PHY3011 Wells and Barriers page 1 of 17

1. The infinite square well

First we will revise the infinite square well which you did at level 2. Instead of the
well extending from 0 to a, in all of the following sections we will use a well that extends
from −a to a, that is, twice as wide and centred at 0. Of course the solutions are the
same: they’re just shifted and scaled (see Robinett, chapter 5).
The potential is
0, for − a ≤ x ≤ a,

V (x) =
∞, for |x| > a,
Outside the well where the potential is infinite there is no probability for the particle to
be found and so ψ(x) = 0 for |x| > a. Inside the well the potential is zero so the time
independent Schrödinger equation reads

h̄2 d2 ψ
− = Eψ
2m dx2
which we write
d2 ψ
= −k 2 ψ
dx2
and
1p
k= 2mE
h̄
There are no normalisable solutions for E < 0 and so k is real and the general solution is
ψ(x) = A sin kx + B cos kx
The boundary conditions are
ψ(−a) = ψ(a) = 0
so
A sin ka + B cos ka = 0
−A sin ka + B cos ka = 0
Adding or subtracting these we get either
A sin ka = 0 or B cos ka = 0
Because the potential has inversion symmetry, V (−x) = V (x), it is natural that the
solutions fall into two classes: even parity solutions having ψ(−x) = ψ(x), and odd parity
solutions having ψ(−x) = −ψ(x). And the two possible boundary conditions give rise to
one of these sets each. Specifically since we cannot have both A and B equal to zero
(since then ψ = 0 everywhere and this cannot be normalised) the two sets correspond
respectively to either A = 0 or B = 0.
The boundary condition A = 0 leads to the even parity solutions. In this case we have
cos ka = 0 which is true if
e = m − 1 π = (2m − 1)π ,
 
k = km m = 1, 2, 3, · · ·
2 a 2a
and the eigenfunctions of the even solutions are
ex
ψm (x) = B cos km
PHY3011 Wells and Barriers page 2 of 17

The other case, B = 0 leads to the odd parity solutions with boundary condition A sin ka =
0 implying
k = kmo = mπ = 2mπ , m = 1, 2, 3, · · ·
a 2a
with eigenfunctions
ψm (x) = A sin kmox

We note that the eigenfunctions can be written as

C cos kn x, n = 2m − 1 i.e., n odd

ψn (x) =
C sin kn x n = 2m i.e., n even
and
nπ
kn = , n = 1, 2, 3, · · ·
2a
The boundary condition fixes the allowed wavelength λn = 2π/kn . kn is called a quantum
number and generally speaking quantum numbers arise as labels which dictate the allowed
solutions under the given boundary conditions. The allowed p energies are the eigenvalues
associated with the values of kn . Since we will have kn = 2mEn /h̄ it follows that

h̄2 kn2 n2 π 2 h̄2

En = =
2m 2m(2a)2
To find the constant C, we normalise the wavefunction.
Z a Z a
 2  2
C  cos2 kx dx = C  sin2 kx dx = a C 2 =1
 
−a −a
√
and so we can take C = 1/ a.
Please note that these ψn (x) are eigenvectors. A stationary state of the infinite square
well is
Ψn (x, t) = ψn eiEn t/h̄
Now comes an important point. The eigenvectors

ψn (x) = |ni

provide a basis in which to express any wavefunction that satisfies the boundary conditions
laid down by the potential (in this case the infinite square well) just like the basis vectors
ı̂ , ̂ and k̂ are a basis for any cartesian vector. Mathematically we can write the state of
any particle in the infinite square well as a linear combination
∞
X
ψ(x) = cn ψn (x)
n=1
X∞
= cn |ni
n=1

(In the second line I have used the vector notation to describe the eigenvector.) Unlike
cartesian space, this space is infinite dimensional. It’s exactly the same as Fourier analysis
which says that any function in some interval, which is zero at the ends, can be expanded
in sines and cosines having the same boundary conditions. Using Fourier analysis we find
PHY3011 Wells and Barriers page 3 of 17

the important result that the coefficients cn in the expansion above can be found using
an integral, Z a
cn = ψn∗ (x) ψ(x)dx
−a
= hn ψ
 

And, again, I have indicated the alternative vector notation that emphasises this as a
scalar product between the state and one of the eigenstates. Another property of the
basis functions is that they are orthonormal,
Z a
∗
ψm (x) ψn (x) dx = δmn
−a
hm |ni = δmn

where δmn is the “Kronecker delta,” it’s zero if m 6= n and one if m = n. The fact that
the scalar product between two eigenstates is zero unless they are identical is completely
analogous to the orthogonality between the cartesian unit vectors, ı̂ , ̂ and k̂ . For example

̂ · k̂ = 0

but
ı̂ · ı̂ = 1
This will become a lot clearer when I do an example in class.
PHY3011 Wells and Barriers page 4 of 17

2. The finite square well

The potential is
− V0 , for − a ≤ x ≤ a,

V (x) =
0, for |x| > a,
with V0 > 0. We first consider states with energy E < 0. Classically they would be
confined within the well. In quantum mechanics a particle is represented by a solution
to the time independent Schrödinger equation. The strategy is to find solutions in the
three regions, left of the well, in the well and right of the well. To the left, x < −a, the
potential is zero and we have
h̄2 d2 ψ
− = Eψ
2m dx2
which we write
d2 ψ
= κ2 ψ (1a)
dx2
with
1p
κ= −2mE (1b)
h̄
The general solution is
ψ(x) = Ae−κx + Beκx
but we must have A = 0 or this blows up at large negative x.
Inside the well the time independent Schrödinger equation reads

h̄2 d2 ψ
− − V0 ψ = Eψ
2m dx2
which we write
d2 ψ
= −l 2 ψ (2a)
dx2
with
1q
l= 2m(E + V0 ) (2b)
h̄
While E < 0 it’s also true that E > −V0 because there is no normalisable solution in the
well for energy less than −V0 (why not?). Please note the signs on κ and l on the right
hand sides of (1a) and (2a). Solutions of equations like (1a) with positive coefficent are
always exponentially decaying, like e±κx , while if the coefficient is negative as in (2a) the
solution is oscillatory like e±ilx which we can also equally correctly write in terms of sines
and cosines:
ψ(x) = C sin lx + D cos lx
Note it is conventional to use the symbols k or l for the quantum number in the oscillatory
case and to use the symbol κ in the decaying case. In your level 2 notes you used α rather
than κ as I shall also do in handwritten notes, since k and κ look very similar.
To the right of the well the potential is again zero and the energy of the state is less
than this so the particle is classically forbidden this region. The signature of this is an
exponentially decaying wavefunction just as to the left of the well. So for x > a we have

ψ(x) = F e−κx
PHY3011 Wells and Barriers page 5 of 17

The procedure is one we will use in all problems of this type. First we write down the
wavefunction and its first derivative with respect to x in the three regions.
κx
 Be , for x < −a


ψ(x) = C sin lx + D cos lx, for − a ≤ x ≤ a,

F e−κx , for x > a


κx
 κBe , for x < −a


ψ ′ (x) = lC cos lx − lD sin lx, for − a ≤ x ≤ a,

−κF e−κx , for x > a


The task at hand is to find those values of the constants B, C, D and F that render the
wavefunction continuous and differentiable over the whole range of x. I will show later
for the case of states with E > 0 how this is done in a general way. We approach this
particular problem, E < 0, in a way that exploits the symmetry of the potential, namely
that it is symmetrical about x = 0 so that the wavefunction solutions must be either of
even parity (ψ(x) = ψ(−x)) or of odd parity (ψ(x) = −ψ(−x)). We then seek first the
even solutions and then the odd solutions. For the even solutions we have

 ψ(−x), for x < −a



ψ(x) = D cos lx, for − a ≤ x ≤ a,

F e−κx , for x > a


For ψ to be continuous at x = a we must have

F e−κa = D cos la

and to be differentiable we need

−κF e−κa = −lD sin la

Dividing these two equations one by the other we get

κ = l tan la even solutions (3)

and this is the answer. Looking at (1b) and (2b) you see that the allowed energies, E,
are those which result in κ and l obeying (3). For each allowed energy, we get κ and l
which we plug into the wavefunctions in the three regions. At the same time we can solve
for the constants F and D (one of which is arbitrary and can be chosen to normalise the
wavefunction over the range of x).
Now, for the odd solutions we write

 −ψ(−x), for x < −a



ψ(x) = C sin lx, for − a ≤ x ≤ a,

F e−κx , for x > a


and matching value and slope at x = a results in

ψ(a) = F e−κa = C sin la

ψ ′ (a) = −κF eκa = lC cos la
PHY3011 Wells and Barriers page 6 of 17

which when we divide one by the other, we arrive at

l
−κ = odd solutions (4)
tan la
Equations (3) and (4) do not have analytical solutions but we can find the answers graph-
ically or by computation. To do this we define some dimensionless parameters,
aq
z = la, and z0 = 2mV0
h̄
and (3) and (4) become, in terms of z and z0

 1/2
 2
 (z0 /z) − 1 , even solutions



tan z =  −1/2
 2
 (z0 /z) − 1 , odd solutions




even solutions

z0=16

z0=8

z0=8 z0=16

odd solutions
π 2π 3π 4π 5π
To find even and odd solutions we just have to plot tan z against the right hand sides.
The figure shows this plot for two cases, z0 = 8 and z0 = 16. Note that z0 is proportional
to the width and the square root of the depth of the well; it therefore is a measure of the
“strength” of the potential well. From (2b) we have

h̄2 l2 h̄2
En + V0 = = z2
2m 2ma2
PHY3011 Wells and Barriers page 7 of 17

In the case of a very deep well, we see from the figure that the intersections occur
near the infinities of tan z, namely
zn → nπ/2 as z0 → ∞
and so in the limit of an infinitely deep well we get that the eigenenergies measured from
the bottom of the well, −V0 are
n2 π 2 h̄2
2m(2a)2
which are of course the energies of a particle trapped in an infinite square well of width
2a.
In the other limit, as z0 → 0, we see from the figure that there are fewer and fewer
states “bound by the potential” and for z0 < π/2 there are no odd states and just one
even state. It is significant that this survives however weak is the potential: even the
weakest and narrowest square well will bind at least one state.
The next figure shows this lowest energy bound state and also the second and sixth
states, which are odd and even respectively, for z0 = 16. These have the shapes of the
states of the infinite square well, but unlike classical particles, the quantum mechanics
admits a non zero probability that a measurement will find the particle outside the walls
even though its energy is less than zero. I also show the eigenvalues of the first 10 bound
states. You note that the probability density of a measurement finding the particle outside
the well increases with the quantum number n. You might think this is inconsistent with
Bohr’s correspondence principle which asserts that the classical limit corresponds to larger
quantum numbers; but in this case the increased probability is due to the energy getting
closer to zero after which the states and are free from the well altogether. These states,
having E < 0, are discussed next.
PHY3011 Wells and Barriers page 8 of 17

So those were the particles whose energies were below zero so that they are trapped
in or near the well. Now we come to particles whose energies are greater than zero so they
can exist as propagating stationary states anywhere in the range of x (−∞ < x < ∞).
I will do this using the general method of “transfer matrices.” First we write down
the wavefunctions and derivatives in the three regions as before.

 Aeikx + Be−ikx ,
 for x < −a
ψ(x) = C sin lx + D cos lx, for − a ≤ x ≤ a,
F eikx + Ge−ikx ,

for x > a



 ikAeikx − ikBe−ikx ,
 for x < −a
′
ψ (x) = lC cos lx − lD sin lx, for − a ≤ x ≤ a,
ikF eikx − ikGe−ikx ,

for x > a


in which
1p
k= 2mE
h̄
and
1q
l= 2m(E + V0 )
h̄
In the well the wavefunction is the same as in the bound state, but outside the well there
are now running wave solutions travelling left (e−ikx ) and right (eikx ). We need to join
the three regions up by matching value and slope at the boundaries of the well as before.
So at x = −a we have

Ae−ika + Beika = −C sin la + D cos la

ikAe−ika − ikBeika = lC cos la + lD sin la (5a)

and at x = a

C sin la + D cos la = F eika + Ge−ika

lC cos la − lD sin la = ikF eika − ikGe−ika (6a)

Now the nub. We can solve these as simultaneous equations bit by bit as we did for
the bound states by adding and subtracting. But you know that you can usually solve
simultaneous equations with matrices. So inspecting these sets of equations for a moment
you see they are equivalent to these:

e−ika eika A − sin la cos la C

     
−ika ika = (5b)
ike −ike B l cos la l sin la D
    ika −ika
sin la cos la C e e F
  
= (6b)
l cos la −l sin la D ikeika −ike−ika G

Remember the quantum numbers k and l are fixed by the energy E for which we are
seeking a solution. What we want are the coefficients A, B, C and D to plug back into
the wavefunction. Before doing that let’s just look at the physics of this problem. We may
well be asking, suppose I launch a wave in from the left at the well. How much probability
amplitude is transmitted and how much is reflected? Then I will fix A from the outset by
the normalisation of the incoming wave. The amplitude A tells me how much is reflected
PHY3011 Wells and Barriers page 9 of 17

and the amplitude F how much is transmitted. As long as I don’t expect any amplitude
coming in from the right (it’s my experiment after all) then I may take it that G = 0.
The next figure shows such a matched solution for an energy just above zero, E =
0.1V0 ,

Re ψ
Im ψ
amplitude

Anticipating ahead, what I really want to know is the transmission coefficient

 2
F 
T=  2
A

In the meanwhile I may want to know what is the wavefunction in the well, so I’ll want
to know C and D as well.
I will rewrite (5b) and (6b) like this
A C
   
M1 = M2 (5c)
B D
C F
   
M3 = M4 (6c)
D G
I can multiply (5c) by the inverse of M1 and (6c) by the inverse of M3 . Then I substitute
(6c) into (5c) and I get
A F
   
−1 −1
= M1 M2 M3 M4
B G
F
 
=P
G
and P is called the transfer matrix. This is a very general way to deduce the relation
between waves going in and waves coming out of a one dimensional scattering problem.
PHY3011 Wells and Barriers page 10 of 17

There may be more than one well or barrier one after the other and this is the way to
tackle such a problem.
For these purposes we just need to solve for A and B in terms of C and D and for
C and D in terms of F and G and finally to get all the coefficients so we can plot the
wavefunction and extract the transmission coefficient. So firstly I have
A C
   
= M1−1 M2 (7)
B D
It’s useful to remember how to invert a 2 × 2 matrix.
a b 1 d −b
   
if T= then T −1 =
c d ad − bc −c a

(Multiply by one over the determinant, swap the diagonals and change the sign of the off
diagonals.) We’ll get
i −ikeika −eika − sin la cos la
  
M1−1 M2 =
2k −ike−ika e−ika l cos la l sin la
eika (ik sin la − l cos la) −eika (ik cos la + l sin la)
!
i
=
2k e−ika (ik sin la + l cos la) e−ika (−ik cos la + l sin la)

Putting this in (7) you’ll find

i ika  
A= e C (ik sin la − l cos la) − D (ik cos la + l sin la)
2k
i
B = e−ika C (ik sin la + l cos la) + D (−ik cos la + l sin la)
 
2k
Next we do the same thing with (6c). But we’ll now set G = 0 as discussed above, and
to keep the algebra simpler. (If there were amplitude coming in from the right also, then
you can retain G 6= 0.)
1 l sin la cos la eika e−ika
  
M3−1 M4 = ika
l l cos la − sin la ike −ike−ika
1 eika (l sin la + ik cos la) e−ika (l sin la − ik cos la)
!
=
l eika (l cos la − ik sin la) e−ika (l cos la + ik sin la)

and (6c) becomes

k
 
C = F eika sin la + i cos la
l
k
 
D = F eika cos la − i sin la
l
Finally I substitute these into my formulas for A and B to get A and B in terms of F . I
find after quite a bit of algebra
iF  2
l − k 2 sin 2la

B=
2kl
l2 + k 2
!
A = Fe2ika cos 2la − i sin 2la
2kl
PHY3011 Wells and Barriers page 11 of 17

So the transmission coefficient must be given by

 2
−1
A V02 2 2a
 q 
T =  2 =1+ sin 2m(E + V0 ) (8)
F  4E(E + V0 ) h̄

The next figure shows the transmission coefficient plotted versus the energy (remember
this is the energy above the top of the well). If the energy is less than zero there are no
travelling solutions outside the well, the particle is trapped inside although it can tunnel
to the outside as seen in the figure on page 6. Note that the figure below plots the same
function twice over, in different ranges of the energy.
1

0.75
transmission

0.5 square well

0.25

0
0 1 2 3 4 5
E / V0

1
transmission

0.98
0
20 10 30 40 50
E / V0
You notice that the transmission coefficient periodically becomes one as a function of
the energy of the incoming wave. This happens as you see from equation (8) whenever
the sine is zero. That is, when
2a q
2m(E + V0 ) = nπ
h̄
for any integer n. So there is perfect transmission (no reflection) whenever

n2 π 2 h̄2
E + V0 = = En
2m(2a)2
but these are the energy eigenvalues of the infinite square well, measured from the bottom
of the well. This is a quite remarkable fact.
PHY3011 Wells and Barriers page 12 of 17

3. The square barrier

The square barrier rises above the zero of energy having a constant potential V0
between −a < x < a. If we look for energies greater than V0 then the problem is the same
as for the finite well as you can see from the following figure.
E
V0
0
−V0

As long as E > V0 then the problem is the same as for the well having E > 0, except
that the kinetic energy in the barrier is E − V0 rather than E + V0 . So the solution is
exactly as for the well except that now
1q
l= 2m(E − V0 )
h̄
This means that wavefunctions approaching the barrier are reflected and transmitted with
a transmission coefficient given by equation (8) with the sign of V0 changed,

−1
V02 2 2a
 q 
T =1+ sin 2m(E − V0 ) (9)
4E(E − V0 ) h̄

A classical particle would also fly over the barrier, but what if the energy is less than
V0 ? This is the most interesting case; classically a particle would bounce off the barrier,
but in quantum mechanics there’s a finite probability for the particle to “tunnel” through
the barrier and appear on the other side.

Re ψ
Im ψ
amplitude

E
PHY3011 Wells and Barriers page 13 of 17

The figure above shows tunnelling of a particle with energy E = 0.1V0 which you
can compare with the figure on page 9. In either case, the amplitude is constant on the
right as there is only an outgoing wave and this must have a constant intensity, or energy
is not conserved. On the left on the other hand what you see is interference between
the incoming and reflected waves, and the combination has a varying amplitude. In the
tunnelling case note how the wavefunction is decaying in the barrier and what remains is
allowed to “leak out into the vacuum” and propagate.
The transmission coefficient in the tunnelling case, E < V0 , as compared to equa-
tion (9) is given by

−1
V02 2 2a
 q 
T =1+ sinh 2m(V0 − E) (10)
4E(V0 − E) h̄
I’ll now derive this result (you may skip this bit) using the method of transmission ma-
trices. It’s almost the same as the case of the potential well. The wavefunction is

 Aeikx + Be−ikx ,
 for x < −a
ψ(x) = Celx + De−lx , for − a ≤ x ≤ a,
F eikx + Ge−ikx ,

for x > a


The only difference is that inside the well the wavefunctions are decaying—it’s like the
case of states bound within a finite well that decay to the outside. These now decay to
the inside. The slopes are

 ikAeikx − ikBe−ikx ,
 for x < −a
ψ (x) = lCelx − lDe−lx ,
′
for − a ≤ x ≤ a,
ikF eikx − ikGe−ikx ,

for x > a


in which
1p
k= 2mE
h̄
and
1q
l= 2m(V0 − E)
h̄
At x = −a
Ae−ika + Beika = Ce−la + Dela
ikAe−ika − ikBeika = lCe−la − lDela
and at x = a
Cela + De−la = F eika + Ge−ika
lCela − lDe−la = ikF eika − ikGe−ika

e−ika eika A e−la ela C

     
−ika ika = −la
ike −ike B le −lela D
 la
e−la
    ika
e C e e−ika F
 
la −la
= ika −ika
le −le D ike −ike G
By solving these matrix equations in the same way as before you will get these expressions
for A and B in terms of C and D,
i
A = eika −Ce−la (l + ik) + Dela (l − ik)
 
2k
i
B = e−ika Ce−la (l − ik) − Dela (l + ik)
 
2k
PHY3011 Wells and Barriers page 14 of 17

and solving for C and D in terms of F (again we set G = 0)

F ika  −la 
C= e e (l + ik)
2l
F
D = eika ela (l − ik)
 
2l
Then
iF  2(ik−l)a 2 2
(l + ik) + e2(ik+l)a (l − ik)

A= −e
4lk
iF 2ika  −2la  2
l − k 2 + 2ilk + e2la l2 − k 2 − 2ilk
  
= e −e
4lk
iF 2ika  2la
e − e−2la l2 − k 2 − e2la + e−2la 2ilk
    
= e
4lk
l2 − k 2
!
= Fe 2ika cosh 2la + i sinh 2la
2lk
so finally forming |A|2 /|F |2 we get the inverse of the transmission coefficient

−1
V02 2 2a
 q 
T =1+ sinh 2m(V0 − E) (10)
4E(V0 − E) h̄

0.75
transmission

0.5 square barrier

0.25

0
0 1 2 3 4 5
E / V0

1
transmission

0.98
0 10 20 30 40 50
E / V0
The figure above shows equations (9) and (10) which is the transmission coefficient of
the barrier for energies below the barrier (10) left of the dotted line; and above (9) right
of the dotted line.
PHY3011 Wells and Barriers page 15 of 17

Again we see perfect transmission when the sine in (9) is zero, that is

2a q
2m(E − V0 ) = nπ
h̄

which is the same as

n2 π 2 h̄2
E − V0 = = En
2m(2a)2

This happens when the energy measured from the top of the barrier is an eigenenergy of
an infinite square well whose energy zero is set to the top of the barrier. Maybe this is
not too surprising because at those energies the boundary conditions of the infinite well
require that exactly one or one-half a wavelength must fit exactly in the well. This allows
for the wave reflected from the start of the barrier at −a to interfere maximally with the
wave reflected from the end of the well at a when they add up to make the reflected wave
in the region x < −a. If the interference is destructive then there is no reflected wave and
the incoming wave is totally transmitted. This property of reflective coatings is employed
in camera lenses to maximise the light entering the camera, and also in non reflective
coatings on spectacles. We see as we noted in the Introduction and Revision notes that
there is a parallel between interference of the probability amplitude and the interference
of waves in optics.

Re ψ Re ψ
Im ψ Im ψ
amplitude amplitude

The above figures show this effect, in which the energy E has been set to V0 +E1 , left;
and V0 + E2 , right. In each case you can see that the amplitude is completely transmitted
from left to right. You can also see that exactly one-half wavelength is contained in the
barrier in the left hand plot, and one wavelength in the right hand plot; this is consistent
with the boundary conditions on the infinite well for quantum numbers n = 1 and n = 2
as expected.
An example of perfect transmission is the Ramsauer–Townsend effect in which low
energy electrons impinging on a gas of neon or argon is found to have certain energies for
which there is no reflexion.
PHY3011 Wells and Barriers page 16 of 17

4. Some manifestations of tunnelling

Here, I will briefly describe some experiments in which tunnelling is revealed.

Look back at equation (10) for the transmission coefficient of a barrier when the
energy is less than the barrier height. If we define some parameters,
1q
α= 2m(V0 − E)
h̄
4E(V0 − E)
E=
V02
w = 2a
then
1
T −1 = 1 + sinh2 αw
E
We consider the limit that αw ≫ 1 and the limit of the sinh is
1 1
sinh x = (ex − e−x ) → ex
2 2

We then neglect the “one” compared to the sinh2 and get

T = 4E e−2αw
This is the famous “tunnelling law” that says that the fraction of incoming wave that
tunnels through a barrier depends exponentially on the thickness, w, of the barrier. So
to observe tunnelling you need a large incoming energy or flux and a thin barrier. There
is a couple of very important applications of this law.
1. alpha decay. The decay of unstable nuclei by emission of an α-particle (he-
lium nucleus) was understood by Gamow, Condon and Gurney in 1928 as a tunnelling
manifestation. The α-particle is bound to the nucleus by the strong nuclear force which
attracts positive protons more strongly than the repulsive Coulomb force. However the
strong force is very short ranged, acting only over distances of around 10−15 m. So a one
dimensional representation of the potential looks like this.
PHY3011 Wells and Barriers page 17 of 17

The combination of the two potential energies (strong and Coulomb) amount to a
barrier that the α-particle might tunnel through. The higher the energy of the particle
the greater the probability of emission (once it escapes the pull of the strong force it’s
immediately repelled from the positive nucleus by the Coulomb force). In fact the tun-
nelling law predicts that the logarithm of the lifetime is proportional to the inverse square
root of the α-particle’s energy. This is confirmed for a large number of nuclei in the plot
below.

2. The scanning tunnelling mi-

croscope was invented in 1981 by Bin-
nig and Rohrer, who received the No-
bel Prize in physics for their invention
in 1986. You probably know how it
works; here I just want to show a fig-
ure from one of their papers showing
the logarithmic dependence of the tun-
nelling current on the gap between the
tip and the sample, as predicted by the
tunnelling law.