ARDB Technical Note 132 - Draft - 10/27/97

Surface Physics Exercises in Quantum Mechanics

David H. Whittum

In this note we review some amusing quantum mechanics problems relating to surface
phenomena. The selection is a bit eclectic, but this is just a tech note after all.

Outline

Introduction: Copper and Quantum Mechanics

1. Bound-States of a Point Charge Above a Conducting Surface
2. Field Emission from a Plain Vanilla Surface
3. Field-Emission Scaling from the Uncertainty Principle
4. RF Field Emission from a Plain Vanilla Surface
5. More Exercises

Introduction
It is helpful to review a couple of items at the outset. Before discussing a surface, it is good
to consider an example of a solid; we pick copper. In addition, a short review of quantum
mechanics may be helpful.

Copper
For the atomic physicist, Copper is a noble metal, with an atomic number Z=29, atomic
mass A=63.5 amu, 1 the orbital configuration [ Ar ]3d 10 4 s , and ground level 2 S1 / 2 , with ionization
energy 7.478 eV . As far as the solid state physicist is concerned, pure copper forms a face-
centered cubic (fcc) lattice. This means that a perfect crystal of copper may be thought of as
depicted in Fig. 1, when viewed up close, and in the bulk, beyond the surface,

FIGURE 1. Pure, single-crystal copper forms a face-centered cubic lattice.2

Algebraically, an fcc lattice is an array of points consisting of all integer combinations of the
1Recall that 1 amu is one atomic mass unit, defined as 1/12 the mass of the 12C carbon atom, or
1.6605x10-27 kg.
2 N. W. Ashcroft and N. D. Mermin, Solid State Physics, (Holt, Rinehart & Winston,
Philadelphia, 1976)
1
following basis vectors,
a a a
a1 = ( yˆ + zˆ ), a2 = ( xˆ + zˆ ) , a3 = ( xˆ + yˆ ) ,
2 2 2
and for copper, the dimension a=3.61Å.
In a pure crystal of copper the density of conduction-band electrons is
ne = 8.47 × 10 22 cm −3 , just one "free" electron per copper atom. The remaining 28 electrons per
atom are bound, or "valence" electrons. If we convert the density of free-electrons to a volume per
electon, then the radius of a sphere having the same volume is rs=1.41Å, at room temperature, and
atmospheric pressure. The Fermi energy is ε F = 7.00 eV, the Fermi temperature is
TF = 8.16 × 10 4 °K , the Fermi wavenumber is kF = 1.36 × 108 cm −1 and the Fermi velocity is
VF = 1.57 × 108 cm/s . The work function is, depending on the crystal plane, 4.59 eV (100), 4.48
3
eV (110), 4.98 eV (111), 4.53 eV (112).
Let us recall the particulars concerning the concept of Fermi energy. One can find these
considerations in a solid-state text (my favorite is Ashcroft and Mermin). Recall that electrons are
fermions, governed by the Pauli exclusion principle. If an ensemble of electrons is confined to a
system at zero temperature, they will fill available states, starting with the lowest energy state, 2
electrons per state (corresponding to the two possible values for spin), until all N electrons are
accomodated. The energy of the highest energy filled state is then called the Fermi energy ε F , and
it may be related to the Fermi wavenumber kF according to

h 2 kF2
εF = .
2m
In depicting electron states in a solid, we look into the bulk of the solid, and consider a cubic
volume of length L, and volume V=L3. Momentum eigenstates are separated by ∆k = 2π / L and
thus the number of eigenstates in a k-space volume Ω is

Ω 1
Nk = = ΩV .
∆k 3
8π 3

Let us fill these states, 2 electrons per state, until the highest wavenumber, kF , is reached. Then we
are filling a k-space volume

4π 3
Ω= kF ,
3
and we have N electrons, so that

1 4π 3 kF3
N = 2 Nk F =2× 3V kF = 2 V .
8π 3 3π

Evidently Fermi wavenumber may be related to the density of free electrons,

3
Physics Vade Mecum, A Physicist’s Desk Reference, Herbert L. Anderson, ed., (American
Institute of Physics, New York, 1989) pp. 306-317 and pp. 318-335.
2
 N k3
ne = = F2 .
V 3π
Let us also notice that the Fermi energy is referred to the ground state. Thus if the ground
state is separated from the potential at infinity (i.e., well beyond the surface of the solid) by an
amount eV0 < 0 , then the highest energy level referred to that at infinity is,

ε F + eV0 = −eφ ,

where eφ is the work function, the amount of energy required to liberate an electron. For copper,
taking a value eφ ≈ 4.65 eV as representative, and using ε F ≈ 7.00 eV, one has V0 ≈ −11.65 V.

Quantum Mechanics
Recall that in the quantum mechanical description of a one-electron system, and neglecting
r
spin, we describe the state of the system with a complex wavefunction Ψ(r , t ) , a function of space
and time coordinates. The evolution of the system is governed by the Hamiltonian operator H ,

∂Ψ
HΨ = ih ,
∂t
and we may express H as

h2 r 2 r
H=− ∇ + V (r ),
2m
with V the potential. The quantity

ρ P = ΨΨ∗ ,

may be interpreted as a probability density. Associated with this is a law of conservation of

probability, taking the form,
r r ∂ρ
∇ • J P + P = 0,
∂t

with
r r r
JP =
h
2m i
{
Ψ ∗∇Ψ − Ψ∇Ψ ∗ . }
r
It is helpful in analyzing the behavior of such systems, to determine the energy eigenstates ψ E (r )
such that
r r
Ψ(r , t ) = ψ E (r )e − iEt / h ,

is a solution, i.e.,

3
  h2 r 2 r
Hψ E = − ∇ + V (r )ψ E = Eψ E .
 2m 
A general solution may then be represented as a superposition of energy eigenstates,
r r
Ψ(r , t ) = ∑ cEψ E (r )e − iE / ht .
E

Let us turn next to consider some amusing problems at a conducting surface.

1. Bound-States of a Point Charge Above a Conducting Surface

A recent article4 notes that the bound states of a point charge above a conducting plane form
a Rydberg series. A hand-waving explanation of this observation would consult the picture of an
image charge and discuss the two-body problem, with constrained center of mass coordinate. In
this note we take a more pedantic approach to check the result quoted in the article.
The image potential of an electron above a conducting plane takes the form5

e2
V (z) = − , (z>0)
4z
with z the distance above the plane. We enumerate bound states of energy E. Schrodinger’s
equation is

 h2 2 
− ∇ + V ( z ) Ψ = EΨ ,
 2m 

and the problem is separable due to invariance under translations parallel to the surface,
r r r
Ψ(r⊥ , z ) = ψ ( z )eik ⊥ • r⊥ .
r r
The coordinates in the plane are r⊥ = xxˆ + yyˆ , and the transverse momentum in this state is hk⊥ .
We have
r
 h2 d 2   h 2 k⊥2 
− + V ( z ) ψ =  E − ψ,
 2 m dz 2   2 m 

or, in normalized coordinates,

 d2 1 2
 2 + − q ψ = 0 .
 dξ ξ 

The notation is

4
U. Hofer, et al., “Time-Resolved Coherent Photoelectron Spectroscopy of Quantized Electronic
States on Metal Surfaces”, Science 277 (1997) pp. 1480-1482.
5
J. D. Jackson, Classical Electrodynamics (John Wiley & Sons, New York, 1975) Ch. 2.
4
 z
ξ= ,
z0

with z0 twice the Bohr radius,

h2 2 °
z0 = 2 = r ≈ 1. 1 A.
me 2 α 2
e

Here

e2 1
α= ≈ ,
hc 137
is the fine structure constant, and

e2
re = 2
≈ 2.82 × 10 −13 cm ,
mc
is the classical electron radius. We have adopted the notation
r
h 2 k⊥2 1 2 2 2
E= − α mc q ,
2m 8
defining the variable q.
Let us solve this equation for positive ξ . For ξ >> 1/ q our system takes the approximate
form

 d2 2
 2 − q ψ ≈ 0 ,
 dξ 

and solutions finite at large ξ take the form ψ ∝ e − qξ . (Notice that solutions with imaginary q are
possible; these correspond to unbounded motion). This suggests looking for a solution in the form

ψ = f (ξ )e − qξ .

Then

ψ ′ = f ′e − qξ − qfe − qξ , ψ ′′ = f ′′e − qξ − 2 qfe − qξ + q 2 fe − qξ ,

and Schrodinger’s equation reduces to

d2 f df f
− 2q + = 0.
dξ 2
dξ ξ

We look for a series solution,

5
 ∞
f (ξ ) = ∑ ck ξ k ,
0

and find c0 = 0 , and

2 qk − 1
ck +1 = ck .
k (k + 1)

Notice that the series truncates if 2qk=1 for some k. On the other hand, if the series does not
truncate, then, for large k we have

2q
ck +1 ≈ c,
(k + 1) k
and in this case the solution for f differs by at most a finite polynomial from the function defined
by Taylor coefficients (for some constant A),

ck ≈A
( 2q)
k

.
k!

This latter series is summable and gives f ≈ Ae 2 qz , corresponding to a divergent solution. We

conclude that a finite solution requires truncation of the series, and therefore a discrete spectrum.
We denote

1
qn = ,
2n
and the corresponding eigenfunction,
n
ψn = e − qnξ
∑c ξ ,
1
n
k
k

with

k / n −1 n
ckn+1 = ck .
k (k + 1)

The corresponding spectrum is

r
r
( )
En k⊥ =
h 2 k⊥2 1 2 2 1
2 m 32
− α mc 2 .
n
The constant

1 2 2
α mc ≈ 0.85 eV ,
32

6
is 1/16th of the Rydberg energy,

1
R = α 2 mc 2 ≈ 13.6 eV.
2
It will be convenient for later work to abbreviate

1 1
ε0 = R = α 2 mc 2 ≈ 3.40 eV.
4 8
For perturbations respecting translational invariance parallel to the surface, and where
symmetry permits, one expects to uncover the equivalent of the Lyman series ( n → 1),
0.64 eV (1.9 µm -"α ") , 0.76 eV (1.6 µm -" β ") , etc., converging to 0.85 eV (1.5 µm ) . The
Balmer series ( n → 2) corresponds to 0.12 eV (10.5 µm -"α ") , 0.16 eV (7.8 µm -" β ") , etc.,
converging to 0.21 eV (5.8 µm ) . The Paschen series ( n → 3), corresponds to 41 meV (30 µm ) ,
61 meV (21 µm ), converging to 95 meV (13 µm ) . In converting to eV we made use of

re 2π 1.24 eV
hω = mc 2 = ,
α λ λ ( µm )

to relate the transition energy to wavelength of the emitted photon, as for single-photon emission.
Note that room temperature corresponds to

1
kBT = kB × 300°K ≈ eV ≈ 26 meV .
38.7

It would be amusing to revisit this problem starting from the Dirac equation. Also, the
treatment of z<0 is a bit fuzzy. To smoothly connect the potential outside to that inside, we should
pay attention to the constellation of charges in the vicinity of the surface.

2. Field-Emission Scaling from the Uncertainty Principle

A fundamental problem for high-gradient linacs is field-emission, the extraction of
electrons from a conducting surface in the presence of a large field. Let us review here the problem
first considered by Fowler and Nordheim. 6 We consider a conducting plane, to which is applied a
DC electric field E0. This field lowers the potential beyond the surface, making it possible for
electrons to tunnel through and be emitted. Let us determine the conditions favorable to emission
with a scaling argument.
In the course of the tunneling process, the uncertainty, ∆p, in the electron momentum is
related to the potential according to

eφ ≈
( ∆p )
2

⇒ ∆p ≈ (2 meφ ) .
1/ 2

2m
From Heisenberg’s principle, the uncertainty in position is

6
R. H. Fowler and L. W. Nordheim, Proc. Roy. Soc (London) A 119 (1928) 173.
7
 h h
∆x ≈ ≈ 1/ 2 .
2 ∆p 2(2 meφ )

This corresponds to an uncertainty in energy on the order of

∆ε ≈ eE∆x .

When this uncertainty is on the order of the work function, ∆ε ≈ eφ , one expects to find electrons
being emitted. In this case

2(2 m) (eφ )
3/ 2
h
1/ 2

eφ ≈ eE∆x ≈ eE ⇒ ≈ 1.
2(2 meφ ) eEh
1/ 2

We may express this in normalized form using

mc 2 1 1 2 2 eφ eφ
E= (α / 2)4 µ , ε0 = α mc , =8 2 2.
e re 16 ε0 α mc

In terms of normalized variables, our condition is, after some algebra,

3/ 2
2  eφ 
  ≈ 1.
µ  ε0 

In practical units, using µ ≈ 3.11 × 10 −2 E0 (GV/m ) , and eφ / ε 0 ≈ 4.65 / 3.40 ≈ 1.37 (for copper),
we may write this as E0 ≈ 100 GV/m. At this field level one would expect the orderly bound states
characteristic of the solid to lose their integrity.
In a similar manner we can estimate the tunneling time,

(2 meφ )
1/ 2
α m c  eφ 
1/ 2
h h
∆t ≈ ≈ ≈ ≈   .
2 ∆ε 2eE∆x eE 2eE  ε 0 

For copper, with eφ / ε 0 ≈ 1.37 , this gives ∆t ≈ 10 −14 s / E(GeV/m ) , quite a short time scale, when
the field is large.

3. Field Emission from a Plain Vanilla Surface

Let us consider next a more quantitative approach to field emission, being interested in the
scaling for the emitted current density as a function of applied field. Continuing with the notation
of the Sec. 1, Schrodinger’s equation may be cast in the form, for ξ > 0,

 d2 1 
 2 + − q + µξ  ψ = 0 , ( ξ > 0)
2

 dξ ξ 

We will consider the simplest of surfaces in this section, and discard the image potential. In this
equation,

8
 eE0 re
4 ≈ 3.11 × 10 E0 ( GeV/m ) .
−2
µ=
mc (α / 2)
2

In the first approximation, we are interested in the limit µ << 1. The quantity q 2 is related to the
energy of this eigenstate according to
r
h 2 k⊥2 1 2 2 2
E= − α mc q .
2m 8

For ξ < 0, we consider a propagating solution, our picture of a conduction band electron,

 d2 2
 2 + Q ψ = 0 .
 dξ 

The quantity Q2 is related to the potential within the surface, V0 < 0 , according to
r
h 2 k⊥2 1 2 2 2
E − V0 = + α mc Q .
2m 8
This is to say that

−V0
Q2 + q 2 = = ρ2 .
R/ 4

Solutions for ξ < 0 are just plane waves,

ψ = A+ eiQξ + A− e − iQξ .

Let us solve Schrodinger’s equation in the region ξ > 0,

 d2  q 2 
 dξ 2 +  − µ   ψ = 0 .
µ ξ
  

We make the change of variables,

 q2 
η = µ1 / 3  − ξ  ,
µ 
so that

 d2 
 2 − η ψ = 0 .
 dη 

9
This we recognize as Airy’s equation, with solutions Ai and Bi. Asymptotic forms are7

1
Ai(η > 0) → π −1 / 2η −1 / 4 e −ς , Bi(η > 0) → π −1 / 2η −1 / 4 eς ,
2
π π
Ai( −η < 0) → π −1 / 2η −1 / 4 sin ς +  , Bi( −η < 0) → π −1 / 2η −1 / 4 cos ς +  ,
 4  4
1
Ai ′(η > 0) → − π −1 / 2η1 / 4 e −ς , Bi ′(η > 0) → π −1 / 2η1 / 4eς ,
2
π π
Ai′( −η < 0) → −π −1 / 2η1 / 4 cos ς +  , Bi′( −η < 0) → π −1 / 2η1 / 4 sin ς +  ,
 4  4
and we abbreviate

2
ς = η3 / 2 .
3

Our interest is to match to an outgoing solution in the region ξ > 0, and this corresponds to

ψ (ξ ) = Bi[η(ξ )] + iAi[η(ξ )] . ( ξ > 0)

We match the solutions at

q2
η0 = η(ξ = 0) = .
µ2 / 3

Matching conditions are

ψ (0 − ) = A+ + A− = ψ (0 + ) = Bi(η0 ) + iAi(η0 ) ,
dψ − dψ − dη
dξ
( 0 ) = iQ( A+ − A− ) =
dξ
( 0 )=
dξ
( Bi ′(η0 ) + iAi ′(η0 )) = − µ 1 / 3 ( Bi ′(η0 ) + iAi ′(η0 )) .

The solution is

1 iµ 1 / 3  i iµ 1 / 3 
A+ =  Bi(η0 ) + Bi ′(η0 ) +  Ai(η0 ) + Ai ′(η0 ),
2 Q  2 Q 
1 iµ 1/ 3
 i iµ 1/ 3

A− =  Bi(η0 ) − Bi ′(η0 ) +  Ai(η0 ) − Ai ′(η0 ) .
2 Q  2 Q 

In the limit of not too large a gradient, we have η0 >> 1, and may evaluate these
coefficients using the asymptotic forms

1  −1 / 2 −1 / 4 ς iµ 1 / 3 −1 / 2 1 / 4 ς  η −1 / 4 eς  iµ 1 / 3η1 / 2 
A+ ≈ π η e + π η e ≈ 1 + ,
 2π
1/ 2
2 Q  Q 
7
M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions (Dover, New York,
1972), p.446.
10
 η −1 / 4 eς  iµ1 / 3η1 / 2 
A− ≈ 1 − .
2π 1 / 2  Q 

For very large gradients, the coefficients must be evaluated numerically. Routines for this purpose
may be found in Numerical Recipes.8
With the solution in hand, we may derive a figure for the net flux far to the right of our
surface, and compare that to the incident flux, to derive a transmission probability, T. Probability
flux, evaluated to the left of the surface, is

h ∗ ∂
JP z =
2m i
(
A+ eiQξ + A− e − iQξ ) (
∂z
A+ eiQξ + A− e − iQξ + c.c.)
h
( ) (iQ)( A e )
∗
= A+ eiQξ + A− e − iQξ +
iQξ
− A− e − iQξ + c.c.
2 m iz0
hQ hQ
= ( A+ + A− )( A+∗ − A−∗ ) + c.c. = ( A+ A+∗ + A− A+∗ − A−∗ A+ + A− A−∗ ) + c.c.
2 m z0 2 m z0
hQ
=
m z0
( A+ A+∗ − A−∗ A− )
Probability flux, evaluated far to the right of the surface, is

h
( [ ] [ ]) ∂∂z ( Bi[η(ξ )] + iAi[η(ξ )]) + c.c.
∗
JP z = Bi η(ξ ) + iAi η(ξ )
2m i
h ∗ µ 
1/ 3
= ( Bi(η) + iAi(η))   ( Bi ′(η) + iAi ′(η)) + c.c.
2m i  z0 
hµ 1 / 3
( Bi(η) + iAi(η)) ( Bi ′(η) + iAi ′(η)) + c.c.
∗
≈
2 m iz0
∗
hµ 1 / 3  −1 / 2 −1 / 4 π   π 
≈ π η exp − i ς +    iπ −1 / 2η1 / 4 exp − i ς +   + c.c.
2 m iz0   4    4 
hµ 1 / 3
≈
m z0 π

The transmission probability is then

hµ 1 / 3
J (ξ > 0) m z0π µ1 / 3
T = Pz = = .
J P z + (ξ < 0)  hQ  ∗ Qπ A+ A+∗
  A+ A+
 m z0 

In the limit of not too large a gradient ( η0 >> 1), this is

8
W.H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes, The
Art of Scientific Computing, (Cambridge University Press, Cambridge, 1992).
11
 4e −2ς
T≈ .
 Q µ 1 / 3η01 / 2 
 1/ 3 1/ 2 + 
 µ η0 Q 

This may be made more explicit using η01 / 2 = q / µ 1 / 3 ,

4  4 q3 
T≈ exp − .
Q q  3 µ
 + 
 q Q

The kinematic dependence may be made more explicit,

4q ρ 2 − q 2  4 q 3  4Q ρ 2 − Q 2  4 ( ρ 2 − Q2 )3 / 2 
T≈ exp − = exp − .
ρ2  3 µ ρ2  3 µ 

Next we would like to convolve this probability, with the incident flux corresponding to the
spectrum of momenta characteristic of conduction band electrons in copper. For copper at zero
temperature, this distribution is simple; the distribution in momenta,
r r Q
k = k⊥ + zˆ
z0
r
is uniform over a sphere in k -space, with radius, kF . Thus for example, the maximum Q is given
by,

Qm = kF z0 ≈ 1.36 × 108 cm −1 × 1.06 × 10 −8 cm ≈ 1.44 ,

where we evaluate this for copper. The quantity,

−V0 11.65 eV
ρ= ≈ ≈ 1.85 ,
ε0 3.4 eV

(with slightly different values depending on the choice of crystal plane), and we employed
ε 0 = R / 4 ≈ 3.4 eV.
r
r r To r simplify our notation, let us define a normalized coordinate in k -space,
Q = kz0 = k⊥ z0 + zQ
ˆ . The probability flux in z associated with a momentum state is

hQ
J Pz = ,
m z0

and the quantity with the classical analog of current density is then

12
 r r hQ
jz = −ene
1
4πQm3 / 3 ∫ d 3
Q T Q
m z0
( )
Q >0

( )
,
r 4Q 2 ρ 2 − Q 2  4 ρ 2 − Q2 3/ 2

3 h  
= −ene ∫ exp −
3
dQ 
4π Qm m z0
3
Q >0
ρ 2
 3 µ 

where the integral is over the half-sphere of radius Q m , in Q-space. Employing spherical
coordinates, this may be expressed as

3 h
Qm 1
4Qr2 χ 2 ρ 2 − Qr2 χ 2  4 ( ρ 2 − Qr2 χ 2 )3 / 2 
jz = −ene ∫0 Q dQr ∫0 d χ exp −
2
.
ρ2 µ
r
2 Qm3 m z0  3 

Let us introduce

ρˆ = ρ / Qm , µˆ = µ / Qm3 ,

and observe the relation to work-function. Fermi energy is just the energy of the most energetic
occupied state at zero-temperature, referred to the ground-state,
1
ε F = α 2 mc 2Qm2 = ε 0Qm2 = EF − V0 = −eφ − V0 ,
8
where E F is the energy of the most energetic occupied state, referred to a potential of zero at
infinity, beyond the surface. This relation may be expressed as

εF −eφ
= Qm2 = + ρ2.
ε0 ε0
Thus
2
 ρ eφ 1 eφ
ρ̂ − 1 =   − 1 =
2
= .
 Qm  ε 0 Qm ε F
2

Our integral may be expressed as

1 1  4 ( ρˆ 2 − r 2 χ 2 )3 / 2 
jz = j0 ∫ dr ∫ d χ r χ 4 2
ρˆ − r χ exp −
2 2 2
,
0 0  3 µˆ 

with

6hQm
j0 = −ene .
m z0 ρ̂ 2

Let us make this expression a bit more explicit. The number density is

13
 3/ 2
1 α 6  εF 
3/ 2
ne = F2 = 2  2 ε F 
k3 1 2m
=   .
3π 3π  h  24π 2 re3  ε 0 

Other terms are

1/ 2 −1
ε  1  eφ  h 1
Qm =  F  , 2 = 1 +  , = α.
 ε0  ρ̂  εF  m c z0 2

Thus
3/ 2 1/ 2 −1
1 α 6  εF  ε   eφ  1
j0 = −ec ×   ×6× F 1 +  × α
24π 2 re3  ε 0   ε0   εF  2
2 −1 2 −1
α 7 ec  ε F   eφ   ε   eφ 
= − 2 3   1 +  ≈ 3 × 10  F  1 +  A/cm 2
12

8π re  ε 0   εF   ε0   εF 

For copper, eφ ≈ 4.65 eV , ε F ≈ 7.00 eV, and using ε 0 ≈ R / 4 ≈ 3.40 eV , we have,

ε F 7.00 eφ 4.65 eφ
≈ ≈ 2.06 , ≈ ≈ 1.37 , ≈ 0.665 .
ε 0 3.40 ε 0 3.40 εF

In addition, making use of only the very latest values for the fundamental constants,9

α ≈ 1 / 137.0359895 (61) , me c 2 ≈ 0.510 999 06(15) MeV ,

e ≈ 1.60217733( 49) × 10 −19 C, c ≈ 2.99 792 458 × 108 m / s,
re ≈ 2.817940 92(38) × 10 −15 m, π ≈ 3.141592 65359K,

one finds

α 7 ec
≈ 7.49 × 1011 A/cm 2 ,
32π re
2 3

so that j0 ≈ 7.6 × 1012 A/cm 2 .

It is straightforward to evalute this integral numerically. We are interested in parameter
values, ρˆ ≈ 1.85 / 1.44 ≈ 1.28, and in gradients from 1 GV/m to perhaps as large as 10 GV/m or
µ ≈ 3.11 × 10 −2 − 3.11 × 10 −1 . This corresponds to µˆ ≈ 1.0 × 10 −2 − 1.0 × 10 −1. With such values of
µ̂ , an analytic estimate of the integral is worthwhile. With a change of variables,
r ≈ 1 − ε r , χ ≈ 1 − ε χ , and a small argument approximation,

9
“Review of Particle Physics”, Particle Data Group, Phys. Rev. D, 54 (1996) 65.
14
 1 1  4 ( ρˆ 2 − r 2 χ 2 )3 / 2 
jz ≈ j0 ρˆ − 1 ∫
2
dr ∫ d χ exp − 
0 0  3 µˆ 
 4 ( ρˆ 2 − 1)3 / 2  1  4 2
).
1

≈ j0 ρˆ − 1 exp −
2

µˆ
 ∫ dε r ∫ dε χ exp − ( ρˆ − 1) ε r + ε χ
1/ 2
(

3
 0 0  µˆ 

We approximate the exponential integrals with

∞
 4 1/ 2   4 1/ 2  µˆ
1

∫ dε exp − ( ρˆ 2 − 1) ε  ≈ ∫ dε exp − ( ρˆ 2 − 1) ε  ≈ 1/ 2 .
 µˆ  0  µˆ  4( ρˆ − 1)
2
0

We then obtain
 4 ( ρˆ 2 − 1)3 / 2   4 ( ρˆ 2 − 1)3 / 2 
2
µˆ 
(ρˆ − 1) µˆ exp− 3 µˆ .
1 2 −1 / 2 2
jz ≈ j0 ρˆ − 1 exp −
2
 1 / 2  ≈ j0
 µˆ  4( ρ − 1)   
3 2
ˆ 16

This may be expressed as

1 ε 
1/ 2
µ2  4  eφ  3 / 2 Q3  1/ 2
1  ε0   ε0 
5/ 2
 4  eφ  3 / 2 
jz ≈ j0  F  exp −    ≈ j0     µ exp −
2
  ,
m
16  eφ  Qm6  3  ε F  µ  4  eφ   ε F   3µ  ε 0  

or
−1
α 7 ec  eφ   ε 0 
1/ 2
 ε0 
1/ 2
 4  eφ  3 / 2 
jz ≈ 1 +      µ exp−    .
2

128π 2 re3  ε F   eφ   εF   3µ  ε 0  

In practical units this is

−6
jz ≈ 6.15 × 10 A/cm 2 (ε F / eφ )1 / 2 E 2 exp−6.838 × 10 7 (eφ )3 / 2  ,
(eφ + ε F ) 0   (eV and V/cm units)
E0 
with work function and Fermi energy in units of electron-volts, and electric field in units of
volts/cm. [The expression given by Gomer 10 has the coefficient 6.2 × 10 6 A/cm 2 , still tracking
down the source of the discrepancy.] In this, we made use of µ ≈ 3.11 × 10 −9 E0 (V/cm ) .
For copper we have

10
Robert Gomer, Field Emission and Field Ionization, (Harvard University Press, Cambridge,
1961)
15
 1/ 2 5/2
 4 3/ 2 
jz ≈ j0 
1 1   1  µ 2 exp − (1.37) 
16 1.37 
  2.06   3µ 
 2.14   2.14 
≈ 0.035 j0 µ 2 exp −  ≈ 2.8 × 10 A/cm × µ exp −
11 2 2

 µ   µ 
 68.8 
≈ 2.5 × 108 A/cm 2 × E 2 (GeV/m ) exp − .
 E(GeV/m ) 

Thus at 4 GV/m jz ≈ 125A/cm 2 . For large enough current density, one may become concerned
with the cold surface approximation.

4. RF Field Emission from a Plain Vanilla Surface

The typical situation of interest for high fields (>100 MeV/m), corresponds to a field
varying sinusoidally in time at microwave frequencies. At such frequencies and fields electrons are
interacting with a large number of photons, and photon energies are small compared to the energy
level spacings, and we may continue to treat the electric field classically. In this section, we derive
the amended scaling for an rf field, with period long compared to the tunneling time. As we’ve
seen, tunneling time is short ∆t ≈ 10 −14 s / E(GeV/m ) , while even at 100 GHz, the rf period (10
ps) is relatively long in comparison.
Consider a function

f ( x ) = x 2 e −1 / x ,

with x = A sin θ and let us average over one-half cycle in angle θ, assuming A << 1. We expand

1 1 1 2
≈ + ε ,
x A 2A
where

π
θ= +ε.
2
We have
π +∞
 ε 2  1 2 −1 / A  ε2   2
1/ 2
1
( )
π ∫0 ∫−∞  − 2 A  ≈  π 
2 −1 / A
f ≈ A sin θ e exp  −  ≈ A e exp A2.5e −1 / A .
 2 A π

This implies that our DC emission law,

j ≈ b E 2 exp −  ,
a
 E
takes the form

16
 1/ 2
 1 
b E 2.5 exp −  ,
a
j≈ 
 2π a   E

when expressed in terms of rf amplitude. An additional factor of 1/2 appears here, since no
emission occurs during one-half cycle of the rf. The net correction is essentially a “duty-cycle”
correction, j ≈ ηrf j peak , amounting to the fraction of time during which the field is sufficiently near
the peak to extract appreciable current,
1/ 2
 E 
ηrf =   .
 2π a 

The approximation made in expanding the term in the exponent amounts to the assumption ηrf << 1
. To compute the time-averaged current density for a pulsed rf signal, one would need an additional
duty cycle correction. For example, for a 1 µs pulse length, and a 60 Hz pulse repetition
frequency, the additional duty cycle correction is 6 × 10 −5 .
More explicitly, the average current density, within the rf pulse then takes the form
−1
31 / 2 α 7 ec  eφ   ε 
5/ 4
 ε0 
1/ 2
 4  eφ  3 / 2 
j ≈ 17 / 2 5 / 2 3 1 +   0    µ 5/2
exp −   .
2 π re  ε F   eφ   εF   3µ  ε 0  

In practical units,

 68.8 
j ≈ 1 × 10 7 A/cm 2 × E 2.5 (GeV/m ) exp − .
 E(GeV/m ) 

Thus at 4 GV/m, j ≈ 10A/cm 2 . Wang and Loew quote a coefficient that equates to
1.6 × 108 A/cm 2 .11 [track down discrepancy, their result is closer than Gomer’s...]

5. More Exercises
There are many more exercises one could cram in here: a general WKB treatment of
tunneling, incoporation of the image potential, a dielectric inclusion layer, space-charge and
thermal effects. Meanwhile, time is up for this note, and we must turn to other things...

11
G. A Loew and J. W. Wang, “RF Breakdown Studies in Room Temperature Electron Linac
Structures”, XIIIth Internat’l Symposium on Discharges and Electrical Insulation in Vacuum,
(1988) SLAC-PUB-4647.
17