a

r
X
i
v
:
1
2
0
7
.
3
7
0
6
v
1


[
a
s
t
r
o
-
p
h
.
H
E
]


1
6

J
u
l

2
0
1
2
LAPTH-024/12
Spectral breaks as a signature of cosmic ray induced turbulence in the Galaxy
Pasquale Blasi
e,p
, Elena Amato
e
and Pasquale D. Serpico
p
e
INAF-Osservatorio Astrosico di Arcetri, Largo E. Fermi, 5 50125 Firenze, Italy
p
INFN-Laboratori Nazionali del Gran Sasso, Assergi, Italy and
p
LAPTh, Univ. de Savoie, CNRS, B.P.110, Annecy-le-Vieux F-74941, France
(Dated: July 17, 2012)
We show that the complex shape of the cosmic ray (CR) spectrum, as recently measured by
PAMELA and inferred from Fermi-LAT -ray observations of molecular clouds in the Gould belt,
can be naturally understood in terms of basic plasma astrophysics phenomena. A break from a
harder to a softer spectrum at blue rigidity R 10 GV follows from a transition from transport
dominated by advection of particles with Alfven waves to a regime where diusion in the turbulence
generated by the same CRs is dominant. A second break at R 200 GV happens when the diusive
propagation is no longer determined by the self-generated turbulence, but rather by the cascading
of externally generated turbulence (for instance due to supernova (SN) bubbles) from large spatial
scales to smaller scales where CRs can resonate. Implications of this scenario for the cosmic ray
spectrum, grammage and anisotropy are discussed.
PACS numbers: 98.70.Sa
IntroductionThe spectrum of cosmic rays (CRs) as
observed at the Earth is not a perfect power law, even
below the knee, at 10
15
eV. Recent measurements by
the PAMELA experiment have shown that the spectrum
has a change of slope at blue 230 GV, e.g., for pro-
tons, from E
2.85
for E < 230 GeV to blue E
2.67
for E > 230 GeV [1]. Moreover, Fermi-LAT -ray data
between 100 MeV and 100 GeV from molecular clouds in
the Gould belt (but o the Galactic disc) show evidence
for a composite CR spectrum, with a slope 1.9 be-
low 10 GeV and 2.9 between 10 and 200 GeV
[3]. This analysis has recently been repeated by [4] who
nd very similar results. The spectrum of CR protons
as measured by PAMELA in the highest energy bins is
compatible with balloon measurements by CREAM [5].
At energies below a few tens of GeV the spectrum ob-
served by PAMELA is aected by solar modulation and
hence it is non-trivial to use these data for comparison
with the ndings of [3].
These spectral features are so far either attributed to
propagation eects from nearby sources or modeled with
corresponding breaks in either the injection spectrum of
CRs, or in the diusion coecient experienced by these
particles on their way through the Galaxy [6]. Since the
CR ux at Earth is dominated by distant sources, it ap-
pears rather unlikely that nearby sources may produce
uctuations of order unity at these energies [2]. Recently
[7] proposed that the change of spectrum may be induced
by a diusion coecient that is not separable in energy
and space.
In this Letter we propose that the complex observed
CR spectrum is telling us something about how and why
particles diuse in the interstellar medium (ISM) and de-
rives from diusion of CRs on a background of waves
partly due to self-generation and partly to wave-wave
turbulent cascading from a large scale, probably corre-
sponding to the size of SN bubbles.
Several authors have discussed the possibility that CRs
could be self-conned by the waves generated through the
streaming instability that they excite in the direction of
their spatial gradient (see [8, 9] for reviews). In particu-
lar [10] and [11] discussed the eect of self-generation in
the presence of ion-neutral damping and non-linear Lan-
dau damping (NLLD). The general conclusion in both
cases is rather interesting: in the Galactic disk and its
vicinities, waves are damped so fast that CR transport is
ballistic, while particles are dragged at the Alfven speed
(connement) only in the halo, much above or below the
Galactic disc, where the neutral gas density is suciently
low to avoid eective ion-neutral damping. However, the
results of these early works were not based on a self-
consistent solution of the transport equation. More re-
cently, Ref. [12] considered the problem of propagation of
CRs on their way out of the sources (e.g. SNRs): the au-
thors nd that bubbles of self-generated waves develop in
few hundred pc around SNRs, and that such waves dom-
inate the diusive CR propagation. The eective diu-
sion coecient is found to be D(E) E
0.6
. This slope is
roughly consistent with the slope of the Boron/Carbon
ratio, but leads to severe problems with CR anisotropy
if extrapolated to
>

TeV energies [13, 14]. Moreover

such approaches cannot explain the complex CR spec-
trum discussed above since the dependence of D(E) on
energy does not change.
Our current knowledge of the Galaxy suggests that,
although the average density of neutral gas is relatively
high ( 1 cm
3
), such gas is actually conned to regions
with small volume lling factor [15]. Most of the Galaxy
is lled with a tenuous ionized gas, where ion-neutral
damping is slow and the wave dynamics is most likely
determined by NLLD. The implications of the resonant
absorption of MHD waves by CRs in the Galaxy have
recently been investigated in [16] using GALPROP.
Moreover we have evidence that turbulence exists in
the ISM with a roughly Kolmogorov-type spectrum up
to spatial scales of order 50 pc or even larger [17]. This
2
turbulence is probably injected by SN explosions on such
scales and then cascades towards smaller scales. It only
becomes eective for CR scattering when the wavelength
becomes as small as the particle Larmor radius. Diusion
models routinely used in calculations of CR propagation
are inspired (implicitly or explicitly) by the assumption
that something like this happens.
In this Letter we present our calculations of the com-
bined eect on CR scattering of turbulence cascading
from some large scale (L
0
= 50 pc) through NLLD and
self-generated waves induced by CR streaming in the
Galaxy. We nd that a change in the scattering prop-
erties of the ISM must occur at 200 300 GV re-
ecting in a change of shape of the CR spectrum at the
same rigidity. While the transition energy can be esti-
mated analytically, we solve the full system of equations
describing CR transport and wave evolution so as to ob-
tain a self-consistent spectrum of CRs. In this way, we
also nd that at energies below 10 GV the advection
of CRs with waves moving with the Alfven velocity leads
to a spectral hardening. Both spectral features are ob-
served [1, 3, 4], and this work was actually stimulated by
these observations.
The calculationWe solve the CR diusion equation

z
_
D
f
z
_
+ v
A
f
z

dv
A
dz
p
3
f
p
= q
CR
(z, p) (1)
coupled with the equation for the waves:

k
_
D
kk
W
k
_
+
CR
W = q
W
(k). (2)
Here f(p, z) is normalized so that the number of particles
in the range dp around momentum p at the location z is
4p
2
f(p, z)dp. The diusion coecient is related to the
wave spectrum through the well known expression [10]:
D(p) =
1
3
r
L
(p)v(p)
1
k W(k)
, (3)
where the power in the form of waves, W(k), satises:
_

k0
dk W(k) =
B
=
B
2
B
2
0
, (4)
with B
2
/4 the power in turbulent elds and B
0
the
regular magnetic eld strength. In Eq. (3) the momen-
tum and wavenumber are related through the resonance
condition k = 1/r
L
(p) = qB
0
/(pc), with r
L
the Lar-
mor radius of particles with momentum p moving in the
magnetic eld B
0
. The underlying assumption is that
B B
0
. In Eq. (1) we use a simple injection model in
which all CRs are produced by SNRs in an innitely thin
disc of radius R
d
:
q
CR
(p, z) =

CR
E
SN
R
SN
R
2
d
I()c(mc)
4
_
p
mc
_

(z) q
0
(p)(z).
(5)
Here
CR
is the fraction of the total kinetic energy
of a SN, E
SN
, assumed to be channelled into CRs,
and the SN rate is R
SN
. The quantity I() =
4
_

0
dx x
2
_
x
2
+ 1 1

comes from the normaliza-

tion of the kinetic energy of the SN that goes into CRs.
Notice that the particle spectrum is assumed to be a
power law in momentum, as expected for diusive shock
acceleration in the test particle regime.
Eq. (2) describes the stationary wave spectrum W(k)
under the eect of wave-wave coupling and amplication
of waves due to streaming instability at a rate
cr
(k).
The cascade is due to NLLD and is described as a diu-
sion process in k-space with a diusion coecient [20]:
D
kk
= C
K
v
A
k
7/2
W(k)
1/2
(6)
for a Kolmogorov phenomenology (C
K
5.210
2
[18]).
One can easily check that, in the absence of a CR-induced
contribution, this diusive process in k-space leads to
the standard Kolmogorov spectrum W(k) k
5/3
(for
k k
0
), if the injection of power occurs at a single k
0
=
1/L
0
. The eect of CRs is to amplify the waves through
streaming instability, with the growth rate [10]:

cr
(k) =
16
2
3
v
A
k W(k)B
2
0
_
p
4
v(p)
f
z
_
p=qB0/kc
(7)
where the spatial gradient of CRs can be found by solving
the transport equation, Eq. (1).
Eq. (1) is solved in the simplifying assumptions that
D depends weakly on the the z-coordinate, and that the
Alfven speed is also independent of z, except for the fact
that Alfven waves move upward (downward) above (be-
low) the disk. This implies that dv
A
/dz = 2v
A
(z). With
these assumptions the solution of Eq. (1) can readily be
found to be in the form:
f(z, p) = f
0
(p)
1 e
(1|z|/H)
1 e

, (p)
v
A
H
D(p)
, (8)
and f
0
(p) has to satisfy the following equation, obtained
by integrating Eq. (1) in the range z = (0

0
+
):
2D(p)
_
f
z
_
z=0
+

2
3
v
A
p
df
0
dp
= q
0
(p). (9)
The space derivative can be easily derived from Eq. (8):
_
f
z
_
z=0
+
=
v
A
f
0
D(p)
1
(p)
, (p) = 1 exp[(p)] . (10)
Solving for f
0
, the CR spectrum in the disc of the Galaxy
is readily found to be:
f
0
(p) =
3
2v
A
_

p
dp

q
0
(p) exp
_
_
p

p
dp

3
(p

)
_
. (11)
In the high energy limit, where diusion prevails upon
convection at speed v
A
, Eq. (11) reduces to the well
3
known solution of the diusion equation in one dimen-
sion, f
di
0
(p) = q
0
(p)H/(2D(p)).
Eqs. (1) and (2) form a set of two non-linear dierential
equations. We solve them iteratively so that the nal re-
sults are the spectrum of CRs in the Galactic disc, f
0
(p),
and the power spectrum of waves, W(k), resulting from
self-generation and cascading. Before illustrating the ex-
act results, it is useful to estimate the energy of CR pro-
tons where one expects a transition from self-generated
waves to waves deriving from Kolmogorov cascade. In
order to do this, we use the fact that CR driven waves
saturate when the NLLD rate (
NL
D
kk
/k
2
, with D
kk
given in Eq. (6)) equals the growth rate
cr
(Eq. (7)). Us-
ing the expression f
0
(p) = A
p
(p/mc)
p
to describe the
spectrum, with A
p
and
p
taken from PAMELA data
above 250 GeV, we nd:
W
CR
=
_
16
2
3C
K
A
p
(mc)
4
B
2
0
H
_
eB
0
mc
2
_
4p
_2
3
k
2
3
(p
13
2
)
,
(12)
where k is related to the particle energy through the res-
onance condition. The transition length-scale, and hence
energy, is then found by simply equating W
CR
to the ex-
ternally generated Kolmogorov spectrum. We take the
latter in the form:
W
ext
(k) = (2
B
/3k
0
) (k/k
0
)
5/3
. (13)
Rather than guessing the value of
B
, we use again our
knowledge of the CR spectrum based on Pamela data,
and express
B
in terms of the CR acceleration eciency

CR
. This can be done by just recalling that in the high
energy regime f
0
(p) is well approximated by f
di
0
(p), with
D(p) given by Eq. (3). At scales at which the Kolmogorov
turbulence is the dominant scattering source, one nds:

B
=
R
2
d
I()A
p
_
mc
2
_
3
cH
CR
E
SN
R
SN
_
mc
2
qB
0
_
1/3
k
2/3
0
. (14)
Using this condition in Eq. (13) and equating it to
Eq. (12) we obtain:
E
tr
= 228 GeV
_
R
2
d,10
H
1/3
3

0.1
E
51
R
30
_
3
2(p4)
B
2p5
2(p4)
0,
, (15)
where R
d,10
= R
d
/10 kpc, H
3
= H/3 kpc,
0.1
=

CR
/0.1, E
51
= E
SN
/10
51
erg, R
30
= R
SN
/30 yr
1
,
B
0,
= B
0
/G.
The estimate obtained for the reference values of the
parameters is tantalizingly close to the energy where
PAMELA data show a change of slope of the spectrum of
protons from E
2.850.015
(E < 230 GeV) to E
2.670.03
(E > 230 GeV). Clearly the fact that al low ks (large mo-
menta) the power spectrum is W(k) k
1/3
implies that
the CR injection spectrum must be q
0
(p) p
4.3
. We
notice that the transition energy turns out to be indepen-
dent of the characteristic scale of turbulence, 1/k
0
: this
fact is especially important given the large uncertainty
on this parameter. What does depend on k
0
, linearly, is
the energy density in turbulent magnetic eld, which for
the above values of the parameters and L
0
= 50 pc turns
out to correspond to
B
8%.
Interestingly enough, taking into account that in the
advection dominated regime the solution is f
adv
0
(p)
q
0
(p)/v
A
and equating this to the solution in the diusion
regime determined by self-generated waves, it is easy to
see that for reasonable values of the Alfven speed the low
energy transition to a spectrum f
0
(p) p

occurs at
E
<

10 GeV.
Results and Discussion The iterative procedure de-
scribed above leads to the spectrum of CRs plotted in
Fig. 1 (solid line). The symbols are PAMELA data.
FIG. 1: Spectrum of CR protons from our calculations (solid
line) compared with the results from observations of -ray
emission from clouds in the Gould belt [3] (shaded region)
and with PAMELA measurements [1].
Since solar modulation aects the low energy part, only
the data above 80 GeV were used in [1] to highlight the
break at 230 GV. The dashed line is the solution of
the same equations but setting the self-generation term
to zero. The shaded area shows the CR spectrum in-
ferred in Ref. [3]. The break at 200 GeV reects
FIG. 2: Diusion coecient induced by streaming instability
of CRs and cascading from a large spatial scale of 50 pc.
the transition from a regime where the scattering cen-
ters are self-generated to a regime where particles diuse
in external turbulence that cascades from larger spatial
4
scales. This is well visible in Fig. 2, where we plot the dif-
fusion coecient with and without self-generation (solid
and dashed line respectively). The energy dependence of
the diusion coecient in the energy range 10 < E < 200
GeV is not a perfect power law, but if we approximate
it as such, we nd D(E) E
0.7
. At larger energies,
E > 200 GeV, the trend is D(E) E
1/3
, as expected
for a Kolmogorov cascade. At low energy the change of
slope is due to the transition to a non-relativistic regime,
but there the propagation is advection-dominated.
FIG. 3: Grammage obtained in our calculations (solid line)
compared with the leaky box t proposed in Ref. [19] (dashed
and dotted lines are the t for = 0.5 and 1/3 respectively).
The grammage traversed by CRs of velocity v
under the combined eect of diusion and advec-
tion at the Alfven speed is easily found to be
X(E) =
v
2 vA
_
1 exp
_

vAH
D(E)
__
, where = 2 nmh =
2.4mg/cm
2
is the surface density of the disc (correspond-
ing to a Galactic disc density n
d
1 cm
3
, for a typical
chemical composition of the interstellar matter). The
grammage obtained in our calculations is shown in Fig. 3
(solid line) and compared with the leaky box ts pro-
posed in [19] (dashed (dotted) line for = 0.5 ( = 1/3)).
The normalization in the 1 10 GeV range comes out
naturally in our calculations, together with the change
of slope in the diusion properties at 200 GeV.
The attening in the energy dependence of the diu-
sion coecient at high energies automatically avoids se-
vere problems with CR anisotropy. The mean anisotropy
amplitude is 10
3
at 1 TeV and increases with energy
as E
1/3
. However, as discussed in [14], this mean value
is not very meaningful in that the amplitude is domi-
nated by the nearest and most recent sources, and may
dramatically dier from the mean value. Nevertheless, it
is encouraging that the physical processes described here
naturally lead to steep energy dependence of the diu-
sion coecient in the low energy regime that does not
necessarily lead to violate observed data on anisotropy.
In summary, we showed that in a simplied but physi-
cally consistent model for the CR propagation in the ISM
the departure from a power-law spectrum in CRs mea-
sured at Earth is a consequence of basic processes. Both
a convective velocity of the order of v
A
in Eq. (1) and the
contribution of CRs to the wave spectrum in Eq. (2) are
unavoidable, albeit often neglected in phenomenological
studies. It is actually remarkable that without ad hoc free
parameters, the basic trend shown by the measured and
inferred CR spectra can be reproduced in a relatively sim-
ple model. The several approximations that have been
made (e.g. treatment of non-linear damping, absence of
re-acceleration, neglect of non-linear eects in diusive
shock acceleration) are not expected to change the qual-
itative picture that emerges. The scenario detailed here
also allows to accommodate the inferred behaviour of the
grammage as well asat least qualitativelythe strin-
gent constraints coming from anisotropy at high energy.
We nd encouraging that the recent bonanza of CR data
allows one to gain some insights on (astro)physical pro-
cesses involving CRs, beyond the mere task of providing
more accurate ts to injection spectra or propagation pa-
rameters; this is a trend which hopefully will be further
boosted in the near future with the expected results of
AMS-02.
Acknowledgements We thank A. Neronov, A. Tay-
lor for discussions and V. Formato and M. Boezio for
providing PAMELA data points. PB thanks LAPTh for
hospitality during the initial phases of this project. The
work of PB and EA is partially funded through PRIN
2010 and ASTRI grants.
[1] O., Adriani et al., Science, 332, 69 (2011).
[2] P. Blasi, and E. Amato, JCAP 1, 010 (2012).
[3] A. Neronov, D.V. Semikoz, and A.M. Taylor, PRL 108
1105 (2012).
[4] M. Kachelrie, S. Ostapchenko, arXiv:1206.4705.
[5] H.S. Ahn, et al., ApJ Lett. 714, 89 (2010).
[6] A.E. Vladimirov et al., ApJ, 752, 68 (2012).
[7] N. Tomassetti, ApJ Lett., 752, 13 (2012).
[8] C.J. Cesarsky, Ann. Rev. of A&A, 18, 289 (1980).
[9] D.G. Wentzel, Ann. Rev. of A&A, 12, 71 (1974).
[10] J. Skilling, ApJ ,170, 265 (1971).
[11] J.A. Holmes, MNRAS, 170, 251 (1975).
[12] V.S. Ptuskin, V.N. Zirakashvili, A.A. Plesser, Adv. Space
Res. 42, 486 (2008).
[13] V.S. Ptuskin, F.C. Jones, E.S. Seo, R. Sina, Adv. Space
Res. 37, 1909 (2006).
[14] P. Blasi, and E. Amato, JCAP 1, 011 (2012).
[15] M.A. Dopita & R.S. Sutherland 2002, Astrophysics of the
Diuse Universe (Berlin: Springer-Verlag).
[16] V. Ptuskin, et al., ApJ, 642, 902 (2006).
[17] J.W. Armstrong, B.J. Rickett, S.R. Spangler, ApJ, 443,
209 (1995).
[18] V.S. Ptuskin, V. N. Zirakashvili, A&A 403 1 (2003).
[19] V.S. Ptuskin, O.N. Strelnikova, L.G. Sveshnikova, As-
trop. Phys. 31 284 (2009).
[20] Y. Zhou, W.H. Matthaeus, J. of Geophys. Res., 95, 14881
(1990).