) mb, indeed close to proton-nuclear
data above 2 GeV [36]. Nuclear inelastic cross sections are weakly energy-dependent in
the GeV-TeV range of primary interest here: For an illustration, see Fig. 9, reporting a
number of inelastic cross sections vs. kinetic energy/nucleon implemented in the FLUKA
code (9 ). At lower energies, some energy dependence is present, for instance related to
thresholds, or manifesting a shallow minimum at few hundreds MeV/nuc and a rise at
tens of MeV/nuc. In general, in this range nucleus-specific details may matter.
An important energy benchmark is the threshold for pion production: The process
pp → ppπ 0 has a kinetic energy threshold Kπ0

m2π
(53) 2m2p + 2(Kp + mp )mp > (2mp + mπ )2 =⇒ Kp > Kπ0 = 2mπ + ' 290 MeV .
2 mp

The associated E-loss term for a CR of mass number A due to pion production can be

(8 ) That is, well above ΛQCD ∼ 300 MeV, the transmutation energy scale at which, loosely
speaking, the running coupling constant of the QCD gauge theory explodes above unity.
(9 ) http://www.fluka.org/fluka.php
Cosmic ray interactions with matter and radiation 21

103
Total inelastic σ (mb)

102

12 14
p-p C-p N-p
16 20 24
10 O-p Ne-p Mg-p
28
Si-p p-4He 4
He-4He
12
C-4He 14
N-4He 16
O-4He
24
20
Ne-4He Mg-4He 28
Si-4He
D-p p-40Ar 4 40
He- Ar
1 3 5
10-1 1 10 102 10 104 10
Kinetic Energy (GeV/n)

Fig. 9. – Total inelastic cross sections as a function of the energy per nucleon of the incoming
projectile. The plot shows the cross sections for all the projectile-target pairs studied in [37].
Reproduced with the permission of the authors.

approximated as [38]
    E 1.28  E −0.2
dE −16
n
gas
(54) − = 3.85 × 10 A 0.79
× + 200 GeVs−1 ,
dt cm−3 GeV GeV

according to the re-assessement of modern parameterization of pion production in [39].

Below the energy (per nucleon) given in eq. (53), the only inelastic processes that
matter are spallation ones, where target and/or projectile nuclei are broken into multiple
nucleon/nuclear fragments. Note that these processes are the leading source of nuclei
such as Li-Be-B, which are fragile from the thermonuclear point of view: Due to their
small nuclear binding energy, they are easily burned in stellar thermonuclear processes
and are only present in traces in conventional astrophysical environments, such as the
ISM. Their sizeable presence among CR nuclear fluxes is interpreted as the consequence
of spallation of heavier CR common in the ISM, such as C or O, onto the ISM gas during
their propagation. Such nuclei are thus also referred to as “secondary” CR species. Since
single-nucleons are rather common among fragments, spallation processes also contribute
as a non-negligible secondary source term for protons.
Due to theoretical limitations, one must rely on semi-empirical formulae to describe
not only the inelastic cross sections, but also the relevant exclusive differential cross-
22 P. D. Serpico

sections to produce the secondary α in the collision of the primary η with the target t.
They can be written as function of kinetic energies of the primary Kη and product K as

dση+t→α inel dNη+t→α

(55) (K, Kη ) = σηt (Kη ) (K, Kη ) ,
dK dK

with dN /dK being the multiplicity spectrum of the species α in the collision of η with
t. Since data are typically scarce, regularities motivating semi-empirical formulae turn
out to be useful in interpolating between and extrapolating beyond measurements, or
to estimate cross-sections involving nuclei for which no measurement exists. Calibrated
models and codes for the nuclear cross sections libraries are implemented and described
in propagation codes, such [10, 40, 41]. Some relevant cross-sections and branching ratios
are also reported in printed form in Tab. 10.1 of [42].
.
5 1. Examples: Spallations, hadronic gamma-rays, antiprotons. – As an example of
some of these regularities, well away from thresholds, dN /dK is prominently a function
of x ≡ K/Kη but is only weakly dependent from K, and depends on η and t mostly via a
normalisation (equivalent to the branching ratio κ). To a good approximation, estimated
to a 5-6% level for B/C in [43], in spallation reactions the kinetic energy per nucleon is
conserved, so that
 
dNη+t→α K Kη
(56) (K, Kη ) ' κη+t→α δ − .
dK Aα Aη

It is worth mentioning that the projectile or target nucleus can end up in an unstable
state and de-excite via gamma-ray emission, at typical energies of few MeV for nuclei at
rest. Although this has long been recognized as a potential exquisite diagnostic tool for
the study of energetic phenomena [44], the observational challenges in MeV gamma-ray
astronomy make this field still in its (relative) infancy, at least for CRs in the ISM.
While inelastic processes must be taken into account for precision calculations of
hadronic CR spectra, at sufficiently high energy (typically above tens of GeV/nuc) they
have a sub-leading role for primaries compared to collisionless propagation effects, as
.
it will be illustrated in a simplified propagation model in Sec. 7 2. Nonetheless, at all
energies they are of crucial importance as source terms not only for secondary CR nuclei,
but also for hadronic gamma-rays, neutrinos, and antinuclei.
As a specific example of this link, let us focus on the hadronic photons, i.e. essentially
those produced via the neutral pion decay process, π 0 → γγ. The photons are back-to-
back in the π 0 frame, each carrying an energy equal to mπ /2 ' 67.5 MeV. If the pion
moves at β, in the Lab frame the energy of the photon is Eγ = mπ γ(1 + β cos θ)/2, θ
being the angle of the emitted photons with respect to the the direction of flight of the
pion. Hence, the maximal and minimal energy of the photons obey

mπ m2
(57) max
Emin = γ(1 ± β) =⇒ Eγmax Eγmin = π , Eγmax + Eγmin = Eπ .
2 4
Cosmic ray interactions with matter and radiation 23

There is a one-to-one correspondence between Eγ and θ, whose law is

mπ
(58) dE = γβd cos θ .
2
Since the pion is a scalar particle, its decay products are emitted isotropically, hence
we can easily deduce the energy distribution of the photons:

dN 1 1 dN 1 1 1
(59) = =⇒ dN = d cos θ =⇒ = = =p .
dΩ 4π 2 dE mπ γβ Eπ β Eπ − m2π
2

This distribution is flat (box-shaped) in energy space between Emin ' 0 and E max ' Eπ
in the relativistic limit. In log-energy space,

1 √ m 
π
(60) (log E min + log E max ) = log E min E max = log ,
2 2
i.e. the center of the interval is half the pion mass, independently of the pion energy
distribution, hence of the parent nucleon distribution. Adding an arbitrary number of
box-shaped spectra, each centred at mπ /2, implies that this is the maximum of the
spectrum. This is dubbed “pion bump” and considered the cleanest (albeit hard-to-
detect!) signature of hadronic origin of a gamma-ray spectrum [45, 46].
Let us assume for simplicity a pure hydrogen target of density n and relativistic
projectiles and products so that K ' E. A useful approximation for the spectrum of
pions from a proton CR interaction is

dNp+H→π
(61) (E, Ep ) ' ζπ δ(E − cπ Ep ) ,
dE
where the pion multiplicity ζπ and the average fraction of the proton energy into a pion,
cπ , are weakly dependent functions of energy. The pion source term writes
Z    
dNp+H→π inel n ζπ inel E E
(62) qπ (E) = n dEp σpp (Ep )φp (Ep ) = σ φp ,
dE κπ pp κπ κπ

inel
where σpp (Ep ) is the proton-proton inelastic cross section. For more advanced treat-
ments, see e.g. [39] or [47].
To a first approximation, the pion source spectrum at Eπ is thus proportional to the
proton flux at an energy Ep = Eπ /κπ (with sizable corrections due to the cross section
energy dependence, relevant close to threshold). The photon spectrum is then
Z ∞ Z ∞
dN q(Eπ )
(63) qγ (Eγ ) = 2 dEπ q(Eπ ) = 2 m2
dEπ p ,
min (E )
Eπ γ
dE Eγ + 4Eπγ Eπ2 − m2π

where the minimum energy of the pion to produce a photon of energy Eγ , Eπmin (Eγ ), is
given by the relation linking the maximal energy of a photon produced by a pion of Eπ .
24 P. D. Serpico

For a given energy Eπ , Eγmax and conversely Eπmin (Eγ ) are given by

m2π m2
(64) Eγmax = Eπ − Eγmin = Eπ − max
, =⇒ Eπmin (Eγ ) = Eγ + π .
4Eγ 4Eγ

103

102
<nsec(>1MeV)> × σinel (mb)

10

p-p
π-
1
π+
πo
10-1 -1 3
10 1 10 102 10 104
Kinetic Proton Energy (GeV)

Fig. 10. – Inclusive cross sections for the production of π 0 (blue), π + (red) and π − (green) in
p−p collision as function of the incoming proton kinetic energy. Lines: FLUKA simulation; points:
data from Ref. [48]. From [37], reproduced with the permission of the authors.

It is worth noting that each process producing a π 0 is associated to a process producing

charged pions: For the pp process, at CR energies one is often well above threshold,
multipion production is frequent, and isospin symmetry yields equal numbers of π 0 , π + ,
and π − , see Fig. 10. Also, the energies of the gammas and the neutrinos are within a
factor of ∼two of each other: each of the there neutrinos emitted in a charged pion plus
following muon decay carries about 1/4 of the energy of the parent pion, while photons
exactly a fraction 1/2 of it. We thus expect a link between neutrino and (hadronic)
photon spectra at emission given by
 
1 dNγ 1 dNν
(65) ' ,
2 dEγ 6 dEν Eν =Eγ /2

which has been (and still is) extremely useful to estimate neutrino target fluxes as coun-
terparts of observed photon sources (assumed hadronic) as well as in constraining inter-
pretations of the diffuse (probably extragalactic) neutrino flux observed by IceCube (for
a review, see [49]) based on gamma observations. Further details on neutrino fluxes can
be found in Denise Boncioli’s lectures.
Cosmic ray interactions with matter and radiation 25

Another important class of secondaries are antinucleons. Currently, only antiprotons

have been measured (rather precisely) in CRs, and their spectrum is consistent within
errors with expectations from secondary production [50]. Their production cross-sections
cannot be computed from first principle either, and are typically parameterised as
Z ∞
dσ 1
(66) = 2πpp̄ (Kp̄ ) dη σinv ,
dKp̄ ηmin '0 cosh2 (η)

where η = − ln[tan(θ/2)] is the pseudo-rapidity, defined in terms of the scattering angle

θ, and

d3 σ √

(67) σinv ≡ E 3
s, xR , pT
dp
√
depends only on manifestly Lorentz invariant quantities: the CoM energy s, the trans-
verse momentum of the produced antiproton (pT ) and the ratio of the antiproton energy
to the maximally possible energy in the CoM frame (xR ). Since in all cases known to
inel
the author one factorises σinv = σpp (s) × F (s, pT , xR ), and (in the case of proton-proton
collisions) s = 2mp (Ep + m), eq. (66) is ultimately equivalent to a rewriting of eq. (55).
Several parameterisation have been proposed for antiproton cross-sections in the past
decade, benchmarked to a growing dataset of collider data [51, 52, 53, 54, 55], with
current uncertainties still among the dominant contributions to the total error budget
of secondary CR antiproton fluxes [50]. Additional cross-sections inputs concern the
antiproton energy-losses: Customarily, the “sink” term includes the total inelastic an-
tiproton cross-section. Then, a so-called tertiary source term enters, which is due to
inelastically scattered secondary antiprotons. Its differential cross section is typically
parameterised as
inel ann
dσp̄+p→p̄ σp̄+p − σp̄+p
(68) (K, Kp̄ ) = .
dK K
References containing fits to the relevant cross sections can be found e.g. in [40].
Antiprotons are an important diagnostic tool especially for indirect searches aimed
at identifying the nature of the still mysterious dark matter pervading the universe (for
a recent study and assessment of the literature, see [56]); hopefully these searches will
be extended in the near future to light antinuclei such as antideuteron [57] and antihe-
lium [58, 59], with peculiar kinematical aspects leading to spectra features advantageous
for exotic searches. For instance, antideuteron production in secondary reactions requires
an energy of at least 17 mp (just apply the consideration developed after eq. (4) to the
process pp → pp̄nn̄pp). This makes an antideuteron low-energy component (which does
not suffer from the Lab frame kinematical suppression above) an exquisite target for
CR-based searches of DM, notably with the innovative GAPS experiment based on X-
ray de-excitation of exotic atoms [60]. Important uncertainties remain in modeling the
production of CR antinuclei both in conventional and exotic processes, an area of fervent
activity today (see e.g. [61, 62]) which we cannot make justice to here.
26 P. D. Serpico

e+
γ

e−
Fig. 11. – Diagram for the QED pair production process.

6. – Photon interactions

We have seen how loss processes of charged particles have their counterpart as source
processes for non-thermal photons. Synchrotron photons associated to GeV to TeV elec-
trons typically fall in the radio to X-ray band (see eq. (51)), while inverse Compton
photons are typically from the soft to very high-energy gamma band (see eq.s (29, 30)).
Similarly, photons from pion decay also fall in the gamma-ray band (see eq. (63)). These
photons are important messengers of their parent charged particles, since they retain
directionality and can be associated way more easily to astrophysical sources. It is im-
portant to realise, however, that photons as well are subject to scattering and absorption.
This process due to the interstellar gas and especially dust (onto which photons can co-
herently diffuse via Rayleigh scattering) is typically known as extinction in low-energy
astronomy and perhaps familiar to the reader. In the extreme UV to X-ray range, photo-
ionisation processes are relevant, but neither of these regimes will be dealt with here. At
very high energies, interactions with interstellar and a fortiori intergalactic matter are
negligible, but the interaction with the radiation backgrounds are increasingly impor-
tant. The most important process is pair production, see Fig. 11, by a photon with high
energy E impinging on a background one of energy . The threshold for the reaction
γ + γ → e+ e− in the Lab is

th m2e
(69) 4Eγγ  = (2me )2 =⇒ Eγγ
th
= .


The cross-sections behaves as

  
3πσT 1+β
1 − β 2 2β β 2 − 2 + 3 − β 4 ln




(70) σγγ (β) = ,
16 1−β

where β is either lepton velocity in the CoM frame; σγγ peaks at about σT /4 at about
twice the threshold energy. It is easy to check that the e± pair production mechanism
constitutes a serious limitation to measure photons from the remote extragalactic sky
(beyond redshift z ∼ 1) already at ∼ 100 GeV, for the near extragalactic sky (beyond
z ∼ 0.1) at ∼10 TeV, and even for Galactic objects in the PeV range: Beyond the TeV
Cosmic ray interactions with matter and radiation 27

range, we can only perform astronomy in the “local neighborhood”, which is one of the
motivations to develop neutrino astronomy at larger energies.
An interesting feature due to collisional effects involving γ’s and e± is that extra-
galactic photons initiate electromagnetic cascades: After the first pair-production, typ-
m2e
ically on EBL photons, affecting photons with Eγ > Eγ ≡ EBL ' 390 GeV, the e±
scatter via inverse Compton onto the CMB resulting into highly energetic photons (no-
tably at early stages in the Klein-Nishina regime, see eq. (30)), which undergo the
same multiplicative process as long as Eγ > Eγ . Another characteristic energy is
EX ≡ 13 Eγ CMB
EBL
' 1.2 × 108 eV , corresponding to the upscattered photon energy (in
the Thomson regime) associated to minimum-energy e± produced by photons at the
threshold Eγ . Below EX , the spectrum is solely due to the further energy losses of
“cascade-sterile” e± via inverse Compton.
While the exact spectral shape of the diffuse γ-ray flux depends on the cosmic evo-
lution and the distance to the sources, in the limit of fully developed cascades, one can
derive analytically an approximate, universal spectral shape for the resulting diffuse γ-ray
flux [63, 64], given by:


−3/2
 (Eγ /EX )
 at Eγ ≤ EX
Es −2
(71) φγ (Eγ ) = 2 × (Eγ /EX ) at EX ≤ Eγ ≤ Eγ
EX (2 + ln Eγ /EX )  0 at Eγ > Eγ

where the normalisation is given in terms of the total injection energy Es . After all the
scatterings, no directional memory remains of the initial photons, but the quasi-isotropic
extragalactic gamma-ray flux below the TeV range puts an upper limit to any extragalac-
tic photon source at supra-TeV energies. This has been used in countless astroparticle
applications. For instance, given the expected link between the spectrum of neutrinos
and the hadronic gamma-ray spectrum (both originating via pion decays) at the source,
this has been used in the literature to raise the possibility that most of the sources con-
tributing to the IceCube flux are actually opaque (i.e. the associated gamma-rays are
degraded to MeV or even thermal energies before escaping the sources). For a recent
calculation in that sense, see [65] and refs. therein.

7. – Applications to cosmic ray propagation

In the hybrid (non-inertial) plasma frame, where positions are measured in the Lab
(Galaxy) frame but particle momenta are measured with respect to where magnetic
inhomogeneities are statistically at rest, the ensemble average (over realisations of inho-
mogeneities) and angle average of the single particle distribution function φα for a CR
28 P. D. Serpico

species α obeys (sum convention over repeated latin indices)

   
∂φα ∂ ∂φα ∂φα 1 ∂ui ∂φα 1 ∂ ∂φα
(72) − Dij + ui − p − 2 p2 D =
∂t ∂xi ∂xj ∂xi 3 ∂xi ∂p p ∂p ∂p
   
1 ∂ dp X
q+ 2 p2 φα −Γtot,α φα + φη ⊗ Γη→α ,
p ∂p dt ` η

where the LHS contains transport terms whose origin is collisionless (from second to
fifth terms: spatial diffusion with diffusion tensor Dij , convective term associated to the
plasma velocity field ui (x), adiabatic energy loss/gains, and reacceleration or magnetic
inhomogeneity diffusion in momentum space, controlled by D), while the RHS, besides
a possible primary source term q, contains terms describing continuous losses (in blue)
and catastrophic sinks (in red) or sources (in magenta). Catastrophic sinks include both
decays if the species is unstable with lifetime at rest τdec (one should account for the
boost γ of the CR in the Galactic frame), and inelastic interactions over all possible
targets t of the ISM with density nt , so that

1 X
inel
(73) Γtot,α = α + βσαt nt .
γτdec t

In practice, only hydrogen and helium have significant densities to be relevant ISM
targets. The last term in eq. (72) represents in general an integral operator acting on
φβ , which is more easily expressed in terms of the kinetic energy K = E − m rather than
momentum variables (10 ), so that

Z
X dση+t→α
(74) φη ⊗ Γη→α = nt dKη β (K, Kη )φη (Kη ) .
t
dK

Since each term in eq. (72) has the dimension of φ/time, one can define character-
istic timescales (analogous to eq. (3)) which allow for a quick parametric assessment of
their importance. Public codes exist that allow for a numerical (such as GALPROP (11 )
DRAGON (12 )) or semi-analytical (USINE (13 )) solution of the problem. It is important
however to grasp the key features of these terms via some analytical, limiting solution,
notably of the steady state problem. We outline two examples below, which have the
merit to show the interplay of fundamental physics properties of the collisional processes
with CR phenomenology.

(10 ) Note that φ(K) = φ(p(K))dp/dK = β −1 φ(p(K)), since p2 = (m + K)2 − m2 = 2mK + K 2 .

(11 ) https://galprop.stanford.edu
(12 ) https://github.com/cosmicrays/
(13 ) https://dmaurin.gitlab.io/USINE/
Cosmic ray interactions with matter and radiation 29

.
7 1. Continuous E-loss dominated propagation. – If continuous energy loss timescales
are the shortest ones (or the only one of relevance, in “quasi-homogeneous” problems),
the steady state equation approximates to
   
1 ∂ dp
(75) − 2 p2 φα = q ,
p ∂p dt `

and is solved by
Z p
1
(76) φ(p) ∝ − 2 dp0 q(p0 ) p02 ,
p (dp/dt)`

which, for q ∝ p−s and (dp/dt)` ∝ −p` , leads to

(77) φ(p) ∝ p−s−`+1 .

Namely, the steady state spectrum is ` − 1 softer than the injected one. It turns out
that for CR leptons the continuous E-loss dominance is close to truth, with ionisa-
tion and Coulomb losses (` ' 0) dominating at low-energies, eventually overcome by
bremsstrahlung energy losses and Compton and Synchrotron energy losses (` ' 2) finally
taking over. As a result, as a first approximation we expect the low-E (sub-GeV) electron
spectrum observed at the Earth to be one power harder than the source one, that the
spectrum matches closes the source one at intermediate energies (few GeV), while being
one power steeper than the injected one at high-energies. Of course, diffusion and other
transport terms modify these basic features, which remain however qualitatively correct.
.
7 2. Catastrofic loss for diagnostics: Secondaries over primaries. – Let us compare
secondaries, i.e. nuclei only produced by spallation during propagation, such as the
above-mentioned Boron, Lithium, and Beryllium, with primaries. For simplicity, let us
assume that primary CR sources are confined to a thin disk of half-width h, practically
infinitesimal if compared to the diffusive propagation halo half-thickness H. We are then
reduced to a 1D problem is considering the disk as infinite, translating the fact that the
radial extent of the Galaxy is large compared to its vertical extension. If we include just
the dominant diffusive transport operator, assumed isotropic, the steady state transport
equation for a purely primary species reduces to

∂φP
 
∂
(78) − D = 2 q0 (p) hδ(z) − 2h Γσ φP δ(z) .
∂z ∂z
2 P
At z 6= 0, eq. (78) reduces to ∂∂zφ2 = 0, whose solution is φP (z, p) = a(p) + b(p)|z|. If we
denote with φP P P
0 the solution in the plane φ0 = φ (0, p), the vanishing at the boundary
|z| = H gives
 
|z|
(79) φP (z, p) = φP
0 (p) 1 − .
H
30 P. D. Serpico

An equation for φP
0 (p) can be found by integrating Eq. (78) over a small interval around
z = 0,

∂ φP
(80) −2D(p) = 2h(q0 (p) − Γσ φP
0) .
∂z 0

We are further assuming the diffusive halo to be homogeneous, although K can depend
on p. Using Eq. (79) one finds
−1
(81) φ0P (p) = q0 (p)τeff (p) , where τeff (p) = τd−1 (p) + τσ−1 (p) ,

where we defined a diffusive timescale τd

Hh H h 1028 cm2 s−1

(82) τd (p) ≡ ≈ 107 yr ,
D(p) 3 kpc 100 pc D

and a catastrophic loss timescale τσ

1 cm−3
  
100 mb
(83) τσ (p) ≡ Γ−1
σ ≈ 10 yr7
.
nISM σ

Note that, as long as D(p) is a growing function of p, the diffusive timescale τd domi-
nates over collisional losses at sufficiently high energies, since Γσ ' σ n has very little
dependence on p, as we described. This remains true in more general models: Collisional
effects are more relevant in shaping hadronic CR fluxes at low energies.
The distribution of secondaries in the plane, φS (p), is sourced by the injected nuclides
per unit time, i.e. q0 (p) → φP ΓP →S , φP being the primary population. Hence we obtain
the solution for the ratio of primary to secondary distribution in the form

φS (p) Hh
(84) ' ΓP →S τeff,P ' ΓP →S ,
φP (p) D(p)

where the second relation holds if collisions are subdominant with respect to diffusion
(which is not true at low energies!). Since, at least in principle, ΓP →S h can be inferred
by independent means, from this ratio (for instance, Boron-to-Carbon ratio in CRs) one
can gauge the value of the “diffusive” ratio H/D, including its energy dependence.
For the secondaries, one has approximately (remember the previous discussion around
eq. (74))

∂φS
 
∂
(85) − D ' 2h ΓP →S φP δ(z) − 2h ΓS φS δ(z) ,
∂z ∂z

which is formally analogous to eq. (78); the z dependence is thus the same, while the
p-dependence in the disk can be obtained via the replacement q0 h → φP
0 ΓP →S , i.e.

(86) φS0 (p) = φP

0 (p)ΓP →S τeff,S (p)
Cosmic ray interactions with matter and radiation 31

where

−1
(87) τeff,S (p) = τd−1 (p) + ΓS .

Hence we finally obtain

φ0S (p)
(88) = ΓP →S τeff,S (p) .
φ0P (p)

Note that the RHS of eq. (88) depends on quantities that are in principle measurable in
the lab, the diffusion timescale and the target density in the ISM, while the LHS comes
from CR measurements at the Earth, allowing one to infer τd . This is the essential
method through which one can break the degeneracy between source parameters and
propagation ones, allowing one to infer source spectra via e.g. eq. (81). The uncertainties
with which hadronic cross-sections are known limits however the precision with which
propagation parameters can be inferred, even in this very simple setup, and do introduce
some correlation between “hadronic” parameters and diffusion ones. These uncertainties
are nowadays determining the ultimate capacity to test propagation frameworks, and
must be reduced in the future if one is to fully exploit the high-precision data currently
collected by experiments like AMS-02 (14 ).

∗ ∗ ∗

The author thanks the organisers for their kind invitation to lecturing in such a
prestigious and fascinating location, where history and science, as well as natural and
artistic beauty meet.

Appendix A.

Rutherford scattering

Rutherford scattering can be defined as the classic problem of elastic scattering of

charged particles by the Coulomb interaction: The target (of charge Zt ) does not recoil
and is the center of the potential. The problem obeys cylindrical symmetry around the
incoming projectile direction. Given the Coulomb potential of a target and the initial
kinetic energy K (at infinity) of the incoming particles of charge Zp , the scattering angle θ
is uniquely determined by the impact parameter b. The cross section relates the number
of particles scattered into dΩ = 2π sin θdθ to the number of particles with associated
impact parameters between b and b + db,

dσ dσ b db
(A.1) 2πbdb = dΩ =⇒ = .
dΩ dΩ sin θ dθ

(14 ) https://ams02.space
32 P. D. Serpico

Solving the EoM in polar coordinates with the target at the center (r = 0)

d2 u Zp Zt e2
(A.2) + u = − , (Binet equation)
dθ2 2 K b2

with u = 1/r subject to boundary conditions u → 0, r sin θ → b when θ → π, one derives

b(θ) = Zp Zt e2 cot(θ/2)/(2K). Using that, one infers

dσ Zp2 Zt2 α2
(A.3) = .
dΩ 16K 2 sin4 θ/2

The change of momentum ∆p = 2Zp2 e2 /(b v) is thus associated to the change of kinetic
energy W = (∆p)2 /(2me ) = 2Zp2 e4 /(b2 v 2 me ), which allows one to compute the stopping
power via eq. (2).

Appendix B.

Maximum energy transfer in 2-body scattering

Here we derive the maximum energy transfer in a 2-body scattering of a particle of

mass M and momentum p on a target particle (eg. an atomic electron) that is at rest
(Lab frame), of mass me . For simplicity, below we will indicate moduli of 3-momenta
in Italic, i.e. |p| = p. Conservation of energy before and after the scattering requires (a
prime denotes the target electron outgoing momentum, and double prime the projectile
outgoing momentum):

p p p p p
(B.1) p2 + M 2 + me = p002 + M 2 + p02 + m2e ⇒ p2 + M 2 = p002 + M 2 + W ,

where the latter follows from the definition of kinetic energy for the electron, W = E−me ,
and the relativistic dispersion relation

(B.2) p02 = (W + me )2 − m2e .

We can rewrite eq. (B.1) isolating p002 as

p
(B.3) p002 = p2 + W 2 − 2W p2 + M 2 .

The momentum conservation implies (θ being the angle between the scattered electron
and the CR incoming direction)

(B.4) p00 = p − p0 ⇒ p002 = p2 + p02 − 2p p0 cos θ .

Cosmic ray interactions with matter and radiation 33

Now, we can replace p0 from eq. (B.2) and p00 from eq. (B.3) in eq. (B.4) getting
p p
−2W p2 + M 2 = 2W me − 2p W 2 + 2me W cos θ
p
p me + p2 + M 2
⇒ 1 + 2me /W =
p cos θ
2m p cos2 θ
2
(B.5) ⇒W = p e
(me + p2 + M 2 )2 − p2 cos2 θ

The maximum energy transfer, i.e. the maximum of W , is obtained for cos θ = 1 (head-on
collision, the electron recoils along the incoming direction of the particle):

2me p2 2me β 2 γ 2
(B.6) Wmax = p = m2e
→ 2me β 2 γ 2
m2e + M 2 + 2me p2 + M 2 1 + 2m
M
e
γ + M2

where the last limit holds when M  me . Eq. (B.6) is the relation presented in the main
text.

Appendix C.

From QED to Rutherford

The notions recalled here are treated in an extensive way in any introductory particle
physics and quantum field theory textbook, such as [66, 67, 68]. The interaction proba-
bility in a two-body process 1 + 2 → f1 + . . . fn is quantified via the cross-section σ. It
can be written in a manifestly Lorentz-invariant form as

|M|2
(C.1) dσ = dQ ,
F

where the phase-space factor dQ (imposing 4-momentum conservation) writes

 
fn
X Y d3 pf
(C.2) dQ = (2π)4 δ 4 p1 + p2 − pf  ,
(2π)3 2Ef
f f =f1

while the flux factor F writes

q
(C.3) F = 4 (p1 · p2 )2 − m21 m22 = 2E1 2E2 |v1 − v2 | .

Finally, M is the initial to final transition matrix element (computed according to stan-
2
dard Feynman rules). If polarisation is not measured, as it is typically the case, |M| in
eq. (C.1) is intended as a sum over all final spin states and an average over initial ones
and denoted as |M|2 . By the way, the differential decay rate of a particle 1 also writes
in the form given by eq. (C.1), with the flux factor however simply reducing to 2E1 .
34 P. D. Serpico

If the final state is made of two particles, 3 and 4, one has

|M|2 d3 p4 4
(C.4) dσ = β 3 p3 dp 3 dΩ3 δ (p1 + p2 − p3 − p4 ) ,
8π 2 F 2E4

an expression one can integrate over the “uninteresting” variables to get the differential
(or total) cross section of interest. In doing that, the following identity is often useful

d3 p0 4
Z Z
(C.5) δ (P − p0 ) = d4 p0 δ 4 (P − p0 ) Θ(E 0 )δ(p02 − m02 ) .
2E 0

Further manipulations require expressing |M|2 and F in a specific frame. In the CoM
frame, for instance, it is a textbook exercise to prove that

2
dσ |M| p0
(C.6) = 2 ,
dΩ CoM 64π 2 ECoM p

where p, p0 are the initial and final momentum of either of the scattered particles. In
the Lab frame, one has F = 4β E M , with E (β) the energy (respectively, speed) of the
projectile and M the mass of the target.
Let us express |M|2 for the case of the scattering e + µ → e + µ, indicating with
a prime (unprimed) symbol the final (initial) momenta, using E, p, m for the electron
energy, momentum, and mass and k, M for the muon momentum and mass, respectively.
This is clearly a proxy also for the electron-proton scattering, if we neglect the non-
elementary nature of the proton and focus on it as an electromagnetically interacting
spinor, much more massive than the electron. Also, here SI denotes the SI charge, to
comply with the standard QFT convention, so that e2SI = 4πe2 = 4πα. One has
(C.7)
8e4 
|M|2 = 4SI (k 0 · p0 )(k · p) + (k · p0 )(k 0 · p) − m2 (k 0 · k) − M 2 (p0 · p) + 2m2 M 2 .

q

If we specify eq. (C.7) to the frame where the heavy particle is initially at rest, in the
relativistic limit for the electron, one has q 2 = −4E 0 E sin2 ϑ/2, ϑ being the angle of the
scattered electron with respect to its incoming momentum,

8e4SI q2
 
2 0 2 2
(C.8) |M|2 → 2M E E cos ϑ/2 − sin ϑ/2 ,
q4 2M 2

and eq. (C.5) applied to the muon momentum reduces to (2M A)−1 δ(E 0 − E/A) where
A = 1 + (2E/M ) sin2 θ/2. This is equivalent to the kinematical relation following from
conservation of energy and momentum, used to express the final electron energy in terms
of the initial one. When combining these relations, after integration over E 0 one obtains

α2 cos2 θ/2 q2
 
dσ 1 2
(C.9) = 1− tan ϑ/2 .
4E 2 sin4 ϑ/2 1 + 2E 2
 
dΩ Lab M sin ϑ/2
2M 2
Cosmic ray interactions with matter and radiation 35

Eq. (C.9) beautifully illustrates a number of notions: First of all, in the limit M 2 → ∞
only the first factor survives, i.e. one recovers Mott scattering, namely the Rutherford
cross-section (15 ) corrected by the cos2 ϑ/2 factor which accounts for the spinorial nature
of the electron. The factor in brackets, going to 1 in the limit M 2 → ∞, accounts instead
for the recoil of the target. Finally, the factor in curly brackets, also going to 1 in the
limit M 2 → ∞, accounts for the spin of the target: The sub-leading term suppressed
by q 2 /M 2 accounts for the scattering via the magnetic moment. This is analogous to
the effect of the electron spin: The term cos2 ϑ/2, truly given by 1 − β 2 sin2 ϑ without
relativistic approximation, reflects the scattering on the electric charge −eSI (the “1”
term) and the magnetic moment −eSI /2m, respectively. This can be seen explicitly
via the so-called Gordon decomposition, amounting to rewrite the spinor current ūγ µ u
as (2m)−1 ū[(p0 + p)µ + iσ µν (p0 − p)ν ]u. The first piece in this decomposition, if used
to compute spinorial amplitudes, leads to the same results as a “scalar” electron with
vertex ieSI (p + p0 )µ instead of ieSI γ µ , and no spinor factors u associated to the particle
currents.

Appendix D.

Derivation of the synchrotron spectrum in its fully glory.

A detailed treatment of this topic can be found in [24] and it is only sketched below.
As discussed around eq. (36), the power per unit solid angle carried by waves across
the surface normal to their propagation direction is given by the Poynting vector. The
emitted spectrum (energy per unit frequency and solid angle) of an accelerated charge
can be thus expressed as Fourier transform of the modulus of eq. (36), leading to

2 ∞ 2
e2 ω 2
Z Z
dW 1 iωt ~ iω(t0 −n̂·~r(t0 )) dt0
(D.1) = <E(t) e dt = n̂ × (n̂ × β)e ,
dωdΩ 4π 2 4π 2 −∞

where the expression at the RHS comes from the differentiation of the Liénard-Wechart
potentials (whose details can be found in Sec. 14.1 of [2]), n̂ is the line of sight unit vector,
β~ is the particle velocity and t0 is the retarded time accounting for the movement of the
source, whose trajectory is described by ~r (t0 ). This expression follows from changing the
integration variable to t0 and assuming the observer far from the source. We remind the
reader that we denote the pitch angle (angle between particle velocity and magnetic field)
with ζ and that the particle trajectory is characterised by the Larmor radius rL = γrg .
Let us assume that the particle is at the origin of the reference system at t0 = 0,
0
~r (t = 0) = 0, with the magnetic field along z and the velocity along x, denoting with
θ the angle between the velocity and the line of sight. As evident from eq. (D.1), the
emission is perpendicular to the line of sight, and it is conveniently decomposed in one
component parallel (ˆ k ) and one perpendicular (ˆ⊥ ) to the magnetic field direction. Since

(15 ) Note that Rutherford cross section can also be recovered from eq. (C.6), if we use ECoM →
M , p = p0 , and |M|2 → (2m)2 (2M )2 /q 4 , with the 2m being the non-relativistic normalisations
of the fermionic states.
36 P. D. Serpico

after a time t0 the particle has covered an angle vt0 /rL , we have

vt0 vt0
   
(D.2) ~ = −ˆ
n̂ × (n̂ × β) ⊥ sin + ˆk cos sin θ
rL rL

and

vt0
 
(D.3) t0 − n̂ · ~r(t0 ) = t0 − rL cos θ sin .
rL

Hence we can rewrite eq. (D.1) as

(D.4)
Z ∞  0  0   2
αω 2
 0
dW vt vt

0 iω t0 −rL cos θ sin( vt
rL )
= dt ˆk sin θ cos − 
ˆ ⊥ sin e .
dωdΩ 4π 2 −∞ rL rL

It makes sense to split this expression as follows

   
dW dW dW
(D.5) = + .
dωdΩ dωdΩ ⊥ dωdΩ k

Let us introduce θγ2 ≡ 1+θ2 γ 2 , which turns out to control the angular dependence, the
new variable of integration y = γt0 /(aθγ ), and the new variable η ≡ ωrL θγ3 /(3γ 3 ) which
retains the complete frequency-dependence. For relativistic particles β ≈ 1 (γ  1) the
observer receives the signal only when vt0 /rL ≈ 0 and θ ≈ 0 (remember Fig.s 6 and
7), so that we can expand the trigonometric functions accordingly, and also consider
η ≈ η(θ = 0) = ω/(2ωc ), where ωc ≡ 3γ 2 ωg sin ζ/2. Then eq. (D.2) yields

!2 Z  2
!2
rL θγ2 ∞ rL θγ2
αω 2 αω 2
  3
dW

3
2 iη y+ y3 2
(D.6) = dy y e = K2/3 (η) ,
dωdΩ ⊥ 4π 2 γ2 −∞ 3π 2 γ2

2 Z ∞  2 2
αω 2 θ2 αω 2
   3

dW rL θγ rL θγ

3
2 iη y+ y3 2
(D.7) = dy e = K1/3 (η) ,
dωdΩ k 4π 2 γ −∞ 3π 2 γ

where Ki (x) are the modified Bessel functions of order i.

We can now integrate these relations over the solid angle, obtaining in this way
the energy emitted per unit frequency for a complete orbit on the plane normal to the
magnetic field. During one such orbit the emitted radiation is almost completely confined
to the solid angle lying within an angle 1/γ of a cone of aperture twice the pitch angle
ζ. Thus dΩ = 2π sin ζdθ and

∞
√
2αω 2 rL
2
  Z
dW sin ζ 3αγ sin ζ
(D.8) ' 4
dθ θγ4 2
K2/3 (η) = [F (x) + G(x)] ,
dωdΩ ⊥ 3πγ −∞ 2
Cosmic ray interactions with matter and radiation 37
∞
√
2αω 2 rL
2
  Z
dW sin ζ 3αγ sin ζ
(D.9) ' dθ θγ2 K1/3
2
(η) = [F (x) − G(x)] ,
dωdΩ k 3πγ 2 −∞ 2

where x ≡ ω/ωc , we have extended the integrals to infinity for analytical ease, since
anyway the integrands are non-vanishing only in a small angular interval, and defined
Z ∞
(D.10) F (x) ≡ x K5/3 (ξ)dξ , G(x) ≡ xK2/3 (x) .
x

The total emitted energy per unit frequency is the sum of the two contributions above,
hence given by
  √
dW 3αγ sin ζ
(D.11) = F (x) .
dωdΩ 2

The function F (x) scales as x1/3 at x  1, and as x1/2 e−x at x  1. The total power
per unit frequency can be obtained by dividing the energy by the giration period of the
particle, 2πγ/ωg = γ/νg , so that:
√
(D.12) P (ω) = 3ανg sin ζF (x) .

Note that this spectrum peaks around ωc ∼ γ 2 ωg , as argued with an heuristic argument
in the main text.

REFERENCES

[1] B. Rossi, “High Energy Particles”, Prentice-Hall, Inc., Englewood Cliffs, NJ, (1952).
[2] J. D. Jackson, “Classical Electrodynamics”, 3rd edition, John Wiley & Sons (1999).
[3] K. Akiba et al. [LHC Forward Physics Working Group], “LHC Forward Physics,” J. Phys.
G 43, 110201 (2016) [arXiv:1611.05079 [hep-ph]].
[4] P. B. Abraham, K. A. Brunstein, T. L. Cline, “Production of Low-Energy Cosmic-Ray
Electrons”, Physical Review, 150, 1088 (1966).
[5] C. D. Orth, A. Buffington, “Secondary cosmic-ray e± from 1 to 100 GeV in the upper
atmosphere and interstellar space, and interpretation of a recent e+ flux measurement,”
Astrophys. J. 206, 312 (1976).
[6] G. Molière, “Theorie der Streuung schneller geladener Teilchen I. Einzelstreuung am
abgeschirmten Coulomb-Feld”, Zeitschrift Naturforschung Teil A, 2, 133 (1947)
[7] G. Molière, “Theorie der Streuung schneller geladener Teilchen II. Mehrfach- und
Vielfachstreuung”, Zeitschrift Naturforschung Teil A, 3, 78 (1948).
[8] H. A. Bethe, “Molière’s theory of multiple scattering,” Phys. Rev. 89, 1256-1266 (1953)
[9] K. Mannheim and R. Schlickeiser, “Interactions of Cosmic Ray Nuclei,” Astron. Astrophys.
286, 983-996 (1994) [astro-ph/9402042].
[10] A. W. Strong and I. V. Moskalenko, “Propagation of cosmic-ray nucleons in the galaxy,”
Astrophys. J. 509 (1998), 212-228.
38 P. D. Serpico

[11] C. Evoli et al., “Cosmic-ray propagation with DRAGON 2: I. numerical solver and
astrophysical ingredients,” JCAP 02 (2017), 015 [arXiv:1607.07886].
[12] E. A. Uehling, “Penetration of heavy charged particles in matter,” Ann. Rev. Nucl. Part.
Sci. 4 (1954), 315-350.
[13] S. P. Ahlen, “Theoretical and experimental aspects of the energy loss of relativistic heavily
ionizing particles,” Rev. Mod. Phys. 52 (1980), 121-173 [erratum: ibidem, pages 653-653].
[14] P. A. Zyla et al. [Particle Data Group], “Review of Particle Physics,” PTEP 2020, no.8,
083C01 (2020).
[15] C. Møller, “Über die Wechselwirkung von Zwei Elektronen”, Zeitschrift fur Physik, 70, 786
(1931).
[16] H. Bethe, E. Fermi, “Über die Wechselwirkung von Zwei Elektronen”, Zeitschrift fur
Physik, 77, 296 (1932).
[17] E. Fermi, “Quantum Theory of Radiation”, Reviews of Modern Physics, 4, 87 (1932),
[18] W. Heitler, “The quantum theory of radiation,” 1936.
[19] S. Schweber, “Enrico Fermi and Quantum Electrodynamics, 192932” Physics Today 55, 6,
31 (2002).
[20] H. J. Bhabha, “The scattering of positrons by electrons with exchange on Dirac’s theory
of the positron,” Proc. Roy. Soc. Lond. A 154, 195-206 (1936)
[21] M. Padovani, D. Galli and A. E. Glassgold, “Cosmic-ray ionization of molecular clouds,”
Astron. Astrophys. 501, 619 (2009) [arXiv:0904.4149 [astro-ph.SR]].
[22] S. Gabici, “Low energy cosmic rays: Regulators of the dense interstellar medium,”
arXiv:2203.14620.
[23] H. Bethe and W. Heitler, “On the Stopping of Fast Particles and on the Creation of Positive
Electrons,” Proc.Roy. Soc., 146, 83 (1934).
[24] G. B. Rybicki and A. P. Lightman, “Radiative Processes in Astrophysics”, Wiley-VCH
(1991).
[25] G. Ghisellini, “Radiative Processes in High Energy Astrophysics,” Lect. Notes Phys. 873,
1-147 (2013) [arXiv:1202.5949].
[26] G. R. Blumenthal and R. J. Gould, “Bremsstrahlung, synchrotron radiation, and compton
scattering of high-energy electrons traversing dilute gases,” Rev. Mod. Phys. 42, 237-270
(1970).
[27] R. J. Gould, “Electromagnetic Processes (Princeton Series in Astrophysics)”, Princeton
University Press (2020).
[28] T. Akahori et al., “Cosmic Magnetism in Centimeter and Meter Wavelength Radio
Astronomy,” Publ. Astron. Soc. Jap. 70, no. 1, R2 (2018) [arXiv:1709.02072].
[29] S. Abdollahi et al. [Fermi-LAT Collaboration], “A gamma-ray determination of the
Universe’s star formation history,” Science 362, no. 6418, 1031 (2018) [arXiv:1812.01031].
[30] O. Klein, T. Nishina, “Über die Streuung von Strahlung durch freie Elektronen nach
der neuen relativistischen Quantendynamik von Dirac,” Zeitschrift fur Physik 52, 853-
868(1929).
[31] M. S. Zolotorev, K. T. McDonald, “Classical Radiation Processes in the Weizsacker-
Williams Approximation,” arXiv:physics/0003096.
[32] S. M. Ressler et al., “Magnetic-Field Amplification in the Thin X-ray Rims of SN 1006,”
Astrophys. J. 790, no.2, 85 (2014) [arXiv:1406.3630 [astro-ph.HE]].
[33] G. Puppi (editor), “Questioni relative alla rivelazione delle particelle elementari, con
particolare riguardo alla radiazione cosmica”, Proceedings of the International School of
Physics “Enrico Fermi”, Vol. I (1953).
[34] J. W. Cronin, “The 1953 cosmic ray conference at Bagneres de Bigorre: The Birth of sub
atomic physics,” Eur. Phys. J. H 36, 183-201 (2011) [arXiv:1111.5338 [physics.hist-ph]].
Cosmic ray interactions with matter and radiation 39

[35] G. Puppi (editor), “Questioni relative alla rivelazione delle particelle elementari, e alle loro
interazioni con particolare riguardo alle particelle artificialmente prodotte ed accelerate”,
Proceedings of the International School of Physics “Enrico Fermi”, Vol. II (1954).
[36] J. R. Letaw, R. Silberberg, C. H. Tsao, “Proton-nucleus total inelastic cross sections - an
empirical formula for E greater than 10 MeV,” The Astrophys. J. Suppl. Series 51, 271-276,
1983.
[37] M. N. Mazziotta, F. Cerutti, A. Ferrari, D. Gaggero, F. Loparco and P. R. Sala,
“Production of secondary particles and nuclei in cosmic rays collisions with the interstellar
gas using the FLUKA code,” Astropart. Phys. 81, 21-38 (2016) [arXiv:1510.04623 [astro-
ph.HE]].
[38] S. Krakau and R. Schlickeiser, “Pion Production Momentum Loss of Cosmic ray Hadrons,”
Astrophys. J. 802 (2015) no.2, 114.
[39] S. R. Kelner, F. A. Aharonian and V. V. Bugayov, “Energy spectra of gamma-rays,
electrons and neutrinos produced at proton-proton interactions in the very high energy
regime,” Phys. Rev. D 74 (2006), 034018 [erratum: Phys. Rev. D 79 (2009), 039901] [astro-
ph/0606058].
[40] C. Evoli et al., “Cosmic-ray propagation with DRAGON2: II. Nuclear interactions with
the interstellar gas,” JCAP 07 (2018), 006 [arXiv:1711.09616].
[41] Y. Genolini, D. Maurin, I. V. Moskalenko and M. Unger, “Current status and desired
precision of the isotopic production cross sections relevant to astrophysics of cosmic rays:
Li, Be, B, C, and N,” Phys. Rev. C 98 (2018) no.3, 034611 [arXiv:1803.04686].
[42] M. S. Longair, “High-energy astrophysics”, 3rd Edition, Cambridge University Press (2011).
[43] D. Maurin, “Propagation des rayons cosmiques dans un modèle de diffusion: une nouvelle
estimation des paramètres de diffusion et du flux d’antiprotons secondaires,” PhD thesis
Savoie U (2001).
[44] R. Ramaty, B. Kozlovsky and R. E. Lingenfelter, “Nuclear gamma-rays from energetic
particle interactions.” Astrophys. J. Supp. 40, 487-526 (1979)
[45] A. Giuliani et al. [AGILE], “Neutral pion emission from accelerated protons in the
supernova remnant W44,” Astrophys. J. Lett. 742 (2011), L30 [arXiv:1111.4868].
[46] M. Ackermann et al. [Fermi-LAT], “Detection of the Characteristic Pion-Decay Signature
in Supernova Remnants,” Science 339 (2013), 807 [arXiv:1302.3307].
[47] E. Kafexhiu, F. Aharonian, A. M. Taylor and G. S. Vila, “Parametrization of gamma-
ray production cross-sections for pp interactions in a broad proton energy range from
the kinematic threshold to PeV energies,” Phys. Rev. D 90, no. 12, 123014 (2014)
[arXiv:1406.7369].
[48] C.D. Dermer, “Binary Collision Rates of Relativistic Thermal Plasmas. II. Spectra,”
Astrophysical Journal 307 47 (1986).
[49] F. Halzen, “The observation of high-energy neutrinos from the cosmos: Lessons
learned for multimessenger astronomy,” Int. J. Mod. Phys. D 31 (2022) no.03, 2230003
[arXiv:2110.01687].
[50] M. Boudaud et al., “AMS-02 antiprotons’ consistency with a secondary astrophysical
origin,” Phys. Rev. Res. 2, no.2, 023022 (2020) [arXiv:1906.07119].
[51] M. di Mauro, F. Donato, A. Goudelis and P. D. Serpico, “New evaluation of the antiproton
production cross section for cosmic ray studies,” Phys. Rev. D 90, no.8, 085017 (2014)
[erratum: Phys. Rev. D 98, no.4, 049901 (2018)] [arXiv:1408.0288 [hep-ph]].
[52] S. J. Lin, X. J. Bi, J. Feng, P. F. Yin and Z. H. Yu, “Systematic study on the cosmic ray
antiproton flux,” Phys. Rev. D 96, no.12, 123010 (2017) [arXiv:1612.04001 [astro-ph.HE]].
[53] M. W. Winkler, “Cosmic Ray Antiprotons at High Energies,” JCAP 02, 048 (2017)
[arXiv:1701.04866 [hep-ph]].
40 P. D. Serpico

[54] F. Donato, M. Korsmeier and M. Di Mauro, “Prescriptions on antiproton cross section data
for precise theoretical antiproton flux predictions,” Phys. Rev. D 96, no.4, 043007 (2017)
[arXiv:1704.03663 [astro-ph.HE]].
[55] M. Korsmeier, F. Donato and M. Di Mauro, “Production cross sections of cosmic
antiprotons in the light of new data from the NA61 and LHCb experiments,” Phys. Rev. D
97, no.10, 103019 (2018) [arXiv:1802.03030 [astro-ph.HE]].
[56] F. Calore et al., “AMS-02 antiprotons and dark matter: Trimmed hints and robust bounds,”
SciPost Phys. 12, no.5, 163 (2022) [arXiv:2202.03076 [hep-ph]].
[57] F. Donato, N. Fornengo and P. Salati, “Anti-deuterons as a signature of supersymmetric
dark matter,” Phys. Rev. D 62, 043003 (2000) [hep-ph/9904481]
[58] M. Cirelli, N. Fornengo, M. Taoso and A. Vittino, “Anti-helium from Dark Matter
annihilations,” JHEP 08, 009 (2014) [arXiv:1401.4017].
[59] E. Carlson et al., “Antihelium from Dark Matter,” Phys. Rev. D 89, no.7, 076005 (2014)
[arXiv:1401.2461].
[60] K. Mori, C. J. Hailey, E. A. Baltz, W. W. Craig, M. Kamionkowski, W. T. Serber
and P. Ullio, “A Novel antimatter detector based on x-ray deexcitation of exotic atoms,”
Astrophys. J. 566, 604-616 (2002) [arXiv:astro-ph/0109463 [astro-ph]].
[61] L. Šerkšnytė et al,, “Reevaluation of the cosmic antideuteron flux from cosmic-ray
interactions and from exotic sources,” arXiv:2201.00925.
[62] A. Shukla, A. Datta, P. von Doetinchem, D. M. Gomez-Coral and C. Kanitz, “Large-scale
Simulations of Antihelium Production in Cosmic-ray Interactions,” Phys. Rev. D 102, no.6,
063004 (2020) [arXiv:2006.12707].
[63] V. S Berezinskii et al. “Astrophysics of cosmic rays” (edited by V.L Ginzburg) Amsterdam:
North-Holland, 1990.
[64] V. Berezinsky and O. Kalashev, “High energy electromagnetic cascades in extragalactic
space: physics and features,” Phys. Rev. D 94, no.2, 023007 (2016) [arXiv:1603.03989].
[65] A. Capanema, A. Esmaili and P. D. Serpico, “Where do IceCube neutrinos come from?
Hints from the diffuse gamma-ray flux,” JCAP 02, 037 (2021) [arXiv:2007.07911].
[66] F. Halzen, A. D. Martin, “Quarks and Leptons: An Introductory Course in Modern Particle
Physics”, Wiley (1984).
[67] M. E. Peskin, D. V. Schroeder, “An Introduction To Quantum Field Theory”, CRC Press
(1995).
[68] M. D. Schwartz, “Quantum Field Theory and the Standard Model”, Cambridge Unv. Press
(2013).