1

PHY104 : Electromagnetism and Light Homework # 2 Due on March 23, 2023

Electrostatic force and potential: application to particle accelerators

Solution

I. LINEAR ACCELERATOR

1) We call x the horizontal coordinate, starting from x = g0 = 0.

x
The electric field between to parallel grates at potentials V1 and V2
and separated by a distance l is uniform, directed along the normal to
g3
the grates (here, ~ex ), and its amplitude is E = (V1 − V2 )/l, so that
the force felt by an electron subject to it is Fl = −q(V1 − V2 )/l. Since
one gate out of two in the linear accelerator is grounded, the potential
difference between two grates is always ±V and alternates.

We want the electron to be always accelerated, which is to say that g2

V1 − V2 = −V along the electron’s path. This is true at t = 0 between
grates 0 and 1, and requires the electron to be at x = gk precisely at
time kτ , so that the switching happens as the electron passes through
the gates. In question 4, we will tune the length lk of each section to g1
make sure that the time it takes the electron to go through it is τ (cf.
figure, where regions in light gray correspond to a negative field, and
those in dark gray to a positive one).
0 τ 2τ 3τ t
To conclude: the field felt by the electron between grates gk and gk+1 is

~ k = − V ~ex
E (1)
lk+1

2) Assuming the electron only moves in the x direction, the work done by the field on the electron between gates gk
and gk+1 is
ˆ gk+1 ˆ gk+1 ˆ gk+1
~ = ~ =q V V
Wk = F~ · dx ~ k · dx
−q E dx = q (gk+1 − gk ) = qV (2)
gk gk gk lk lk

which is to say that the electron gains an energy qV in each section of the accelerator. More precisely, consider
Newton’s second law
d
m ~v = F~ . (3)
dt

Replacing F~ in the formula for the work and changing the variable to time yields
ˆ gk+1 ˆ (k+1)τ  ˆ (k+1)τ  (k+1)τ
d |~v |2 m|~v |2
    
d ~ d
Wk = m ~v · dx = m ~v · ~v dt = m dt = = Kk+1 − Kk . (4)
gk dt kτ dt kτ dt 2 2 kτ

We therefore have Kk+1 = Kk + Wk = Kk + qV = qkV . Since ~v is positive and in the x direction, we also have
that
r r
2Kk 2qV
~vk = ~ex = k ~ex (5)
m m
 2
p
3) For k large, we have vk+1 /vk = 1 + 1/k ∼ 1, i.e. the velocity does not change much between gk and gk+1 . We
therefore have
r
2qV
lk ∼ vk τ = τ k . (6)
m
• The total length of the accelerator is approximately
N N ˆ N √
2qV X √
r r r
X 2qV 2 3/2 2qV
gN = lk = τ k∼τ k= N τ (7)
m m k=0 3 m
k=1 k=1

which we approximate as a Riemann integral for N large.

• The final velocity out of the linear accelerator with N + 1 gates is vN = 2N qV /m and its kinetic energy is
p
2
KN = mvN /2 = N qV .
Doubling the kinetic energy means doubling the number of sections KN 0 /KN = 2 = N 0 /N , which increases the
length by a factor gN 0 /gN = 23/2 ∼ 2.83.

4) The equation of motion for the electron between gk−1 and gk , between times (k − 1)τ and kτ , is
qV
mẍ = (8)
lk
which we can twice integrate, first into
qV
ẋ ((k − 1)τ + t) = vk−1 + t (9)
mlk
where vn−1 is the initial velocity in this case, which is the velocity of the electron as he reaches x = gk−1 , and then
into
qV 2
x ((k − 1)τ + t) = gk−1 + vk−1 t + t . (10)
2mlk
The equation for lk is therefore
qV 2
x(kτ ) = gk = gk−1 + lk which is to say lk = vk−1 τ + τ , (11)
2mlk
which is a second degree polynomial equation in lk solved by
r
vk−1 τ 1 2 qV 2
lk = + vk−1 τ2 + 2 τ . (12)
2 2 m
This depends on the values of the velocities vk , which in turn verify
qV
vk = ẋ(kτ ) = vk−1 + τ. (13)
mlk

The straightforward way to solve these equations is by recursion. Let us look at the first few values of vk and lk :
r
qV
v0 = 0, l1 = τ (14)
2m
r r r r
2qV τ 2qV 1 2qV 2 qV 2 qV √
v1 = , l2 = + τ +2 τ =τ (1 + 2) (15)
m 2 m 2 m m 2m
r r r r r r
2qV 2qV 1 4qV τ 4qV 1 4qV 2 qV 2 qV √ √
v2 = + √ = , l3 = + τ +2 τ =τ ( 2 + 3) (16)
m m 1+ 2 m 2 m 2 m m 2m
r r r r r r
4qV 2qV 1 6qV τ 6qV 1 6qV 2 qV 2 qV √ √
v3 = + √ √ = , l4 = + τ +2 τ =τ ( 3 + 4) (17)
m m 2+ 3 m 2 m 2 m m 2m
 3
√ √ √ √
where we have used that 1/( a + b) = ( a − b)/(a − b). From these, it is easy to guess that

√
r r
2qV qV √
vk = k and lk = τ ( k − 1 + k) (18)
m 2m

and to check that this solves the equations

√
r r r r
2qV 2qV 1 2qV √ √ 2qV
vk = (k − 1) + √ √ = ( k − 1 + k − k − 1) = k (19)
m m ( k − 1 + k) m m
√
r r r
τ 2qV 1 2qV 2 qV 2 qV √
lk = (k − 1) + (k − 1) τ +2 τ =τ ( k − 1 + k). (20)
2 m 2 m m 2m

II. CYCLOTRON

1) The Lorentz force felt by the electron is F~ = −q~v × B~ez , which is orthogonal to ~ez . If ~v0 is in the plane xy, and
therefore also orthogonal to ~ez , then the whole trajectory will be in that plane.
Let us decompose ~v in Cartesian coordinates and write the equations of motion:

m(v̇x~ex + v̇y ~ey ) = −qB(vx~ex + vy ~ey ) × ~ez = −qB(−vx~ey + vy ~ex ). (21)

i.e.

mv̇x = −qBvy and mv̇y = qBvx . (22)

There are two equivalent ways to solve those equations. The first is to differentiate one equation and plug it into the
other:

mv̇x = −qBvy and mv̈y = qB v̇x so m2 v̈y = −q 2 B 2 vy (23)

and equivalently for vx . Using x(0) = y(0) = 0 = vy (0) and vx (0) = v0 as initial conditions, we find

vx (t) = v0 cos(ωt) and vy (t) = v0 sin(ωt) (24)

with ω = qB/m, which is called the cyclotron frequency of the electron. Equivalently, we can define Z = vx + ivy ,
which verifies the equation

mŻ = iqBZ solved by Z = v0 eiωt . (25)

In turn, we get, for the position, y

v0
v0 v0
x(t) = sin(ωt) and y(t) = (1 − cos(ωt)) (26)
ω ω
which describes a circle of radius vω0 and of center (0, v0 /ω) (cf. figure).
The electron will cross the plane x = 0 after half a turn, at time
t = π/ω, which doesn’t depend on v0 . Its position is then (0, 2v0 /ω) v0 m
qB B
and its velocity −~v0 (in blue on the figure).
The kinetic energy K = m~v · ~v /2 evolves according to

K̇ = m~v̇ · ~v = −q(~v × B~ez ) · ~v = 0 (27)

which is to say that it is constant under the action of a magnetic field v0 x

alone.

2) The electrons are injected with a small velocity close to the right cavity. From the previous question, and what
we saw in the first exercise, the trajectory of the electrons will be the following (all the fields and velocities will be
written with respect to the horizontal axis):
 4

• the electron is accelerated to the left by the field E = −V /d and

reaches the left cavity ;
• it enters the left cavity with a velocity −v1 , does a u-turn for a
duration t = π/ω, and exits with a velocity v1 ; the electric field
is switched in the meantime ;
• the electron is accelerated to the right by the field E = V /d and
reaches the right cavity ;
• it enters the right cavity with a velocity v2 , does a u-turn for a
duration t = π/ω, and exits with a velocity −v2 ; the electric
field is switched in the meantime...

and so on. The velocity is increasing during each acceleration, so the

radius of the u-turns are also increasing.

• As we have mentioned before, each acceleration phase gives an extra kinetic

p energy qV to the electron. After n
accelerations, it has a kinetic energy
p K n = nqV , and so a velocity v n = 2nqV /m. The radius of the next u-turn
will therefore be Rn = vn /ω = 2nmV /qB 2 . If the radius of the cavities is R, and provided that the electrons are
injected at the right spot to maximise the kinetic energy of the electrons coming out, they will have had a number of
accelerations of at most 2NR = R2 qB 2 /2mV (one tour of the system corresponds to two accelerations). Their kinetic
energy will then be KR = R2 q 2 B 2 /2m.
√
• To double the kinetic energy, we need KR0 /KR = R02 /R2 = 2, which is to say R0 /R = 2 ∼ 1.41. That is a bit
better than for linear accelerators, but building large magnets is difficult.

• We saw that each u-turn takes the same amount of time t = π/ω. Each acceleration takes less and less time,
however, because the velocity of the electrons is quickly increasing. The trajectory during the (n + 1)th acceleration,
from one grate to the other, is up to a sign

qV 2
x(t) = vn t + t (28)
2dm
p √ √ p
and reaches a distance d at time tn = d 2m/qV ( n + 1 − n) ∼ md2 /2nqV . The field has to be of the right
sign only during those periods. Since that duration goes to 0 for large n, a field of the form V (t) = −V cos(ωt) is
appropriate (with perhaps a small offset due to the first few traversals). In the following figure, we see the trajectory
of the electrons projected on the x axis (with the grey zones corresponding to the two cavities). An example of an
appropriate potential V is drawn in blue. The important parts of V are marked in red, the rest could be anything.

x V

t
 5

III. SYNCHROTRON

1) Just as in the previous exercise, the kinetic energy of the electrons K is constant inside of the magnetic bends,
and increases by qV in each linear portion of the accelerator. The kinetic
p energy after n accelerations is therefore
Kn = nqV and the entry velocity in a bend after n accelerations is vn = 2nqV /m. Since the radius R of the bends
is fixed, and that we needp
to have R = mv/qB, we have to make sure that the magnetic field after n accelerations is
fixed to Bn = mvn /qR = 2nmV /qR2 .

We can also express everything as a function of the time spent in the accelerator.
The time spent in the bends ispequal to one-fourth of a period for a given B, i.e. τ = πm/2qB, so that the time
spentpin the k-th bend is τk = π 2 mR2 /8qkV , and as before the time spent in the k-th linear section goes like
tk ∼ md2 /2kqV . This implies that the total time spent in the system after n accelerations goes like
n
X p p n
X p p √
Tn = τk + tk = π 2 mR2 /8qV + md2 /2qV k −1/2 ∼ π 2 mR2 /2qV + 2md2 /qV n. (29)
k=0 k=0

Plugging n as a function of Tn into Kn yields a quadratic function of time:

t2 8q 2 V 2 2
K(t) ∼ qV p 2 = 2t . (30)
π 2 mR2 /2qV +
p
2md2 /qV m (2πR + 4d)

√
Moreover, since both B and T grow like n, the magnetic field will have to increase linearly with time in the long
time limit:
p
2mV /qR2 4V
B(t) ∼ p p t= t. (31)
2 2 2
π mR /2qV + 2md /qV R(2πR + 4d)

Surprisingly, the charge and mass of the electron have disappeared! Notice that (2πR + 4d) is the total length of the
accelerator.
0
2) We do the same reasoning with an N -gonal accelerator with a radius
p R such thatp the length of a section 2πR0 /N
0 0 02 2 2
is the same as previously, i.e. R = N R/4. We need to have Bn = 2nmV p /qR = 32nmV p /qN R , and the time
0
spent in the k-th bend is one-N th of a period: τk = 2πm/N qB = 4/N π 2 mR02 /8qkV = π 2 mR2 /8qkV , i.e. the
exact same as before. The kinetic energy is therefore exactly the same as before, but we have
p
32mV /qN 2 R2 4V 4V
B 0 (t) ∼ p t= t= t (32)
R(2πR0 + N d)
p
π 2 mR2 /2qV + 2md2 /qV N R(πR/2 + d)

which is inversely proportional to the total length (2πR0 + N d) of the accelerator. This means that the longer it is,
the easier it is to produce the appropriate magnetic fields, while not affecting how fast K(t) grows.