Lecture 5: MHD Shocks and

Particle Acceleration Processes

Aims, learning outcomes, and overview

Aim: To develop a detailed understanding of the physics of magnetohydrodynamic
shock waves at the fluid level, and of two particle acceleration mechanisms that
operate at shocks: shock drift acceleration, and diffusive shock acceleration.

Learning outcomes: At the end of this lecture, students are expected to:
explain qualitatively what shocks are and have a basic physical understanding
of them;
explain why shocks are important in space and astrophysical plasmas;
understand and explain qualitatively the jump conditions which represent
conservation of physical quantities at shock fronts and other plasma disconti-
nuities;
be able to derive and use the Rankine-Hugoniot relations and jump conditions
describing general magnetohydrodynamic shocks;
understand and describe qualitatively the varieties of MHD shocks and their
properties, including fast mode and slow mode shocks;
understand the processes sometimes called shock drift acceleration and
sometimes magnetic mirror reflection, describe their basic physics and prop-
erties, and be able to explain them in terms of both particle motions at the
shock front and reflection by a moving magnetic mirror;
understand Fermi acceleration and be able to describe its properties and phys-
ical origin.
Overview: A shock wave is a disturbance that moves through a fluid faster than
the characteristic speed of propagation of small amplitude waves in the medium (e.g.
faster than the adiabatic sound speed in an unmagnetised plasma, or faster than
the Alfven speed in a magnetised plasma). Shocks develop because of nonlinear
steepening, in which the propagation (group) speed of a wave disturbance increases
with amplitude. In this case, large amplitude Fourier components of a relatively
localized disturbance move faster than the small amplitude components, causing an
initially Gaussian disturbance to steepen towards an almost vertical front (Figure
5.1). This nonlinear steepening then leads to either wave breaking and overturning,
which lead to increased dissipation and intrinsic time variability for the disturbance,
or else to increased dissipation that is large enough to balance the steepening and
produce a time-stationary shock solution. The overall qualitative picture is one

1
Figure 5.1: Illustrations of (Top) the nonlinear steepening of a wave into a shock
and (bottom) wave breaking and overturning.

in which excess directed kinetic energy is redistributed into increased thermal

energy and sometimes magnetic field energy.
Shocks may be produced by explosions, or when a fluid moving faster than
the relevant characteristic speed encounters a slower-moving fluid, or they may be
produced when a wave encounters a region where the local characteristic speed
decreases in the direction of propagation of the wave. At a shock front, fluid pa-
rameters change discontinuously, because waves cannot propagate ahead to smooth
out the discontinuity.
Shocks are important in space physics because of their various effects, which in-
clude the heating, compression, and changed flow velocity (magnitude and direction)
of the plasma plasma, the amplification (or compression) and changed direction of
magnetic fields, the acceleration of small numbers of particles to high energies, and
the generation of . In this lecture the theory of MHD shocks is presented, together
with descriptions of two of the mechanisms whereby shocks accelerate particles
shock drift acceleration, and diffusive shock acceleration.

5.1 Jump conditions for shocks

Shock theory relates plasma properties behind the shock (referred to as downstream,
subscript 2) to those ahead of the shock (upstream, subscript 1).
A boundary (an interface) between two semi-infinite plasmas with different mass
densities, , temperatures, T , and magnetic fields, B, is illustrated in Figure 5.2.
The normal to the boundary is denoted by the unit vector n. By definition the
normal points upstream perpendicular to the boundary, pointing in the x direction
in Figure 5.2. This boundary represents a shock in a shock frame, namely a frame
in which the shock is at rest.
A special shock frame is the shock normal frame illustrated in Figure 5.2, in
which the flow u1 is normal to the shock front. It is always possible to make a
Lorentz transformation to the shock normal frame. Having found the properties of
a shock in that frame, a Lorentz transformation can be made back to the frame of

2
 Figure 5.2: The changes in ve-
upstream z downstream locity and magnetic field across
a shock front in the shock nor-
B2 mal frame. The shock is at
rest, the upstream plasma ap-
u2
u1 proaches with velocity u1 along
the shock normal n, and the
2
x
downstream plasma recedes at
1 u2 at an oblique angle to n. The
upstream magnetic field is at an
B1
angle 1 to n, and this changes
to 2 in the downstream region.
shock
Shocks with 1 = /2 are called
perpendicular and shocks with
1 = 0 (or 1 = ) are paral-
lel.

physical interest. These include the spacecraft frame, the laboratory frame, frames
in which the Sun or Earth are at rest, etc. In general in these frames the shock
is propagating obliquely from region 2 (downstream of the shock) into region 1
(upstream of the shock).
Boundary conditions that relate parameters across the interface between the two
media in a shock frame are called jump conditions, although they are also referred
to as Rankine-Hugoniot relations. If we denote the change in any quantity across
the boundary by square brackets, then the change in a scalar quantity, A, is

[A] = A1 A2 . (5.1)

The normal and tangential components of a vector V are, respectively,

Vn = V n, Vt = V (V n) n. (5.2)

and thus the change in Vn is

[Vn ] = (V1 V2 ) n. (5.3)

The MHD jump conditions are derived from the equations of continuity for mass,
momentum and energy, together with boundary conditions arising from Maxwells
equations. The conservation law for a scalar quantity, Q, with associated flux, F,
and source term SQ is
dQ Q
= + F = SQ . (5.4)
dt t
RIf the flow creating the shock is steady then the shock is not accelerating and
dV Q/t R = 0 where V is an arbitrary volume. If in addition the source
R term is
zero, then dV dQ/dt = 0, and the volume integral of (5.4) reduces to dV F =
0. Converting this into a surface integral gives, for any surface S,
Z
dS F = 0. (5.5)
S

Applying this to the surface illustrated in Figure 5.3, and letting the thickness of
the box shrink to zero leads to the jump condition

(F1 F2 ) n = [Fn ] = 0 (5.6)

3
 Discontinuity Figure 5.3: The surface S over
which the integral (5.5) is per-
formed is a box across the shock
Upstream Downstream with sides parallel and normal
Region 1 Region 2 to discontinuity.

which expresses the obvious requirement that for a conserved quantity, the incoming
flux (amount per unit time, per unit area) from region 1 is equal to the outgoing
flux to region 2.
For a steady flow with no source terms, the explicit form of (5.4) for mass
conservation (Q = mass density, F u, u = flow velocity) is

(u) = 0 , (5.7)

and so the jump condition that expresses mass conservation is, from (5.6) and (5.7),

[un ] = 0 . (5.8)

The corresponding form of (5.4) for energy is

 
1 2 P EB
u u + u+ = 0, (5.9)
2 1 0
where the first term describes the kinetic energy flux, the second is the flux of
thermal energy (the enthalpy flux) with the ratio of the specific heats, and the final
term is the electromagnetic energy flux (the Poynting vector). The jump condition
related to energy conservation follows by eliminating E from (5.9) by assuming that
the plasma is infinitely conducting so that E = u B. The Poynting vector is
then given by E B/0 = (B 2 u u B B)/0 , and it follows that

B2
 
1 2 uB
u un + P un + un Bn = 0. (5.10)
2 1 0 0
The conservation equation for momentum is more complicated because the mo-
mentum density u is a vector, and the counterpart of the flux is the stress tensor,
S. Denoting the ij-component of the stress tensor by Sij (= Sji ),

0 E 2 B2
 
1
Sij = ui uj + P ij + + ij 0 Ei Ej Bi Bj . (5.11)
2 20 0

The first term in (5.11) describes the stress due to the bulk motion, the second is
the isotropic pressure resulting from random thermal motions, the next two terms
are the electric and magnetic pressures, and the final terms describe the stresses
associated with the tensions along the electric and magnetic field lines. The coun-
terpart of (5.8) or (5.9) for the ith component of the momentum flux is of the form
Sij /xj = 0 (where a sum over j = x, y, z is implied). The quantity S n is

4
a vector with normal and tangential components. These components give jump
conditions expressing conservation of the normal and tangential components of the
momentum flux, respectively (terms involving the electric field are neglected):

Bt2
 
2
un + P + = 0, (5.12)
20

and  
Bn
un ut Bt = 0. (5.13)
0
The quantity Pram = u2n will be called the ram pressure below. It is vital in
understanding the qualitative physics and location of shocks and other related dis-
continuities.
In addition to the jump conditions (5.8), (5.10), (5.12) and (5.13), Maxwells
equations also require that the normal component of B and the tangential compo-
nent of E are continuous. The first of these conditions may be written

[Bn ] = 0. (5.14)

The continuity of the tangential component of E is equivalent to the continuity of

n E = (u B) n = u n B B n u,

which implies
[un Bt Bn ut ] = 0. (5.15)
We now combine (5.8), (5.10), and (5.12)(5.15) into a single equation. that can
be solved and used to describe the properties of shocks and related discontinuities.

5.2 Magnetohydrodynamic Shocks

The first step is to choose the shock normal frame, in which u1 = u1n . It is also
convenient to introduce the Alfven Mach number:
u1 2 B12
MA := , where v1A := . (5.16)
v1A 0 1
The Mach number and the angle between the shock normal direction and the
upstream magnetic field ( = 1 in Figure 5.1) are the two independent parameters
that define a magnetized shock. We also introduce the compression ratio, r := 2 /1
and note that the jump condition (5.8) implies
2 u1n
1 u1n = 2 u2n or r = = . (5.17)
1 u2n
The motivation is to eliminate all the variables relating to the downstream region.
The normal component of the magnetic field in the downstream region is removed
using the jump condition (5.14), which implies that B2n = B1n . Then u2t may be
eliminated between jump conditions (5.13) and (5.15), leading to

r(MA2 cos2 )
B2t = B1t . (5.18)
MA2 r cos2

and
u1 (r 1) cos
u2t = B1t . (5.19)
B1 (MA2 cos2 )

5
Finally, eliminating P2 between jump conditions (5.10) and (5.12) and using equa-
tions (5.18) and (5.19) leads to the equation for the Alfven Mach number
2
aMA2
 2
MA2
  2   
MA 2 2 MA r1 2
cos sin a+ a cos = 0,
r r r r 2
(5.20)
where := c21s /v1A
2
, with c1s = (P1 /1 )1/2 , and where 2a = + 1 r( 1).
Equation (5.20) has a variety of solutions, each of which describes a discontinu-
ous transition, but not all of which are shocks. The description shock is usually
restricted to compressive discontinuities (i.e. those for which r = 2 /1 > 1). This
restriction is physically motivated: it may be shown that entropy increases at the
discontinuity only for r > 1. Hence rarefactional shocks do not exist. In the fol-
lowing discussion we focus on solutions to equation (5.20) of relevance to space
physics.
First note that (5.20) is a cubic in MA2 , and so in general it has three solutions.
In the limit of a weak discontinuity, i.e. for r 1, a 1, (5.20) reduces to the
dispersion equation for the three MHD wave modes, viz. the Alfven mode, and the
fast and slow magnetoacoustic modes, as discussed in Lecture 2.
Second, accordingly, the solutions of (5.20) are classified according to the related
MHD modes, that is, as fast mode or slow mode shocks. The counterpart of Alfven
waves is not normally considered to be a shock because, like small amplitude Alfven
waves, it does not compress the plasma in general the Alfven mode corresponds
to a tangential discontinuity, or rotational discontinuity, at which the transverse
component of the magnetic field reverses direction. The solution in this case is
r = 1, MA2 = cos2 . A phenomenological distinction between fast and slow mode
shocks is that the magnetic field increases in a fast mode shock, and decreases or
stays the same in a slow mode shock.
Third, shocks involve a finite compression of the plasma, a property that can be
demonstrated by dividing equation (5.20) by (MA2 )3 and taking the limit MA .
This leads to the requirement r ( + 1)/( 1). Hence the density ratio r cannot
be arbitrarily large; for = 5/3 the limiting value is r = 4, so that no shock can
lead to a compression in density greater than a factor of four (provided = 5/3).
By reference to equation (5.18) it is clear that the increase in the magnetic field at
a strong shock is also limited to a factor of four.
Fourth, the Rankine-Hugoniot analysis above is based only on conservation of
mass, momentum, energy, and electromagnetic field quantities on either side of a
suitably narrow discontinuity and with suitable time and volume averaging. It does
not address the detailed structure of the transition layer, but only the properties of
the upstream and downstream plasma suitably far from the transition. Accordingly,
significantly larger compressions in density and magnetic field strength are possible
in the immediate vicinity of the shock, as discussed in later lectures. Moreover,
the jump conditions apply also to other types of boundaries, including planetary
magnetopauses and both rotational and tangential discontinuities.
For parallel propagation ( = 0), (5.20) has solutions
MA2 = r, MA2 = r/a, (5.21)
with MA2 = r being a double solution. The solution MA2 = r/a is an example of
a slow mode shock. This solution may be written (u1 /c1s )2 = r/a and corresponds
to an unmagnetized shock solution (the same solution is obtained from the jump
conditions with B = 0). In this case the magnetic field does not change across the
front, and so does not participate in the shock process. The other solution (MA2 = r)
is a fast mode shock. In this case neither u2 nor B2 is along n, and the solution is
called a switch-on shock because the tangential component of B switches on as the
front passes.

6
 Figure 5.4: Discontinuities in a
B1
B2
magnetized plasma: (a) a fast
mode shock, (b) a slow mode
B2 shock, (c) a switch-on shock,
B1
and (d) a tangential discontinu-
ity (an Alfvenic shock).

(a) (b)

B1

B1 B2
B2

(c) (d)

For perpendicular propagation ( = /2), (5.20) has the solution

MA2 = (r/a) [ + 1 + (r 1)(1 /2)] . (5.22)
In this case B2t /B1t = B2 /B1 = r, and since the magnetic field increases, this is a
fast mode shock. There is no perpendicular slow mode shock, which might be ex-
pected because (small amplitude) slow mode waves do not propagate perpendicular
to the magnetic field (Lecture 2).
The various phenomenological cases are illustrated in Figure 5.4.

5.3 Qualitative shock physics and properties

The requirement that r > 1 for a shock mean immediately from Eq. (5.8) that
U2n < u1n , so that the flow is slowed (in the shock normal frame) downstream of
the shock. From Eqs (5.18) and (5.19) it is clear that u2t 6= 0 and B2t 6= 0 in
general. Thus shocks tend to slow and deflect the incoming flow, in shock normal
frame.
Turning to Eq. (5.12), taking a factor of 1 u21n outside of the upstream term
it can be seen that the pressure and magnetic terms have characteristic size M S2
and MA2 , respectively. When MA , MS  1 the ram pressure term Pram = 1 u21n
dominates the upstream momentum. On the downstream side, however, the term
2 u22n = Pram /r so that the thermal pressure and magnetic terms must dominate.
Typically the thermal pressure terms dominate. Qualitatively, then, stationary
shocks can be understood in terms of a balance between the upstream ram pressure
and the downstream thermal pressure. Furthermore, normal momentum balance
allows us to estimate the amount of downstream heating required for a stationary
solution:
P2 = 2 kB T2 = Pram = 1 u21n . (5.23)
Typically the amount of heating required is large. This conclusion is robust since
it is essentially based only on conservation of mass, momentum, energy, and field
quantities. Put another way, the shock system must adapt until it finds a heating
mechanism able to satisfy the conservation laws or else must be intrinsically time-
variable and non-stationary.

7
 Turning to Eq. (5.18) for fast mode shocks in the limit MA2  r cos2 it is easy
to see that the MA2 terms dominate the numerator and denominator, but cancel so
that B2t rB1t . With B1n = B2n from Eq. (5.14) it is now clear that both B and
are amplified by the same approximate factor of r.
Characteristically, then, fast mode shocks tend to slow, deflect, compress, and
heat the incoming plasma, as well as amplifying and rotating the magnetic field.
These characteristics lead to significant acceleration and heating of plasma particles,
as described next.

5.4 Shock drift acceleration and magnetic mirror reflec-

tion
Shock drift acceleration and magnetic mirror reflection both involve the energization
of particles by shock fronts due to the particles undergoing B (and curvature)
drifts parallel to qEcon , where Econ is the convection electric field in the upstream
(and downstream) plasma. These particles can be reflected by or transmitted from
the shocks magnetic gradient, with the largest energy gains occurring for reflected
particles. Historically the former term was applied primarily to particles with large
gyroradii, which saw the shock as an abrupt transition, while the latter was applied
primarily to particles with small gyroradii for which adiabatic motions characteristic
of orbit theory apply.
Shock drift acceleration and magnetic mirror reflection occur at a shock front
when a particle is reflected or transmitted by the front. A number of conditions
need to be met for the process to be an efficient acceleration mechanism. First,
the speed of the particle needs to be larger than the speed of the shock relative
to the upstream plasma, so that a particle undergoes many orbits as it passes the
front. Second, the largest energies are achieved at shock fronts that are nearly
perpendicular (i.e. 1 /2 in Figure 5.2).
The basic mechanism may be understood from Figure 1.5 of Lecture 1 and
Figure 5.5 of this Lecture. Figure 5.5(a) shows a perpendicular shock viewed in
the shock normal frame (the shock surface lies in the x-y plane). Figure 5.4(b)
illustrates the same shock viewed along the x-axis, together with the orbit of an
electron crossing the shock. Because the magnetic fields (and hence the gyroradii)
are different on the two sides of the shock, there is a gradient in the magnetic field
and so the particle undergoes a B drift along the shock front (see Lecture 1), as
illustrated in the figure. There is an identical electric field on both sides of the shock
front, actually the standard convection electric field associated with the frozen-in
magnetic field, given by
E = u1 B1 = u2 B2 . (5.24)
This electric field accelerates the electron during its motion on the large circle (in
the upstream region) and decelerates the electron during its motion on the small
circle (downstream). However, the particle spends more time moving along the
large circle than it does along the small circle, and so it gains energy during the
drift along the shock front. Put another way, the particle gyrocenter has qv d .E > 0,
where vd is the B drift velocity given by Eq. (1.25), so that there is a continual
net positive acceleration. Note that in this case there is zero curvature drift.
In the more general case of an oblique fast mode shock both the B and curva-
ture drifts can contribute, often competing. Which one dominates the energy gain
then depends of the magnetic field gradient, curvature, and the values of v and vk
for the incident particle.
Shock drift acceleration and magnetic mirror reflection result in relatively mod-
est increases in particle energies. Quantitative treatments of the theory show that

8
 x y Figure 5.5: (a) A perpen-
dicular shock in the nor-
mal incidence frame. (b)
z The orbit of an electron
B1 B2 crossing the shock.

z
E E
u1 u2

(a) (b)

the increase is at most a factor of approximately 10 for a single encounter with

the shock. However, larger gains are possible if the particles encounter the shock
multiple times, due to either some scattering process or due to the combination of
shock curvature and upstream magnetic field curvature.
This process can also be treated directly in terms of reflection of a particle by the
shocks moving magnetic mirror, providing an alternative way to think about shock
drift acceleration. An analogy with hitting a ping-pong ball is helpful, immediately
telling us that a more rapidly moving reflected particle results from either a faster
moving ping-pong bat or a particle with larger initial velocity directed towards the
shock (larger relative velocity).
A particle encountering a shock may either be reflected or transmitted at the
shock, and in both cases the shock drift mechanism operates to increase the energy
of the particle. A condition for reflection of a particle may be obtained as follows.
First, it turns out that the magnetic moment of the particle, which can be written
as p2 sin2 /B where is the pitch angle, is almost conserved (in an average over
gyrophase) when a particle crosses a shock front, so that

p22 sin2 2 p2 sin2 1

= 1 . (5.25)
B2 B1
Remarkably this is true not just for particles with Larmor radii rl  L, the shock
thickness, which we expect from adiabatic theory, but also for particles with rl  L.
For a quasi-perpendicular shock there is a special shock frame, called the de
HoffmannTeller frame (dHT), in which the convection electric field vanishes (and
the points of intersection of field lines and the shock front are at rest). In the dHT
frame the fluid velocity is parallel to the field lines on both sides of the shock (if
this were not so, the plasma motion perpendicular to the field lines would imply
a motion of the magnetic field at the shock front, since the field is frozen in to
the plasma). According to Eq. (5.24) there is no electric field in the dHT frame,
and so the energy (and hence the magnitude of the momentum) of the particle
is conserved. (More accurate non-MHD theory shows that a cross-shock potential
exists; this effect can be included easily but is ignored here.) Denoting quantities
in the dHT frame by primes, we have |p01 | = |p02 |, and hence from equation (5.25),
sin2 02 = (B20 /B10 ) sin2 01 . Particles with sin2 01 > B10 /B20 would have pitch angles
such that sin2 02 > 1 if they were transmitted through the shock. This is of course
impossible, and so these particles are reflected rather than transmitted. Hence the
condition for reflection is
sin 01 > (B10 /B20 )1/2 . (5.26)

9
This type of magnetic bottle condition was discussed in Lecture 1. Eqs (1.41)
and (1.42) apply also, and can be generalized easily to include the cross-shock
potential. As in Lecture 1 it also follows that the reflected particles have a loss-
cone distribution, while the transmitted particles are those that lie in the shocks
loss cone. Thus, both the reflected and transmitted particles will have distribution
functions that are significantly modified by the magnetic mirror, with gradients in
velocity space that are conducive to wave growth.
Consider a reflected particle, with a velocity before reflection denoted by the
subscript i, and a velocity after denoted by f . Transforming the parallel velocity of
the particle from the dHT frame back to the frame of the upstream plasma gives
(in the nonrelativistic approximation)
k k0 k k0
vi = vi u01 , vf = vf u01 . (5.27)

k0 k0
By the choice of the dHT frame we have vi = vf . From these equations

k k0
vf = vi u01
k
= vi 2u01 . (5.28)

Also, the velocity of the shock in the normal direction is u = u01 cos 10 , and
noting that the magnetic field does not change between frames in nonrelativistic
transformations (so that the angle of the magnetic field is unchanged) we have
k k
vf = 2u sec 1 vi . (5.29)

The speed usec1 is the speed of the dHT frame along the shock surface. Equation
(5.27) demonstrates that larger energy increases (for reflected particles) occur for
shocks that are more nearly perpendicular and/or moving faster relative to the
upstream plasma. This recovers the result expected from the ping-pong bat analogy.
Note that for exactly perpendicular shocks the dHT speed exceeds the speed of
light, so there is no dHT frame and a different approach must be used. In this cases
there are no reflected particles, though, and all incident particles are transmitted
through the shock. This can be seen directly from Figure 5.5, because u1 equals
the E B velocity and B is along the shock surface so that all particle gyrocenter
velocities are necessarily directed downstream.
The strongest evidence for shock drift acceleration and magnetic mirror reflec-
tion comes from observations of shocks in the heliosphere, in particular the Earths
bow shock and interplanetary shocks. These cases will be treated in detail in later
lectures (Lectures 10-14).

5.4 Diffusive shock acceleration

Diffusive shock acceleration is an example of a mechanism first recognised by Fermi
in 1949 and now called the Fermi mechanism. The specific problem discussed by
Fermi was the acceleration of cosmic rays via collisions with magnetized interstellar
clouds. The magnetic field of a moving cloud can reflect (or scatter) a cosmic ray,
as illustrated in Figure 5.6. A particle reflected from an approaching surface gains
energy, and a particle reflected from a receding surface loses energy, just as for a
ping-pong bat, so in general the particles energy changes as a result of the reflection.
This process can be treated quantitatively by noting that there is no energy
change in a frame in which the reflecting surface is at rest. Let u denote the
velocity of the reflecting surface in the observers frame, v1 and v2 denote the
particles velocity before and after reflection, and indicate velocities measured in the

10
 Figure 5.6: Fermi ac-
u
celeration: a cosmic
ray is reflected by an
approaching magnetised
cloud (shaded region).
p1

p
2

rest frame of the reflecting surface by primes. Assuming all motions are collinear
for simplicity, and restricting ourselves to the non-relativistic case (the relativistic
generalisation is straightforward), we have

v10 = v1 u, v20 = v2 u . (5.30)

In the rest frame of the reflecting surface, the particle does not gain energy in
reflection and so v20 = v10 . Equations (5.30) then give v2 = 2u v1 , analogously to
magnetic mirror reflection in the previous section, and hence the observed change
in energy of the particle is
1
= m(v22 v12 ) = 2m(u2 v1 u). (5.31)
2
As stated above, energy gains result from head on collisions (v1 u < 0) and energy
losses result from overtaking collisions (v1 u > u2 ). When energy changes are aver-
aged over large numbers of collisions, the terms which are first order (in u) sum to
zero. However, the second order changes, due to the term 2mu2 in (5.31), do not
average to zero and so a acceleration of the particle results. This process is now
often referred to as second-order Fermi acceleration.
In a subsequent paper in 1953. Fermi pointed out that the acceleration can
be much more efficient if it is first order, and that first order changes need not
always cancel. A simple example is when reflecting surfaces approach each other.
A particle bouncing back and forth between approaching surfaces is always reflected
head on and so the first order changes in energy are always positive. Processes of
this type are called first-order Fermi acceleration.
First-order Fermi acceleration occurs naturally at any shock front, a process
called diffusive shock acceleration, provided only that fast particles are scattered
on either side of the shock. The only important point needed to understand this
process is that the fluid velocity changes across the shock. Imagine a fast particle,
with speed v  u1 , in the upstream plasma about to cross the shock and enter the
downstream plasma. Recall the general shock configuration shown in Figure 5.2.
For simplicity we consider a simple case in which u1 and u2 are normal to the shock.
MHD waves can efficiently scatter fast particles, and these waves are convected along
with the plasma. MHD waves in the downstream plasma are therefore moving at
the fluid velocity u2 , and so these centers are approaching with velocity (u2 u1 )
with respect to the upstream plasma, as shown in Figure 5.7(a). When the particle
enters the downstream plasma and is scattered, its energy is increased because
the reflection is head on. In effect the scattering centers act like particles with
infinite mass, so that the reflection is analogous to that from a moving wall. Now
consider a particle in the downstream plasma about to cross the shock and enter the
upstream plasma. As viewed from the downstream plasma, the upstream plasma
is approaching with velocity u1 u2 , and the shock is receding with velocity -u2 ,

11
as illustrated in Figure 5.7(b). The scattering centers in the upstream plasma are
again approaching the particle. As before, when the particle crosses the shock and is
scattered, the scattering is head on and so causes an increase in the particle energy.

upstream downstream upstream downstream

u1 u2

z z

u1 u2
(u 1 u 2)

(a) (b)

Figure 5.7: (a) A particle, denoted by the black circle, at rest in the upstream
plasma sees both the shock, and (more slowly) the downstream plasma behind the
shock, approaching it. (b): As for (a) but for a particle at rest in the downstream
plasma which again sees the plasma on the other side of the shock approaching, but
sees the shock receding.

Fermi also gave a simple argument that the first-order process leads to a power-
law distribution for the energies of accelerated particles, as follows. The energy
change from one head-on reflection, according to (5.31) is = 2m|v1 u|, where we
have assumed that the speed of the particle is much greater than the speed of the
obstacle, and so have neglected the first term in (5.31). Now consider subsequent
reflections from approaching obstacles, separated by a distance L. For |u|  |v1 |,
we can consider L to be almost constant. The time between collisions is then
t = L/|v1 |, and dividing by t gives the rate of gain of energy of the particle,
d/dt = 2m|u|v12 /L = 4(|u|/L). Hence the energy of the particle after a time t is

= 0 exp(t), (5.32)

where = 4|u|/L. Now consider how the particle escapes from the system. A
reasonable assumption is that the particle has a constant probability of escaping per
unit time, described by a mean loss rate . In that case the probability distribution
for the particle lasting a time t is the Poisson interval distribution,

prob(t) = et (5.33)

[prob(t) dt is the probability that a particle lasts a time t in the system]. The
probability distribution function for the energy of the particle, prob() is obtained
by changing variables: prob()d = prob(t) |dt/d| d. Using (5.32) and (5.33) leads
to prob() (1+/) .
Provided that particles cross the shock many times, this mechanism allows ef-
ficient acceleration. Moreover, fast particles are expected to cross the shock many
times if they are efficiently scattered on each side of the shock. This is because
scattering causes spatial diffusion, so that the motion of a typical scattered particle
has a random component that gives it a high probability of returning to the shock
many times. Collectively, particles diffuse away from the shock. Particles wander-
ing upstream reach a steady state with those wandering back to the shock, and the
upstream particle distribution falls off with distance from the shock, approaching
zero far from the shock. Particles wandering downstream can reach arbitrarily far
downstream and never return, so that the downstream distribution approaches a
constant arbitrarily far away from the shock.

12
 A quantitative treatment of diffusive shock acceleration predicts that, given a
mono-energetic spectrum of injected particles, the downstream spectrum of accel-
erated particles follows a power law form f (p) pb , where the power-law index b
depends only on the compression ratio of the shock: b = 3r/(r 1). This result is
consistent with observed cosmic ray energy spectra.

References
This lecture is based on a lecture by M. Wheatland, itself based on an outline of mine and
earlier lectures by D.B. Melrose and L.T. Ball. In addition the following references were
used.

Huba, J.D. 2006, NRL Plasma Formulary, Naval Research Laboratory, Washington DC
(may be ordered for free from http://wwwppd.nrl.navy.mil/nrlformulary/)

Melrose, D.B. 1986, Instabilities in Space and Laboratory Plasmas, Cambridge University
Press, Cambridge, pp. 23.-250

Thompson, W.B. 1962, An Introduction to Plasma Physics, Addison-Wesley, Reading,

Massachusetts, pp. 8695

Toptyghin, I.N. 1980, Space Science Reviews, 26, 157213

Yuan, X., I.H. Cairns and P.A. Robinson, Numerical simulation of electron distributions
upstream and downstream of high Mach number quasiperpendicular collisionless shocks,
J. Geophys. Res., 113, A08109, 2008.

13