Astroparticle Physics 34 (2011) 447–456

Contents lists available at ScienceDirect

Astroparticle Physics
journal homepage: www.elsevier.com/locate/astropart

Non-linear diffusive acceleration of heavy nuclei in supernova remnant shocks

D. Caprioli ⇑, P. Blasi, E. Amato
INAF/Osservatorio Astrofisico di Arcetri, Largo E. Fermi, 5 – 50125 Firenze, Italy

a r t i c l e i n f o a b s t r a c t

Article history: We describe a semi-analytical approach to non-linear diffusive shock acceleration in the case in which
Received 12 July 2010 nuclei other than protons are also accelerated. The structure of the shock is determined by the complex
Received in revised form 1 October 2010 interplay of all nuclei, and in turn this shock structure determines the spectra of all components. The
Accepted 18 October 2010
magnetic field amplification upstream is described as due to streaming instability of all nuclear species.
Available online 26 October 2010
The amplified magnetic field is then taken into account for its dynamical feedback on the shock structure
as well as in terms of the induced modification of the velocity of the scattering centers that enters the
Keywords:
particle transport equation. The spectra of accelerated particles are steep enough to be compared with
Supernova remnants
Shock waves
observed cosmic ray spectra only if the magnetic field is sufficiently amplified and the scattering centers
Galactic cosmic rays have high speed in the frame of the background plasma. We discuss the implications of this generalized
Nuclei approach on the structure of the knee in the all-particle cosmic ray spectrum, which we interpret as due
Knee to an increasingly heavier chemical composition above 1015 eV. The effects of a non trivial chemical com-
position at the sources on the gamma ray emission from a supernova remnant when gamma rays are of
hadronic origin are also discussed.
Ó 2010 Elsevier B.V. All rights reserved.

1. Introduction particles. Two noticeable exceptions are represented by the work

of [10] and that of [38]. In both papers the calculations consist of
A satisfactory understanding of the origin of cosmic rays (CRs) a numerical solution of the coupled equations of CR transport
must deal with the issue of chemical composition. Probably the and conservation of mass, momentum and energy flux of the over-
most striking instance of the role played by the different chemicals all CR plus background plasma. In the paper by Berezhko and Völk
is the origin of the knee in the all-particle CR spectrum. In a sce- [10] the calculations were not illustrated in detail, and it is difficult
nario in which the maximum energy of accelerated particles scales for us to appreciate the assumptions that were adopted there. The
with the charge of the particles involved, a knee arises naturally as spectra from individual sources was found to be very flat (flatter
a superposition of spectra of chemicals with different charges Ze. than E2 at high energies), so that the observed CR spectrum could
Even more important, the change of spectral slope on the two sides be recovered only by assuming a Galactic diffusion coefficient as
of the knee is determined by the relative abundance of different steep as D(E) / E0.75, which however is not consistent with mea-
chemicals as a function of Z, convolved with the effects of rigidity surements of the CR anisotropy at the Earth. In the paper by
dependent propagation in the Galaxy. Ptuskin et al. [38], some more details were provided, and the
While there has been much work on the propagation of nuclei authors discussed the important role of the velocity of the scatter-
in the Galaxy (mainly because of the importance it has for the pre- ing centers on the shape of the spectrum of accelerated particles,
diction of travel time of CRs in the Galaxy and ratios of secondary which, as a consequence, is here much steeper than that found
to primary fluxes), not much attention has been devoted to the by Berezhko and Völk [10].
acceleration of nuclei in the sources. CR acceleration in SNRs is be- A fully non-linear theory which includes nuclei is made rather
lieved to take place through the mechanism of diffusive shock complex by at least two issues: (1) nuclei change the structure of
acceleration, in its non-linear version that allows one to take into the shock, making the problem harder to tackle and (2) the injec-
account the reaction of accelerated particles on the plasma and tion of nuclei in the accelerator is more challenging to be modelled
on local magnetic fields. Several versions of the non-linear theory than it is for protons, especially because nuclei can be produced as
of diffusive acceleration at shocks have been developed (see [34] a result of dust sputtering. In a non-linear theory, the second point
for a review), but most of them include only protons as accelerated clearly feeds back onto the first one.
It is important to recall that the injection and the non-linear
⇑ Corresponding author. acceleration of nuclei in the Earth’s Bow Shock (EBS) has been suc-
E-mail addresses: caprioli@arcetri.astro.it (D. Caprioli), blasi@arcetri.astro.it
cessfully described using a Monte Carlo simulation in the pioneer-
(P. Blasi), amato@arcetri.astro.it (E. Amato). ing paper in Ref. [25]. These calculations showed how the case of

0927-6505/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.astropartphys.2010.10.011
448 D. Caprioli et al. / Astroparticle Physics 34 (2011) 447–456

the EBS is and remains, 20 years later, the most clear cut instance Section 4 we illustrate our results for the spectra of accelerated
of occurrence of Non-Linear Diffusive Shock Acceleration (NLDSA) particles and the structure of the shock. In Section 5 we discuss
in collisionless shocks. However some caution should be adopted the implications of the presence of accelerated nuclei at the shock
in extending these results to the case of SNRs, in that the physical for the prediction of a gamma ray flux, as due to production and
conditions at the EBS might be somewhat different: the lack of decay of neutral pions. In Section 6 we compare our findings with
electron injection, the fact that magnetic field amplification in the CR flux detected at Earth, discussing the spectra of individual
the EBS appears to remain in the quasi-linear regime and the cru- chemicals and the appearance of the knee at E  106 GeV. We dis-
cial role played by dust sputtering for ion injection in SNRs might cuss our general results and we compare them with previous find-
be considered as possible evidences of such differences. ings in Section 7.
The last point is particularly relevant to our problem in that, in
order to account for the observed discrepancy between the chem-
2. Equations and solution techniques
ical composition of typical interstellar medium and CRs in our
Galaxy, refractory elements (such as Mg, Al, Si and Fe) have to be
In this section we generalize the semi-analytical formalism
injected in a preferential way in the acceleration process with re-
developed in [7,8,21] to the case in which nuclei heavier than
spect to volatile elements [35,36]. Refractory elements are usually
Hydrogen (hereafter simply Heavy Nuclei, HN) are also injected
trapped in dust grains, and their preferential injection has been
and accelerated at a stationary, plane, parallel (background mag-
interpreted as a consequence of the sputtering of the grains when
netic field parallel to the shock normal), newtonian shock wave.
they are swept up by the SNR shock. In particular, the processes
We label with a subscript i quantities referring to different chem-
which lead to the injection of suprathermal iron nuclei as a conse-
ical elements, so that the convection–diffusion equation for the
quence of the sputtering of accelerated dust grains have been put
isotropic part of the distribution function, fi(x, p), reads, for each
forward quantitatively for the first time in [23].
species [see e.g. 39]:
Nevertheless, a self-consistent description of this process would
 
require both a detailed knowledge of the dust chemistry close to a @fi ðx; pÞ @ @fi ðx; pÞ ~ ðxÞ @fi ðx; pÞ
p du
~ ðxÞ
u ¼ Di ðx; pÞ þ þ Q i ðx; pÞ:
SNR shock and an accurate physical description of the grain sput- @x @x @x 3 dx @p
tering, along with a time-dependent treatment of the ionization ð1Þ
of dust and atoms during their acceleration. The intrinsic complex-
ity of such a phenomenon led us (as well as the authors of the pre- Here Di(x, p) is the parallel diffusion coefficient, which may depend
vious work mentioned above [10,38]) to a simplified treatment of both on space and momentum, u ~ ðxÞ ¼ uðxÞ þ v W is the total velocity
the nuclei injection: the measured spectra of relativistic particles of the scattering centers in the shock frame, given by the sum of the
at Earth are reproduced without explicitly taking into account fluid velocity u(x) and wave velocity vW, and Qi(x, p) is the injection
the microphysics either of the nuclei ionization or of the dust sput- term. The shock is at x = 0 and subscripts 0, 1 and 2 label quantities
tering (see Section 3). taken, respectively, at the upstream free-escape boundary x = x0,
One might argue that the relatively low abundances of nuclei in immediately upstream and downstream of the subshock.
the cosmic radiation observed at Earth might make their influence An especially important issue, when taking into account nuclei,
on the shock structure negligible, so that in describing the acceler- is that of particle injection, as stated in the introduction and as
ation process, the shock structure could be treated as determined should become clear below. Unfortunately, the microphysics of
by protons alone, with nuclei behaving as test particles. However, this process is not yet fully understood even for the case of protons
after correcting for propagation effects, it is easy to show that the alone, and much more so for nuclei. In the following we simply as-
nuclear contribution to the total pressure and magnetic field sume that protons are injected from downstream via thermal leak-
amplification in the vicinity of a typical supernova remnant shock age as described in [14], while the injection of HN is tuned in such
may be as important as that of protons. a way as to reproduce the relative abundances observed in the CRs.
In this paper we describe the generalization of the non-linear We do not account for the details of the HN injection, likely related
theory of DSA developed by Amato and Blasi [7,8] and Caprioli to the complex physics of the dust sputtering process. Indeed, the
et al. [21] to include nuclei of different charges. We calculate the required preferential injection of HN is likely related to the fact
spectrum of all species as accelerated at the shock and the struc- that partially ionized heavy particles (i.e., thermalized particles
ture of the shock (including magnetic effects) induced by them. with large mass/charge ratios) have large Larmor radii and are
We also comment on the implications of acceleration of nuclei hence preferentially injected in a thermal leakage scenario (see
on the spectra of secondary products of particle interactions, espe- [24]), and/or in the fact that refractory nuclei can be efficiently in-
cially gamma rays. Finally we show the all-particle CR spectrum at jected via dust grain sputtering [23] (we will discuss these points
Earth resulting from this calculation. in Section 3).
The calculations discussed here are semi-analytical, and from More precisely, we assume that all protons with momentum
the computational point of view very inexpensive. This allows us p > pinj,H have a large enough Larmor radius to cross the shock (sub-
to explore a wide region in parameter space, which is particularly shock) from downstream and start being accelerated. Since the
important when dealing with the goal of explaining the CR spectra shock thickness is expected to be of the order of the Larmor
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi radius
and chemical abundances observed at Earth. For simplicity we con- of particles with thermal momentum pth ¼ 2mH kB T H;2 (where TH,2
sider here only supernovae exploding in a homogeneous interstel- is the downstream proton temperature and kB is the Boltzmann
lar medium (ISM). While more complex situations can be treated in constant), we take pinj,H = nH pth,H, with n  3–4. Furthermore, at a
the context of our formalism, they introduce a wide range of new given momentum p the Larmor radius of a HN with charge Zie is a
and hardly accessible parameters, which overshadow the main factor Zi smaller than that of a proton. Hence, in this scheme, it is
physical results. We will comment further on this point whenever very natural to assume that pinj,i = Zi pinj,H.
we deem it necessary. Clearly, if nuclei are not completely ionized or if they are in-
The paper is organized as follows: in Section 2 we illustrate the jected via dust sputtering this simple recipe may fail to describe
generalization of the equations and solution techniques following the low-energy tail of the CR distribution. Nevertheless, since the
the non-linear theory of [7,8,21]. In Section 3 we comment on spectra of the relativistic particles observed at Earth are consistent
the abundance of nuclei in CRs and at the sources, and how it with power-laws extending from a few GeV/nucleon up to the
relates to the injection of the nuclear component at the shock. In knee, it seems reasonable to assume that injection always occurs
 D. Caprioli et al. / Astroparticle Physics 34 (2011) 447–456 449

" #
at energies between thermal and mildly relativistic. Therefore, the UðxÞc M20 1  U cþ1=2 ðxÞ
spectra we work out are expected to be accurate throughout the Pg ðxÞ ’ ½1 þ fHðxÞ; HðxÞ ¼ cðc  1Þ :
cM20 M A;0 c þ 1=2
entire energy region relevant for the SNR emission, and for the
measured Galactic CRs as well. ð11Þ
The injection term in Eq. (1) can be written as The pressure in cosmic rays is, as usual,
g n0 u0 Z
Q i ðx; pÞ ¼ Q i;1 ðpÞdðxÞ ¼ i 2 dðp  pinj;i ÞdðxÞ; ð2Þ X 4p X þ1
4ppinj;i Pc ðxÞ ¼ P i ðxÞ ¼ dpp3 v ðpÞfi ðx; pÞ: ð12Þ
i
3 i pinj;i
where gi has the usual meaning, i.e. the fraction of particles of each
species crossing the shock that is injected in the acceleration pro- Finally the pressure in the form of magnetic turbulence, which has
cess. We will discuss in Section 3 how to determine reasonable val- been shown to play a key role in the shock dynamics [16], can be
ues of gi as inferred from what is measured at the Earth. approximated, assuming that only standard Alfvén waves are
We solve Eq. (1) along with the spatial boundary condition generated via resonant streaming instability and for M0, MA  1,
fi(x0, p) = 0, which mimics the presence of a free escape boundary as [17]
upstream placed at x = x0. For each chemical element, the distribu- " #
tion function fi(x, p) and the escape flux /esc,i(p) can be written as 1  UðxÞ2
Pw ðxÞ ¼ ð1  fÞUðxÞ3=2 : ð13Þ
[see 21]: 4M A;0
 Z 0  
~ ðx0 Þ
u W i ðx; pÞ A solution of the problem of NLDSA can be obtained through the
fi ðx; pÞ ¼ fsh;i ðpÞ exp  dx0 1 ; ð3Þ
x Di ðx ; pÞ
0 W i;0 ðpÞ same recursive method described in [21]. We summarize it here.
 
@fi ~ 0 fsh;i ðpÞ
u
/esc;i ðpÞ ¼  Di ðx; pÞ ¼ ; ð4Þ  Start from a guess for U(x) and for the distribution function fi(x)
@x x0 W i;0 ðpÞ
(e.g. the test-particle solution) and fix a value of U 1 ¼ Rsub =Rtot .
where we have introduced the functions fsh,i(p) = fi(0, p), and This corresponds to a value of P c;1 , through Eq. (9), since Pg
Z 0 Z 0  and Pw depend only on U.
dx0 ~ ðx00 Þ
u
~0
W i ðx; pÞ ¼ u exp dx00 : ð5Þ  Compute the new distribution functions fi(x, p) by using U(x)
x Di ðx0 ; pÞ x0 Di ðx00 ; pÞ
and fi(x, p) from the previous step in Eqs. (5)–(7).
The distribution function at the shock reads:  Compute the CR pressure through Eq. (12). In general, the value
of Pc,1 that is obtained will be different from Pc;1 . Then Pc(x) and
gi n0 3Rtot fi(x, p) are renormalized by multiplying them by a factor
fsh;i ðpÞ ¼
4pp3inj;i R
e tot U p;i ðpÞ  1 k ¼ Pc;1 =P c;1 .
( Z )  Compute and update the fluid velocity profile U(x) using the
p e tot W 0;i ðp0 ÞU p;i ðp0 Þ  1
dp0 3 R
 exp  ; ð6Þ new Pc(x) in Eq. (9).
pinj;i p W 0;i ðp Þ
0 0 e tot U p;i ðp0 Þ  1
R  Iterate the steps above until k does not change between a step
and the following.
where we used
Z e
e1 
0
d UðxÞ fi ðx; pÞ Starting from an arbitrary U 1 , in general one has k – 1 after conver-
U p;i ðpÞ ¼ U dx ð7Þ gence is reached. The process must then be restarted with a new
x0 dx fsh;i ðpÞ
choice of U 1 until convergence to k = 1 is achieved (within a pre-
and introduced the subshock Rsub = u1/u2 and the total Rtot = u0/u2 scribed level of accuracy). The fluid profile and the distribution
compression ratios for the fluid. In this approach in which the functions so obtained are, at the same time, solutions of both the
waves move with velocity vA with respect to the fluid, the cosmic convection–diffusion and the conservation equations, i.e. solutions
rays which scatter against them feel an effective velocity given by of the full problem.
u + vA and hence a subshock and a total compression ratios given by It is worth recalling that from the computational point of view
e sub ¼ u1 þ v A;1 ;
R e tot ¼ u0 þ v A;0 :
R ð8Þ
this approach to non-linear shock acceleration is the most efficient
u2 þ v A;2 u2 þ v A;2 presented so far in the literature: for a given set of parameters it
takes less than one minute (on a laptop) to calculate the distribu-
As we discuss below, the particles may feel compression ratios lar- tion function fi(x, p) of all accelerated particles and the space
ger or smaller than the fluid, depending on the relative sign of u and dependent structure of all thermodynamical quantities in the
vA . shock region.
The convection–diffusion equation is coupled with the standard
conservation equations for mass, momentum and energy flux,
3. Abundances of heavy elements
where the cosmic ray terms are intended to be summed over all
the elements. In particular, the conservation of momentum, after
The most challenging aspect of a non-linear calculation of diffu-
dividing by the bulk pressure q0 u20 , reads:
sive shock acceleration in the presence of accelerated nuclei is the
1 tuning of the relative abundances at the source in a way that may
UðxÞ þ Pg ðxÞ þ Pc ðxÞ þ Pw ðxÞ ¼ 1 þ ; ð9Þ
cM20 fit the abundances observed at the Earth after propagation. The dif-
ficulty is manyfold: first, the non-linearity of the problem makes it
where the gas pressure, if the heating in the precursor is purely adi- very difficult to establish once and for all what the required rela-
abatic, can be written as tive abundances are; second, the spectrum observed at Earth from
UðxÞc an individual SNR is the superposition of the instantaneous spectra
Pg ðxÞ ¼ : ð10Þ of particles escaping the remnant at different times, and the rela-
cM20
tive abundances are, most likely, time dependent. One final com-
If turbulent heating is also at work converting a fraction f of the plication to keep in mind in establishing a connection between
magnetic pressure into thermal energy, the pressure of the gas source and local abundances of ions comes from the spallation pro-
would instead read [see Section 4.2 of 17]: cesses suffered by nuclei during propagation in the Galaxy. This
450 D. Caprioli et al. / Astroparticle Physics 34 (2011) 447–456

simply means that the relative abundances of chemicals in a given This reflects the rigidity dependent nature of particle propagation in
remnant at a given time cannot be univocally inferred from the CR the Galaxy, namely the fact that the diffusion coefficient is
abundances that we measure at Earth. D(p) / Rd = (p/Zi)d (R is the particle rigidity). The simple scaling in
It is worth recalling that the situation is rather different for the Eq. (15) holds as long as spallation in the ISM can be neglected,
EBS, where real-time measurements of the spectra of different ele- therefore it should only be used at sufficiently high energies (we
ments are available, thus allowing a more detailed interpretation use p* = 105 GeV/c as normalization point) so that the escape time
of the ongoing physical processes. In this case, in fact, the injection from the Galaxy is much shorter than the spallation time scale for
of both protons and nuclei via thermal leakage provided a success- all the relevant species.
ful explanation of the EBS properties as due to efficient NLDSA [25]. The ratio of abundances between ions and protons at the same
On the other hand, the same thermal leakage scheme seems momentum p* as measured at the Earth is therefore:
unsatisfactory when applied to the case of SNRs as the sources of
!3b
Galactic CRs. In fact, a thorough study of the elemental and isotopic ni g pinj;H gi dþb3
ratios in CRs compared with solar ones shows that refractory ele- K iH ¼ ¼ i Zd ¼ Z : ð16Þ
nH gH i pinj;i gH i
ments are preferentially accelerated with respect to volatile ones.
We refer to the original works [35,36] (and to a recent review
[41]) for a comprehensive discussion on these topics. In this expression, the ratio KiH comes from the measured spectra
Here we only want to stress that this fact has been explained by and, at least in the context of this simple (leaky-box-like) approach,
invoking the fundamental role of the interstellar dust in the injection b + d is fixed by observations to be 4.7. The ratio of the g’s to be
of nuclei, as put forward in [23]. In few words, when a grain of refrac- used in the calculations can therefore be easily estimated from
tory material is caught by a SNR shock, it is efficiently accelerated by measurements. The values resulting from this procedure are illus-
crossing the shock many times thanks to its large Larmor radius. trated in Table 1. In the same table we also show the flux of differ-
When its velocity reaches about 0.01c it begins to sputter and to re- ent chemicals at 1 TeV and the spectral slope at Earth as inferred
lease atoms. These have the same velocity of the parent grain: they using the so-called Poly-Gonato parametrization by Hörandel [29].
are hence well suprathermal and therefore can be easily injected These slopes reflect the presence of spallation and most likely the
and take part in the acceleration process. Such an effect can also ac- superposition of different types of sources (see discussion in
count for the fact that among refractory nuclei there is no preferen- Section 7).
tial acceleration of heavy elements with respect to lighter ones, As stressed above, the procedure just outlined is based upon a
which is instead the case for volatile elements. This latter evidence test-particle picture and should only be used as a qualitative esti-
is usually explained as a consequence of the fact that heavy elements mate of the efficiency of acceleration of the different chemicals.
may be ionized only partially and hence preferentially injected be- Many factors can affect the result of this simple estimate: (1) the
cause of their relatively large rigidity (see [22,24]). non-linear diffusive acceleration at shocks does not lead to power
In this section we illustrate for simplicity how to estimate the law spectra; (2) the values of gi can be (and in general are expected
relative normalization between source spectra of different ions in to be) time dependent, which is especially relevant since the spec-
a test-particle approach. Such an estimate is expected to hold for trum of CRs observed at Earth results from the superposition of
energies above a few GeV/nucleon, where the spectra detected at instantaneous spectra of particles escaping the SNR at different
Earth are very close to power-laws. The situation may be more times; (3) there are different types of SNRs exploding in different
complicated in the region between slightly suprathermal and environments, which leads to different ratios at the sources; (4)
weakly relativistic particles because of the effects of partial ioniza- spallation changes the spectra in a complex way, and the difference
tion and dust sputtering discussed above. in the spectral slopes of different chemicals as reported in the table
The resulting abundances, not too dissimilar from the exact above can partially be due to these reactions, although it is likely
ones derived in Section 6, are used to make several important that the differences also reflect the contributions from SNRs in dif-
points, mainly concerning the instantaneous spectra of relativistic ferent ambient media; (5) the diffusion coefficient is not strongly
particles (for instance to address the dynamical role of nuclei at the constrained by the secondary to primary ratios and the slope d
shock, and their effect on the generation of secondary radiation). could well be in the range 0.3–0.7, though the upper end of this
This discussion can be carried out regardless of a detailed knowl- range leads to problems with anisotropy (e.g. [28]); (6) although
edge of the non-relativistic region of the spectra. the spectrum observed at Earth is very close to a power law, differ-
The starting point of our estimate is represented by the ob- ent SNRs or the same remnant at different stages of its evolution
served relative abundances in CRs at Earth [see e.g. 30 for a recent may have both different spectra and different values of the relative
review of the observational results]. In a test-particle approach the abundances.
spectrum of particles accelerated at the shock can be written as a In Sections 4 and 5 we adopt the simple procedure outlined above
power law with slope b: to estimate the abundances of ions in a source at a given time, while
in Section 6, where we calculate the spectrum of the different
!b
C gi p chemicals at the Earth and the all-particle spectrum, we follow a
fi ðpÞ ¼ ; ð14Þ
4pp3inj;i pinj;i
Table 1
with C the same constant for all i. Let us also assume pinj,i = ai pinj,H. Flux at 1 TeV (in units of [m2 sr s TeV]1), spectral slope (ci) and ratio relative to H at
In the following we shall assume that the injection is charge depen- 105 GeV (KiH) measured at Earth for the most abundant elements in galactic CRs. CR
dent, with ai = Zi, being Zie the electric charge of ions of type i. The data are from [29], Table 7, except the ones denoted with (*) which are from [33]. C, N
number of particles with momentum > p can be estimated as and O are taken as one effective element with ZCNO = 7 and ACNO = 14. The same is
done for Mg, Al and Si, with ZMgAlSi = 13 and AMgAlSi = 27.
Fi(>p) 4pp3fi(p) = Cgi (p/pinj,i)3b. The spectrum at Earth is affected
gi
by propagation as Element Flux (1 TeV) ci KiH gH
!3b
p H (8.73 ± 0.07)  102 2.71 ± 0.02 1 1
ni ðpÞ F i ðpÞsi ðpÞ ¼ C gi Z di sH ðpÞ; ð15Þ He (5.71 ± 0.09)  102 2.64 ± 0.02 0.90 0.28
pinj;i
C–N–O(*) 5.4  102(*) 2.64(*) 0.85 0.03
Mg–Al–Si 1.7  102 2.66 0.25 3.2  103
where we used the fact that the confinement time si of nuclei of
Fe (2.04 ± 0.26)  102 2.59 ± 0.06 0.41 1.6  103
charge Zi is Z di times larger than that of protons at given momentum.
 D. Caprioli et al. / Astroparticle Physics 34 (2011) 447–456 451

somewhat more refined procedure. It is however wise to keep in For comparison, in the right panel of Fig. 1 we also show the
mind that, for all the reasons mentioned above, it is extremely diffi- pressure associated to the thermal plasma and to the magnetic tur-
cult, if at all possible, to infer in a realistic way the abundances at the bulence (taken as homogeneous upstream) and the normalized
sources that may fit the whole set of data available at Earth. fluid velocity as well. The latter shows the spatial profile typical
of cosmic ray modified shocks, with a precursor, a weak subshock
4. Particle spectra and shock hydrodynamics with compression ratio Rsub ’ 3.9 and a total compression ratio lar-
ger than 4, namely Rtot ’ 4.5.
We consider a shock moving with velocity u0 = 4000 km/s in a The resulting spectra of the accelerated particles reflect the
homogeneous medium with particle density 0.01 cm3, tempera- modified dynamics, being steeper than the test-particle prediction,
ture T0 = 106 K (sonic Mach number  34) and magnetic field p4, at low energies, and flatter than p4 at the highest energies.
B0 = 5 lG aligned with the shock normal. This choice of parameters The finite shock size (spatial boundary), which allows for particle
corresponds to a 2000 year old SNR with radius Rsh  14.4 pc, i.e. a escape from upstream, induces a rigidity-dependent cut-off in
SNR at the beginning of its Sedov–Taylor stage for a SN explosion of the spectra of the various species, leading to maximum momenta
1051 erg and an ejecta mass of 1.4 solar masses. It is worth recalling scaling with atomic charge (left panel of Fig. 1). More precisely,
that the highest cosmic ray energy is thought to be achieved at this since the diffusion is rigidity-dependent, and since in diffusive
evolutionary stage [13]. shock acceleration the spectral slope at a given momentum,
The free-escape boundary is placed at x0 = 0.2Rsh upstream of qi ðpÞ ¼  d dloglogfi ðpÞ
p
, depends only on the compression ratio actually
the shock and the diffusion coefficient is taken as Bohm-like, experienced by the particles with that momentum, we have the
simple scaling: qi(p/Zi) = qH(p). In principle, when the relative nor-
1 pc malization and the ionization state are fixed, and when the CR
Di ðx; pÞ ¼ v ðpÞ ; ð17Þ
3 Z i BðxÞ pressure is dominated by relativistic particles, one could use this
simple scaling to calculate the distribution functions of all the ele-
in the amplified magnetic field at the shock position, namely
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ments at the shock, their spatial dependence via Eq. (3) and finally
BðxÞ ¼ B1 ¼ 8pq0 u20 P w;1 upstream and B(x) = B2 = RsubB1 the total pressure in cosmic rays Pc(x) and the whole shock struc-
downstream. ture. However, within the fast semi-analytic formalism outlined
In Fig. 1 we show the spectra of accelerated particles and the in Section 2, the computational effort is far from being a severe is-
quantities related to shock hydrodynamics, obtained through the sue demanding numerical optimization.
iterative method described in Section 2, in a case of efficient parti- For comparison, we solved the set of equations describing a
cle acceleration (we used nH = 3.8, corresponding to gH = 5.7  105 shock with exactly the same characteristics, except for the fact that
in Eq. (2)). Notice that the gas pressure upstream is very low and no HN were considered. We found PH ’ 0.07 at the shock. Compar-
lies outside the plot boundaries. ing this result with those in the right panel of Fig. 1, the conclusion
The most noticeable feature is the fact that, for the standard is that when acceleration of HN is taken into account, the pressure
abundances deduced in Section 3, the dynamical role of nuclei over- in accelerated protons is decreased, but the total pressure in accel-
all is twice as important as that of protons: at the shock position the erated particles is larger, and the shock correspondingly more
pressure of accelerated protons is PH ’ 0.05, in units of the ram modified. Indeed, in the absence of HN, the compression ratios turn
pressure far upstream, while the pressure in the form of relativistic out to be Rsub ’ 3.95 and Rtot ’ 4.2. At the same time, also the
HN is about 0.1 (right panel of Fig. 1). The latter is mainly associated magnetic field amplification via resonant streaming instability of
to He nuclei (PHe ’ 0.045) but also CNO and Fe nuclei carry non-neg- the accelerated particles gives different results when the accelera-
ligible fractions of the momentum in the form of accelerated parti- tion of HN is included. We find that the proton-only case returns a
cles (PCNO ’ 0.017 and PFe ’ 0.003). The relative importance of HN downstream magnetic field of 33 lG, while with HN B2 ’ 47 lG.
with respect to protons should not be affected by the exact shape This result is easily understood given the dependence of the
of the spectra between pinj,i/Zi and a few GeV in rigidity (which is re- amplified magnetic field on the pressure of accelerated particles.
lated to the injection details discussed above), provided that the All these effects contribute to make the case for the necessity of
spectra are not much steeper than p4 and/or that the maximum taking HN into account properly: the acceleration of HN cannot
momentum is larger than few GeV/c. These peculiar conditions be treated in a test-particle approximation, nor can it be linearly
might only be met rather late in the remnant evolution, and defi- added to a proton only case. In order to correctly describe the
nitely not as early as at the beginning of the Sedov stage, i.e., when shock, all species need to be taken into account in a fully non-linear
the acceleration efficiency is at its maximum. calculation.

Fig. 1. Left panel: all-particle spectrum (thick line) and spectra of individual elements. Right panel: spatial dependence of the hydrodynamical quantities (the gas pressure
upstream is very low and lies outside the plot boundaries).
452 D. Caprioli et al. / Astroparticle Physics 34 (2011) 447–456

A crucial point to stress is that the mild shock modification and of the result on the Alfvénic Mach number with respect to the
the correspondingly steeper spectra found here and illustrated in assumptions of [38], even if in both cases MA is calculated in the
Fig. 1 are the consequence of the assumption of scattering centers amplified magnetic field.
moving with a wave velocity that equals the Alfvén velocity calcu- Moreover, in our model the strength of the amplified magnetic
lated in the amplified magnetic field. As discussed by Caprioli et al. field is not fixed a priori but is rather a non-linear output of the full
[20], without this assumption the spectra produced in SNRs are calculation, depending on the efficiency of the CR acceleration
very flat and can hardly be related to the CR spectrum observed through Eq. (13). In this sense our approach includes a self-regulat-
at Earth, unless, as done by Berezhko and Völk [10], one assumes ing effect which returns almost the same result in terms of particle
a very strong dependence of the diffusion coefficient in the Galaxy spectra and shock modification for a wide range of values of the
on energy, D(E) / E0.75, which however leads to too large anisot- Alfvénic Mach number at upstream infinity, MA,0, and of the parti-
ropy at the knee compared with observations. It is therefore of cle injection efficiency ni.
the highest importance to realize how the main result in this type
of calculations is actually due to one of the aspects that are least 5. c-Rays from accelerated nuclei
known.
The fact that taking into account the velocity of the scattering In the typical environment of a SNR, in addition to playing an
centers could lead to steeper spectra was already recognized in important role on the shock dynamics, accelerated nuclei may also
the early literature on the topic [see e.g. 9], but the issue has be- give a contribution in terms of c-ray emission via p0 production in
come much more discussed in recent times, after discovery of nuclear interactions with thermal protons. In order to calculate
highly amplified magnetic fields in the shock region. There are this emission, we model each accelerated nucleus with atomic
two levels of the problem: (1) we are unable to determine the number Ai and energy E as an ensamble of Ai protons, each with en-
velocity of the waves responsible for the scattering from first prin- ergy E/Ai, and adopt the parametrization of the nuclear proton–
ciples, especially in the case of relevance for us in which magnetic proton cross section worked out by Kamae et al. [32].
field is strongly amplified by CR induced instabilities. If one as- A simple estimate of the HN contribution to pion production
sumes (unrealistically) that even in these extreme circumstances can be obtained as follows. Let us assume that a nucleus with en-
the waves remain Alfvén waves, namely with magnetic field per- ergy Ei produces monochromatic photons with energy Ec such that
pendicular to the (much smaller) background magnetic field, then Ei = vAiEc, with v 10. The photon spectrum produced by nuclei of
the Alfvén speed is well defined, but its value is so small to induce species i would thus be:
negligible effects on the spectrum of accelerated particles. On the
other hand, one might expect that an effective value of the wave c dEi
Ni ðEc Þ / Ai Ni ðEi Þ ¼ vA2i N i ðEi Þ: ð18Þ
speed is close to the Alfvén speed calculated using the strength dEc
of the amplified field as reference value. In this case the Alfvén
For spectra Ni ðEi Þ / Eq up to an energy Emax,i we have:
speed is much larger and its effect on the transport equation is i
 
not negligible. (2) Even if we knew how to calculate the wave Nci Ec
velocity (for instance by assuming that it corresponds to the Alfvén   ¼ K i Ai2q ; ð19Þ
NcH Ec
speed in the amplified field), the helicity of the waves is unknown,
being related to the type of wave, the mechanism that generated which means that, for the standard case q = 2, all the species con-
them and the transport of the waves through the shock (reflection tribute to the c-ray flux proportionally to their abundances at any
and transmission). In other words, it is not easy to calculate the photon energy below the minimum of Emax,i/(Aiv). Since typically
wave velocity with respect to the shock surface. Emax,i / Zi, the photon spectrum of any HN is cut-off at an energy
Since the slope of the CR spectrum depends on the compression which is a factor 2 lower than for protons. It is also interesting to
ratio of the scattering centers, as felt by the particles, it is clear notice that, if the accelerated particle spectra are flatter than E2,
from Eq. (8) that, when vW is not negligible compared with u, the the contribution of HN with respect to protons is boosted by a factor
actual spectrum may be flatter or steeper than the standard predic- A2q
i . This turns out to be as large as a factor 5 in the case of Fe nu-
tion depending upon the relative sign of u and vW, both in the up- clei accelerated with a spectrum E1.5. There may be stages in the
stream and in the downstream regions. If vW is calculated in the evolution of a SNR when the shock is strongly modified and such
amplified magnetic field, it typically turns out to be a not-negligi- hard spectra may appear.
ble fraction of u, and thus the helicity of the waves strongly affects The spectrum of c-rays, calculated using the particle spectra
the shape of the CR spectrum. shown in Fig. 1, is illustrated in Fig. 2. The HN contribution to
While upstream the turbulence is very likely generated by the the c-ray flux turns out to be dominant over that of protons, with
CR gradient, and thus vW is expected to have sign opposite to u, a total predicted flux which is about a factor 2.5 larger than the
downstream the waves may retain the same helicity (as one would standard prediction for the case when accelerated particles are
expect based on the transmission and reflection coefficients at the only protons. Moreover, since the contribution of all species of
subshock discontinuity appropriate for Alfvén waves), or rather HN is truncated a factor 2 below that obtained with protons alone,
may be fully isotropized (vW = 0). Another possibility, recently ex- the shape of the high-energy end of the photon spectrum is some-
plored by Ptuskin et al. [38], is that downstream waves may also be what different when HN are taken into account. Although this may
generated via streaming instability by the advected CRs, whose in principle represent a spectral feature flagging for the presence of
gradient is induced by the postshock evolution and in particular accelerated nuclei, we find it unlikely to be detectable, given the
by adiabatic losses in the expanding shell: this scenario leads these many intrinsic systematics of experimental and theoretical nature
authors to assume that in the downstream vW has the same sign as and given the topology of the emitting regions which in general are
u. This mechanism may be at work in SNRs, but one would expect complex, so that different regions of the remnant (where the high-
that it should act in the downstream plasma on hydrodynamical est energy could be slightly different) contribute to the flux along
spatial scales. It is therefore hard to assess the relevance of this the same line of sight.
process as compared with that of the turbulence produced up- For the modeling of the gamma ray emission from SNRs a cru-
stream and then advected through the subshock in the down- cial parameter is the gas density in the shock region. Moreover,
stream region. In this work we conservatively assume that in the context of non-linear theory, the temperature of the
downstream vW,2 = 0, which leads to a less dramatic dependence downstream plasma is an output of the problem and the thermal
 D. Caprioli et al. / Astroparticle Physics 34 (2011) 447–456 453

our computation technique, but only to our belief that the addi-
tional uncertainties that would follow overcome the benefits of
adding a class of sources to our calculations, thereby making the
results not more, but in fact less trustworthy. Nevertheless, some
of these pieces of information may be available when dealing with
a given SNR: in this respect, our approach could be easily modified
in order to include a more complex evolution of the remnant
accounting also, e.g., for the presence of either a dense Supergiant
wind or the final encounter between the forward shock and the or-
dinary ISM (see for instance [37] for an analytical treatment of the
evolution of different SNR types).
The spectra of the accelerated particles are calculated as de-
scribed in [20], i.e. taking into account the instantaneous escape
flux from the upstream boundary [see 18], the advection in the
downstream region, which leads to adiabatic losses as a conse-
Fig. 2. Spectra of c-rays due to the decay of p0s produced in nuclear interactions quence of the shell expansion, the escape of particles from down-
(thick line). The partial contributions by each element are also shown. stream because of shell breaking and/or inhomogeneities in the
circumstellar medium [the fraction of downstream escaping parti-
cles is taken as 10%: see also Section 3.3 of [20] for a discussion of
emission from downstream (continuum due to bremsstrahlung
this point]. The relative abundances at the sources are iteratively
and lines from non-equilibrium ionization) can be calculated. The
adjusted in order to fit the fluxes measured at Earth.
latter, scaling with the square of the plasma density, is more sen-
The damping of the magnetic field is heuristically taken into
sitive than the gamma ray emissivity to the value of the gas den-
account by assuming that a fraction f of the generated turbulence
sity. In some cases the gas density is inferred from gamma ray
is damped into gas heating (Alfvén heating), with f(t) = 1  exp
observations, with the strong assumption that the gamma ray
[u0(t)/udamp], where the velocity udamp at which the effect be-
emission is due to production and decay of neutral pions and
comes relevant may range between about 200 and 1000 km/s,
checked versus the signal in the form of lines of thermal origin
depending on the details of the turbulence generation and damp-
[see e.g. 26 for the case of RX J1713.7-3946]. The results shown
ing [see [37], for a study of this topic]. We checked a posteriori that
in this section illustrate the fact that the inclusion of ions in the
our findings depend only weakly on the choice of udamp.
calculations of the gamma ray emission from a SNR leads to an
The convolution over time of the spectra of accelerated particles
estimate of the ambient density which may easily be a factor
injected into the Galaxy returns the source spectrum, which has
2–4 smaller than when the gamma ray emission is calculated
been corrected by accounting for the propagation in the Milky
using only protons. The thermal emission decreases correspond-
Way. We adopt a simple leaky box model of the Galaxy, which is
ingly by 4–16. Claims on the detectability or non-detectability
taken as a cylinder with radius and half-height equal to 10 and
of thermal emission should take into account the intrinsic uncer-
3.5 kpc, respectively, where 3 SN explode each century. The escape
tainty associated with this phenomenon.
time is a function of the nucleus rigidity R as inferred by standard
secondary-to-primary measurements and reflects into a grammage
6. The Galactic spectrum of CRs kesc(p) = 7.3(p/Z/10GV)db(p) g cm2, with d = 0.55 and b(p) the
dimensionless speed of a nucleus of momentum p. Particle losses
In this section we calculate the total spectrum of CRs acceler- due to spallation against interstellar nuclei during the propagation
ated by a class of benchmark SNRs during the different stages of is also taken into account and described as in Section 4 of [31].
evolution, as described by the analytical evolutionary scheme of The results for the different chemicals are shown in Fig. 3, as
[40]. Our reference SNR has the same environmental parameters plotted on top of the experimental data collected by Blümer
as in Section 4, and we follow its evolution from 0.1 to about et al. [15]. The agreement is very good, and both the spectra and
22TST, where TST ’ 2000 yr is the beginning of the Sedov–Taylor the absolute and relative normalization are consistent with the
stage. We assume that the ambient ISM is homogeneous, hot and data from direct and indirect experiments able to resolve the
rarefied (T0 = 106 K, n0 = 0.01) in order to mimic the explosion of chemical composition of the CR flux at Earth. Notice that some dis-
a core-collapse SN into the hot bubble excavated by its pre-SN crepancies at low energies are due to the fact that the results of our
wind, since about 85% of Galactic SNe are expected to be of this calculations have not been corrected for solar modulation.
type [type II + type Ib/c, see e.g. 27]. Such a choice should also be It is interesting to notice that the spectra of nuclei at Earth have
representative of the typical hot medium inside superbubbles, roughly the same slope at given rigidity, except for the effect of
where most [about 75%, according to 27] of the core collapse SNe spallation, which produces the low energy turnover. For iron nuclei
are found. We do not account here either for the presence of a the flattening due to spallation is visible up to energies of
dense, cold wind, occasionally produced during the Red Supergiant 105 GeV. This conclusion should however be taken with caution:
stage of the progenitor, or for the finite size of the hot bubble. the recent data from the balloon-borne experiment CREAM-II [33]
These ingredients add to the problem of particle acceleration suggest that spectra of nuclei are somewhat flatter than the proton
numerous difficulties and require the introduction of several al- spectrum. This might be a mild signature of the contribution of dif-
most unconstrained parameters, like for instance the mass-loss ferent classes of sources operating in environments with different
rate and the total mass blown in the wind, the temperature and chemical composition and contributing to the CR spectrum in a dif-
the density profile and, more important, the strength and the ferent way than described here.
topology of the magnetic field. By the same token, here we did In Fig. 4 the all-particle spectrum in the knee region is shown
not include the contribution of type Ia-like SNe, which are ex- (thick line), again compared with data from [15]. For comparison
pected to be less frequent and to explode into the denser and two power-laws are also drawn (thin lines), corresponding to the
colder homogeneous ISM. fiducial slopes of the flux below (E2:7 0 ) and above (E3:1
0 ) the knee.
It is important to stress that our decision to not include these The all-particle spectrum that we calculated is, as expected,
situations in our calculations is not due to technical problems of somewhat steeper than the standard best fit to the data above
454 D. Caprioli et al. / Astroparticle Physics 34 (2011) 447–456

Fig. 3. CR flux measured at Earth for the different chemicals considered in the text [the experimental data are from Fig. 9 of [15] [where the proper references to the single
experiments can be found as well]. The thin solid lines in the H, HE and Fe panels correspond to the Poly-Gonato fits [29, see also 1].

the knee (/E3:1

0 ). For this effect, also found by other authors, sev- explained this region of the spectrum with rare (only 2% of all
eral possible explanations have been proposed: in Section 2 of the Galactic SNe) but very energetic (ESN = 3  1051 erg) type IIb SNe.
review by Hillas [28] the author discussed the need for an ‘‘ex- As an alternative, [30] suggested the importance of nuclei heavier
tended tail’’ in the Galactic flux in order to explain the CR data in than iron, which are accelerated up to energies 100 (ZU = 92) lar-
the region between 107 and 109 GeV (see Fig. 2 of that paper), ger than that of protons.
and suggested that such a high-energy feature might be granted We wish to stress however that the spectrum of CRs in this en-
by a proper account of a peculiar class of Type II SNe. A similar sce- ergy region is affected by the relative fluxes of Galactic and extra-
nario has been put forward by Ptuskin et al. [38], who successfully galactic CRs, and this makes the claim for excesses rather weak.
 D. Caprioli et al. / Astroparticle Physics 34 (2011) 447–456 455

Fig. 4. All-particle CR flux at Earth in the knee region. The thick line represents our model output, while the thin lines guide the eye by showing two power laws, / E2:7
0 and
/ E3:1
0 below and above the knee, respectively [data from 15, Fig. 8]. The partial contributions from H, He, CNO, MgAlSi and Fe are also shown.

The only point that appears rather strong in terms of the view of process is most likely diffusive shock acceleration in its non-linear
the origin of CRs illustrated in the papers listed above, as well as version, which accounts for the dynamical reaction of accelerated
in the present paper, is that the spectrum of Galactic CRs is unlikely particles on the shock and for the crucial phenomenon of self-gen-
to extend to very high energies, thereby making the transition from eration of amplified magnetic fields in the acceleration region. In
Galactic to extragalactic CRs at the ankle rather poorly motivated. this respect we have been successful, in that several versions of
The two physical scenarios which are compatible with a transition the theory exist and the different formulations compare well with
in the energy region around 107109 GeV are the mixed composi- each other [see for instance 19]. Our semi-analytical approach has
tion scenario [1–4] and the dip scenario [11,12,5,6]. The former the advantage, when compared to all others, of simplicity and re-
model describes the extragalactic CR contribution as the superpo- duced computation time, which is the very reason why it is
sition of different nuclei, so that the chemical composition in the increasingly more adopted in complex hydrodynamical codes of
transition region is mixed. The latter model is stunning for its sim- the SNR evolution including the reaction of CRs.
plicity: the extragalactic CR contribution is made of protons only In the present paper we completed the theoretical framework
and the transition from galactic to extragalactic CRs is completed previously put forward by our group with the inclusion of acceler-
at 1018 eV. The transition occurs through a dip, produced by the ated ions, which we find to contribute considerably to the shock
onset of pair production, a feature which is very well defined in modification. We show that the spectra of different chemicals ob-
the CR spectrum. The chemical composition in the transition re- served at the Earth can be reproduced reasonably well and we also
gion changes suddenly [6] from an iron dominated Galactic one calculate the implications of acceleration of nuclei on the gamma-
to a proton dominated extragalactic one. ray emission of an individual SNR. The all-particle spectrum of CRs
In both scenarios the Galactic CR spectrum is cut off well below is also reproduced in a satisfactory way.
the ankle. The lowest energy part of the extragalactic CR spectrum The one presented here is not the first attempt at describing
is however affected, in both scenarios, by unknowns related to the acceleration of nuclear species in SNRs in the context of non-linear
propagation of extragalactic CRs in weak magnetic fields that diffusive shock acceleration: [10] published previous work on the
might be present in the intergalactic medium, therefore the contri- topic using a numerical approach to the acceleration problem.
bution of extragalactic sources to CRs in the energy region around a Unfortunately a detailed comparison of our results with their work
few 107 GeV remains poorly constrained. This uncertainty makes is made difficult by the fact that the procedure they used to include
the need for an additional class of sources in the transition region nuclei and the amplification of magnetic field are not fully illus-
somewhat weak. trated: actually it is not even clear to us whether the nuclei were
introduced in the non-linear chain or rather included as test parti-
7. Discussion and conclusions cles in a shock structure mainly modified by CR protons. The spec-
tra obtained in that paper for the individual nuclear species were
The problem of explaining the origin of CRs as observed at the very flat (and in fact concave) leading the authors to require a dif-
Earth is a complex one: it requires us to find the sources, describe fusion coefficient with a strong energy dependence, D(E) / E0.75.
the acceleration of nuclei (and in fact of electrons as well) in such Such diffusion is however known to result in the breaking of the
sources, propagate all particles through the Galaxy taking into ac- diffusive approximation well below the knee and consequently in
count diffusion and losses, and finally correct for local effects (such excessive CR anisotropy, at odds with observations.
as the solar modulation at low energies). One can easily realize that More details were given in the work recently presented by
this is a hard task, and yet one in which we have been successful in Ptuskin et al. [38]. Their calculation consists again in a finite differ-
many respects, less in others. ences scheme for the solution of the coupled transport and fluid
The most plausible sources of Galactic CRs remain SNRs, equations, as for [10]. The authors discuss the effect of assuming
although a solid proof of the supernova paradigm satisfying all sci- a large velocity of the scattering centers, which leads to steepening
entific standards has not been obtained as yet. The acceleration of the resulting spectra (and to a reduction of the CR acceleration
456 D. Caprioli et al. / Astroparticle Physics 34 (2011) 447–456

efficiency). As discussed in Section 4, this very important issue, at Earth. Taking into account the wide variety of situations that are
that was also discussed by Caprioli et al. [20], is very poorly realized in Nature and the fact that in principle the CR spectra ob-
known: the velocity of the scattering centers can hardly be esti- served at Earth are the result of the convolution of so many differ-
mated reliably, and even more important the helicity of the waves ent environmental situations, even this latter task would be
is unknown. In fact, if the helicity is not chosen properly (but arbi- formidable. At present, we think it is difficult to envision going be-
trarily) the net effect may well be that of flattening the spectra yond the point already reached and still maintain the same stan-
even further. In the paper by Ptuskin et al. [38], as well as in ours, dard of scientific credibility. On the other hand, the modeling of
it is assumed that an estimate of the speed of the waves is repre- individual SNRs which leads to predict their appearance in differ-
sented by the Alfvén velocity evaluated in the amplified magnetic ent frequency ranges may provide several breakthroughs in the
field at the shock. In [38] the authors assume that the waves down- way SNRs accelerate nuclei and electrons, and this is probably
stream of the shock move in a direction opposite to that of the the most promising avenue to follow for future progress.
shock. In the present paper we make the more conservative
assumption that somehow waves are isotropized downstream, so Acknowledgments
that hvWi = 0. Both choices lead to spectra of accelerated particles
which are steeper than the standard predictions of non-linear the- We are grateful to the referee, Don Ellison, for his valuable com-
ory. In the absence of this very important ingredient it is not pos- ments. This work was partially supported by ASI through Contract
sible to explain CR spectra as observed at Earth. The main ASI-INAF I/088/06/0.
difference between our assumption and that made by Ptuskin
et al. [38] is that in the latter case the spectral slope is strongly References
dependent on the Alfvénic Mach number, which leads the authors
to suggest that some, yet unknown, self-regulation mechanism [1] D. Allard, E. Parizot, A.V. Olinto, E. Khan, S. Goriely, A&A 443 (2005) 29.
[2] D. Allard, E. Parizot, A.V. Olinto, Astropart. Phys. 27 (2007) 61.
might be at work to guarantee the appropriate level of field [3] D. Allard, A.V. Olinto, E. Parizot, A&A 473 (2007) 59.
amplification. [4] D. Allard, N.G. Busca, G. Decerprit, A.V. Olinto, E. Parizot, JCAP 10 (2008) 033.
Ptuskin et al. [38] also made an attempt at taking into account [5] R. Aloisio, V. Berezinsky, P. Blasi, A. Gazizov, S. Grigorieva, B. Hnatyk, APh 27
(2007) 76.
the possibility that different classes of SNRs may contribute to the [6] R. Aloisio, V. Berezinsky, P. Blasi, S. Ostapchenko, Phys. Rev. D 77 (2008) 5007.
CR spectrum at Earth. However the relative abundances of chemi- [7] E. Amato, P. Blasi, MNRAS Lett. 364 (2005) 76.
cals injected in each class is the same and only the environmental [8] E. Amato, P. Blasi, MNRAS 371 (2006) 1251.
[9] A.R. Bell, MNRAS 182 (1978) 147.
conditions around the SNR (for instance gas density and tempera- [10] E.G. Berezhko, H.J. Völk, ApJ Lett. 661 (2007) 175.
ture) are changed in the different cases. [11] V. Berezinsky, A. Gazizov, S. Grigorieva, Phys. Lett. B 612 (2005) 147.
Although the introduction of these and other imaginable com- [12] V. Berezinsky, A. Gazizov, S. Grigorieva, Phys. Rev. D 74 (2006) 3005.
[13] P. Blasi, E. Amato, D. Caprioli, MNRAS 375 (2007) 1471.
plications does not present any serious technical problem for the
[14] P. Blasi, S. Gabici, G. Vannoni, MNRAS 361 (2005) 907.
calculation procedure discussed here, we decided to focus on the [15] J. Blümer, R. Engel, J. Hörandel, PrPNP 63 (2009) 293.
main physical ingredients and avoid additional assumptions that, [16] D. Caprioli, P. Blasi, E. Amato, M. Vietri, ApJ Lett. 679 (2008) 139.
[17] D. Caprioli, P. Blasi, E. Amato, M. Vietri, MNRAS 395 (2009) 895.
in our opinion, take away clarity from the results rather than mak-
[18] D. Caprioli, P. Blasi, E. Amato, MNRAS 396 (2009) 2065.
ing them more reliable. In our view, at the present stage of our [19] D. Caprioli, H. Kang, A. Vladimirov, T.W. Jones, MNRAS 417 (2010) 1773.
knowledge, the assumptions that are required to carry out more [20] D. Caprioli, E. Amato, P. Blasi, APh 33 (2010) 160.
sophisticated calculations are too many to lead to a scientifically [21] D. Caprioli, E. Amato, P. Blasi, APh 33 (2010) 307.
[22] D. Eichler, ApJ 229 (1979) 419.
trustable result. [23] D.C. Ellison, L.O’C. Drury, J.-P. Meyer, ApJ 487 (1997) 197.
This leads us to a question that is of relevance for the investiga- [24] D.C. Ellison, F.C. Jones, D. Eichler, JGZG 50 (1981) 110.
tion of the problem of CR origin from the theoretical point of view, [25] D.C. Ellison, E. Möbius, G. Paschmann, ApJ 352 (1990) 376.
[26] D.C. Ellison, D.J. Patnaude, P. Slane, J. Raymond, ApJ 712 (2010) 287.
namely ‘‘can we move beyond this point in a way that really adds to [27] J.C. Higdon, R.E. Lingenfelter, ApJ 628 (2005) 738.
our knowledge of the problem?’’. [28] A.M. Hillas, J. Phys. G 31 (2005) 95.
There are clearly several open issues that are worth thinking [29] J.R. Hörandel, APh 19 (2003) 193.
[30] J.R. Hörandel, AdSpR 41 (2008) 442.
about: (1) why are the observed spectra of protons and nuclei [31] J.R. Hörandel, N.N. Kalmykov, A.V. Timokhin, APh 27 (2007) 119.
somewhat different? (2) what is the effect of the winds of the [32] T. Kamae, N. Karlsson, M. Tsunefumi, T. Abe, T. Koi, ApJ 647 (2006) 692.
presupernova stars? (3) what is the best way to describe magnetic [33] P. Maestro et al., in: Proceedings of the 31st ICRC, Lodz 2009, 2010.
[34] M.A. Malkov, L.O’C. Drury, Rep. Prog. Phys. 64 (2001) 429.
field amplification and/or complex topology of the magnetic field
[35] J-P. Meyer, L.O’C. Drury, D.C. Ellison, ApJ 487 (1997) 182.
in the shock region? (4) how can we reliably describe the escape [36] J.-P. Meyer, D.C. Ellison, in: Ramaty, Vangioni-Flam, Cass, Olive (Eds.), LiBeB
of CRs from a SNR? and (5) what is the Physics of injection of nuclei Cosmic Rays and Related X- and Gamma-rays, ASP Conference Series, vol. 171,
1999, p. 187.
in a collisionless shock?
[37] V.S. Ptuskin, V.N. Zirakashvili, A&A 429 (2005) 755.
It is possible (although far from guaranteed) that each one of [38] V.S. Ptuskin, V.N. Zirakashvili, E.S. Seo, ApJ 718 (2010) 31.
these questions may be tackled from first principles, and if this [39] J. Skilling, MNRAS 172 (1975) 557.
ever happens, then the challenge will be putting together the con- [40] J.K. Truelove, C.F. Mc Kee, ApJ Suppl. Ser. 120 (1999) 299.
[41] M.E. Wiedenbeck et al., Space Sci. Rev. 130 (2007) 415.
clusions in a convincing way to reproduce the CR spectra observed