Lecture 4

M. Siddikov

Marat.Siddikov@usm.cl

April 11, 2023

Outline

Today’s plan:
Electrostatic fields
−Electrostatic fields in medium. Border
conditions
−Numerical methods
Magnetostatic fields
−Law of Biot-Savart
−Law of Ampere
−Vector and scalar potentials
Electrostatics in brief

Electrostatics Border conditions

(1) (2)
ρ E|| = E||
∇·E = ,
εε0 (1) (2)
D⊥ = D⊥ ⇔ ε(1)
(1)
(2) (2)
r E⊥ = εr E⊥
∇×E =0

⃗ = εr ε0 E
⋆“electric displacement field” D ⃗, Caution
so Note that in general ∇ × D ⃗ ̸= 0 near the
border of two dielectrics (εr ̸= const).
∇ · D = ρ,
For this reason we define potential ϕ as
∇×E =0 E = −∇ϕ, NOT as D ⃗ = −∇ϕ

⋆χ = εr − 1 - electric susceptibility
⋆P⃗ = ε0 χ E
⃗ -polarization density (dipole
moment per volume)
Electrostatics inside dielectrics

Problem is relatively simple. Please

let me know the final result.
The pointlike charge Q is placed in the
center of a dielectric sphere of radius R.
The relative permittivity of the material
of the sphere is ε. Find the electric field
inside and outside the sphere.
Electrostatics inside dielectrics
Due to spherical symmerty, the field Let’s construct hypothetical surface
E ∼ r̂ (is radially symmertic), and ϕ = made of two concentric sphere of in-
ϕ(r ). ternal radius R + 0 and external radius
Let’s construct hypothetical sphere r > R.
of radius r < R. Applying the Gauss  Since there is no charges inside this
theorem, can get for the electric field hypothetical surface, from Gauss the-
on the surface orem expect full flux of electric field
(through internal and external spheres)
Q
E (r ) = r̂ , r ≤R is zero.
4πε0 εr r 2 ⇒Evaluating fluxes, can get for the
At the surface r = R we need to electric field on the surface
apply the border conditions (continu- Q
ity for Dr = εr ε0 Er ), so E (r ) = r̂ . r > R
4πε0 r 2
Q -thus we can see that field outside the
E (r ) = r̂ , r ≤ R
4πε0 εr r 2 spherical dielectric is NOT modified,
Q same as for pointlike charge in vacuum.
E (r + 0) = 
εr r̂ , r ≈ R + 0
4πε0εr R 2
Problem 2
Please explain, how we should solve the
problem ?
A hollow cylindrical tube with internal
radius a and external radius b is made of
a dielectric with relative permittivity ϵ.
The tube is placed into a homogeneous
external electric field E 0 which is
perpendicular to its symmetry axis.
Evaluate the electric field E (r ) in the
system (both inside and outside).
Problem 2
Have 3 independent regions with differ-
ent ε and border conditions:
A hollow cylindrical tube with internal Region A: r < a: no singularity, ε = ε0
radius a and external radius b is made of Region B: r ∈ (a, b): no singularities,
a dielectric with relative permittivity ϵ. ε = ϵ ε0
The tube is placed into a homogeneous Region C: r > b: no singularities, ε =
external electric field E 0 which is ε0 . Additional constraint:
perpendicular to its symmetry axis.
Evaluate the electric field E (r ) in the lim E (r ) == E0 x̂
r →∞
system (both inside and outside).
= E0 (r̂ cos φ − φ̂ sin φ)
h i
= E0 Re (r̂ + i φ̂) e iφ

 In this problem we don’t have concen-

trated charges, so should solve Laplace
equation in all 3 regions, though with
proper value of permittivity ε.
Problem 2
 We found in Lecture 1 a general Border conditions:
z-independent solution (in polar coordi- Region A : no singularities at r → 0⇒ all
nates) in theform 
(A)
βm = 0
X βm Region C : expected behaviour at large
ϕ= αm r m + m e imφ ,
m
r r → ∞ implies that
Er = −∂r ϕ = (C ) (C )
αm̸=1 = 0, α1 = −E0
X  βm

 Border of regions AB and BC :
=− m αm r m−1 − m+1 e imφ , (A) (B)
r Er (a) = ϵEr (a)
m (B) (C )
∂φ ϕ ϵEr (b) = Er (b)
Eφ = − = (A) (B)
Eφ (a) = Eφ (a),
r  (B) (C )
X βm
  Eφ (b) = Eφ (b)
=− im αm r m−1 + m+1 e imφ , We don’t impose any conditions on po-
m
r
tentials
 αm , βm are complex numbers, and Since each harmonic e imφ is indepen-
Re(...) is always implied. dent, should impose border conditions in-
could rewrite this result using two inde- dependently on each term in the sum
pendent sums including both sin(mφ) and
cos(mφ) or sin(mφ + δm ).
 αm , βm are different in each region, so
use superscript indices (A), (B), (C ) to
distinguish
Problem 2
Since each harmonic e imφ is independent, should impose border conditions
independently on each term in the sum
!
(B)
(A) m−1 (B) m−1 βm
αm a = ϵ αm a − m+1
a
! !
(B) (C )
(B) m−1 βm βm
ϵ αm b − m+1 = −E0 δm,1 b m−1 − m+1
b b

!
(B)
(A) m−1 (B) m−1 βm
αm a = αm a + m+1
a
! !
(B) (C )
(B) m−1 βm m−1 βm
αm b + m+1 = −E0 δm,1 b + m+1
b b

For m ̸= 1 homogeneous linear algebraic system, trivial solution αm = βm =

cm = 0
Problem 2
For m = 1 system is inhomogeneous, Solutions:
(A) 4b 2 E0 ϵ
(B)
! α1 = ,
(A) (B) β1 a2 (ϵ − 1)2 − b 2 (ϵ + 1)2
α1 = ϵ α1 −
a2 (B) 2b 2 E0 (ϵ + 1)
α1 = ,
(B)
!
(C )
! a2 (ϵ − 1)2 − b 2 (ϵ + 1)2
(B) β1 β1
ϵ α1 − = −E0 − (B) 2a2 b 2 E0 (ϵ − 1)
b2 b2 β1 = ,
a2 (ϵ
− 1)2 − b 2 (ϵ + 1)2
b 2 a 2 − b 2 E0 ϵ 2 − 1



(C )
β1 = 2 ,
!
(B)
(A) (B) β a (ϵ − 1)2 − b 2 (ϵ + 1)2
α1 = α1 + 12
a
! ! Some special cases:
(B) (C ) (A) (B) (B)
(B) βm βm ϵ → 1: α1 ≈ α1 ≈ −E0 , β1 ∼
α1 + 2 = −E0 + 2 (C )
b b β1 ∼ O (ϵ − 1)
-field at r < a almost the same as outside
ϵ → ∞:

(A) (B) (B) −b 2 E0

α1 ∼ α1 ∼ β1 ∼ ,
(a2 − b 2 ) ϵ
(C )
β1 ∼ b 2 E0

-field at r < a suppressed by 1/ϵ.

Electrostatics inside dielectrics

Simple analytical solutions like in previous problems are possible only for very simple
geometries.
In realistic problems we end up with complicated geometries which cannot be solved
analytically
Now we’ll discuss some numerical methods which allow to construct numerical so-
lutions for arbitrary geometry.
Electrostatics - Numerical methods
(we’ll discuss very basic introduction from Applyˆ Gauss theorem:
ˆ
Sec. 2.12 of Jackson; actually it is a very 1
dV ρψi = ∇ψi · ∇ϕ dV −
large topic) ε0
˛
Different methods, we’ll see just one of
− dS ψi ∇ϕ
them (Galerkin-style method)
Assume Dirichlet boundary conditions
ϕ|S = 0 (we’ll see Neumann b.c. later)  The last term is zero since ψi |S = 0
Poisson equation:  Let’s expandXϕ over ψi :
ϕ(r ) = αi ψi (r ) (1)
ρ(r )
∆ϕ(r ) = −
ε0
ˆ ˆ
⇒Assume now we have arbitrary complete 1
dV ρψi = αk ∇ψi · ∇ψk dV
set of function ψi (r ), such that ψi |S = 0 ε0
| {z } | {z }
(we do NOT impose any other conditions Ri Mik
on ψi !)
⇒Got linear matrix
X equation
ˆ ˆ ⇒ Mik αk = Ri
1
⇒ ψi ∆ϕ dV = − dV ρψi k
⇒ αi = M −1 ik Rk


ε0
ψi ∆ϕ = ∇ (ψi ∇ϕ) − ∇ψi · ∇ϕ so can use (1) to reconstruct ϕ(r ). Usually
can cut full (infinite) sum (1) and work
with finite set of functions.
Electrostatics - Numerical methods
Choice of functions ψi : Rectangular grid is very simple, might
 Consider arbitrary discrete grid with be ineffective (require too many nodes N)
nodes r i in the region of interest if the body has sharp borders/angles
introduce functions ψi (r ) = f (r − r i ), In numerical analysis there are special
where f (0) = 1 but vanishes very rapidly techniques how to add nodes in a sys-
as a function of r − r i tematic and efficient way (see e.g. mesh-
Example from Jackson: ing techniques for Finite Element Anaylisis
use rectangular grid with step h in all (FEA) in Google)
dimensions Increase number of nodes near sharp
Choose f (r ) as borders, where field should change rapidly:

1 − |rh| , |r | ≤ h

f (r ) =
0, |r | > h

 all ψi are concentrated near the node r i

and vanish at large distances > h (as well
as on the surface S, as required)
. For legibility consider 2D
problem, surface S =red
contour; use rectangular
grid. Nodes are located at
(implemented in all modern software)
intersection of black lines
Electrostatics - Numerical methods
 Dirichlet boundary conditions ϕ|S = 0
ρ(r )
∆ϕ(r ) = − Control question
ε0
In our consideration (see left column) we
⇒Assume ∃ set of functions ψi , such that assumed that ϕ|S = 0. Explain in which
∀ψi (r ), ψi |S = 0 place in our derivation we implicitly used
ˆ ˆ this condition. How would change the
1
⇒ ψi ∆ϕ dV = − dV ρψi solution if we had Dirichlet boundary
ε0
condition ϕ(r )|S = b(r )|S , where b(r ) is
ˆ ˆ
1 an arbitrary (known) function.
dV ρψi = ∇ψi · ∇ϕ dV −
ε0
˛ Please let me know your
− dS ψi ∇ϕ replies/suggestions

 The last term

X is zero since ψi |S = 0
ϕ(r ) = αi ψi (r ) (1)
ˆ ˆ
1
dV ρψi = αk ∇ψi · ∇ψk dV
ε0
| {z } | {z }
Ri Mik
⇒ αi = M −1 ik Rk


Electrostatics - Numerical methods
Assume Dirichlet boundary conditions Reply:
ϕ|S = 0  In (1) we assumed that all ψi vanish at
Poisson equation: the border, so any finite sum will eventu-
ally lead to ϕ(r )|S = 0
ρ(r )
∆ϕ(r ) = −  For ϕ(r )|S ̸= 0 we may rewrite main
ε0 equation as
X
ϕ(r ) = αi ψi (r ) (1) ρ(r ) + ε0 ∆b(r )
∆ (ϕ(r ) − b(r )) = −
| {z } ε0
Control question ϕ̃(r )

In our consideration we assumed that ρ̃(r )

∆ϕ̃ := − ,
ϕ|S = 0. Explain in which place in our ε0
derivation we implicitly used this and work with ϕ̃, ρ̃ from now on.
condition. How would You change the
Obviously, ϕ̃(r ) = 0, so now can use

solution if we had Dirichlet boundary S
condition ϕ(r )|S = b(r )|S , where b(r ) is the result (1) found earlier.
some known function.
Electrostatics - Numerical methods
 Dirichlet boundary conditions ϕ|S = 0
ρ(r )
∆ϕ(r ) = − Control question 2
ε0
In our consideration we assumed Dirichlet
⇒Assume ∃ set of functions ψi , such that boundary conditions for ϕ|S . Analyze
∀ψi (r ), ψi |S = 0 how the derivation would change if we
ˆ ˆ had Neuman boundary condition
1
⇒ ψi ∆ϕ dV = − dV ρψi 
ε0 ∂ϕ(r ) 
ˆ ˆ = b(r )|S
∂n S
1
dV ρψi = ∇ψi · ∇ϕ dV −
ε0 where b(r ) is some known function.
˛
− dS ψi ∇ϕ
Please let me know your
replies/suggestions
 The last term
X is zero since ψi |S = 0
ϕ(r ) = αi ψi (r ) (1)
ˆ ˆ
1
dV ρψi = αk ∇ψi · ∇ψk dV
ε0
| {z } | {z }
Ri Mik
⇒ αi = M −1 ik Rk


Static electric fields (summary)
Electrostatics Border conditions
For linear medium (weak fields) just (1)
E|| = E||
(2)
replace ε0 → εε0 :
(1) (2) (1) (2)
D⊥ = D⊥ ⇔ ε(1) (2)
r E⊥ = εr E⊥
ρ
∇·E = , Sometimes we might have surface
εε0
charges at the border of two dielectrics
∇×E =0
(e.g. charge on the surface of the plas-
εr -relative permittivity. tic stick in the air). In this case
(1) (2)
Additional useful quantitites: D⊥ − D⊥ = Σ,
⋆“electric displacement field” D ⃗,
⃗ = εr ε0 E
where Σ is the surface charge density
so

∇ · D = ρ, I will stop here discussion of static

electric field. In the book of Jackson You
∇×E =0
may find a lot of nice examples which I
⋆χ = εr − 1 - electric susceptibility recommend You to see in detail.
⋆P⃗ = ε0 χ E
⃗ -polarization density (dipole
moment per volume)
Magnetostatics

. Please try to recall everything You’ve seen

earlier (in undergraduate courses) about
static magentic fields, associated forces,
etc
Magnetostatics
. Experimental evidence:
. Force acting between
two conductors
µ0 I1 I2
F /ℓ =
2π r
Lorentz’ force acting on charge q in elec- µ0 -permeability
tromagnetic field
 
⃗ =q E
F ⃗ + ⃗v × B
⃗
⇒ A conductor creates a magnetic field
Force acting on a thin wire with electric around it,
µ0 I
current I in external magnetic field B: B∼
ˆ ˆ 2πr
⃗ mag = I d ⃗ ⃗ = ⃗j × B ⃗ dV and the other conductor interacts with its
F ℓ×B
magnetic field¸
Note that B dℓ = µ0 I where I is the
 Since conductor’s net charge is zero, so
current inside closed contour
we don’t have electric component
Technicial comment

In what follows we’ll frequently get structures like

ˆ
d 3 r ′ j r ′ (...)

 In lab setup current usually flows in a thin conducting wire. In this case can split
volume integral into transverse and longitudinal parts,
ˆ ˆ ˆ
d 3 r ′ → dℓ dS⊥ ,

so ˆ ˛
3 ′ ′

d r j r (...) ≈ dℓ I (ℓ) (...)

where ˆ
dℓ
dS ⊥ · j r ′


I (ℓ) =
dℓ
is the current in the conductor.
Magnetostatics
. So far we wrote everything for magnitude.
Now rewrite everything for vectors:

Biot-Savart law

ˆ
⃗ (r ) = µ0 I d⃗
ℓ1 × r̂ 12
B =
Lorentz’ force acting on charge q in elec- 4π 2
r12
tromagnetic field
|{z}
const
ˆ ⃗
µ0 j (r 1 ) × r̂ 12
 
⃗ =q E
F ⃗ + ⃗v × B
⃗ = d 3 r1 2
4π
|{z} r12
Force acting on a thin wire with electric const

current I in external magnetic field B:

ˆ ˆ where r 12 = r − r 1 is a vector between
⃗ mag = I d ⃗
F ℓ×B ⃗ = ⃗j × B ⃗ dV observer and current

Magnetic field around conductor There is no magnetic monopoles,

µ0 I ⃗ ≡∇·B =0
divB
B∼
2πr
˛ div j = 0 in magentostatics
⃗ = µ0⃗j
B dℓ = µ0 I ⇒ rotB Charge conserv.: ρ̇+div j = 0 ⇒ ρ̇ = 0
Static magnetic fields
The magnetic field B does not change if
we add gradient of arbitrary function f
Magnetostatics
⃗ →A
A ⃗ + ∇f
⃗ = µ0⃗j
rotB (1)
(this is the so-called gauge transforma-
⃗ ≡∇·B =0
divB (2) tion).
(2) implies that we can introduce the
⃗
(1,2)⇒equation for A:
⃗ such that
vector potential A,
⃗ := rot A
⃗ ∇ (∇ · A) − ∆A = µ0 j (3)
B (3)
⃗ is not completely determined from Am-
A
pere’s law, there is a freedom of recalibra-
tion
Maxwell’s equations
Gauge freedom Gauge fixing
Statics: A → A + ∇f (x), ⃗
Should fix “longitudinal part” of A
Choice is not unique, can have various
⃗ =∇×A
B (1) gauge conditions

Our definition (1) is incomlete, doesn’t

⃗ Coulomb gauge : ∇·A=0 (k · a = 0)
fix properlyˆlogitudinal part of A:
d 3
k Axial gauge : n·A=0
⃗
A(x) = ⃗a (k)e ik·x ,
(2π)2
ˆ where n is some fixed vector (e.g. for n =
⃗ d 3k ⃗ (0, 0, 1) get A3 = 0)
B(x) = b(k)e ik·x ,
(2π)2 ⋆Longitudinal component:
⃗
a(k) = ⃗
a|| (k) + ⃗
a⊥ (k)
a|| = |a|| |k̂,
⃗
a|| (k) = const ⃗
⃗ k ⇒ k × a|| = 0
so if we know ⃗
a⊥ , we get
⃗ a(k) − ⃗
a⊥ (k) = ⃗ a|| (k)
0 = n · a = n · a|| + n · a⊥
⃗ fixes only ⃗
Physical field B a⊥ part, ⃗
a|| does
  (n · a⊥ )
not contribute: ⇒ a||  = −  
⃗b = ⃗
k × ⃗a (k) ⃗ k · ⃗a =??? n · k̂
  ⃗ do not depend on this
Physical field B
ki kj
(3) :k 2 δij − 2 aj = µ0 j˜i choice
k
Static magnetic fields

Magnetostatics Control question

⃗ = µ0⃗j
rotB ⃗ assuming
Solve the equations for A
additional (Coulumb) gauge condition
⃗ ≡∇·B =0
divB ∇·A=0

⃗ := rot A
B ⃗
∇ (∇ · A) − ∆A⃗ = µ0⃗j (1)
Electro- & Magnetostatics (comparison)
Electrostatics in brief Magnetostatics in brief

ρ ⃗ = µ0⃗j
rotB
⃗ = −∆ϕ =
∇·E ,
ε0
⃗ =0⇒E
∇×E ⃗ = −∇ϕ. ⃗ ≡∇·B =0⇒B
divB ⃗
⃗ = rotA
∇ (∇ · A) − ∆A = µ0⃗j (1)

⃗ is fixed up to gauge transformation

A
A → A + ∇f
Assumed Coulomb gauge ∇ · A = 0:

Eq. for A in Coulomb gauge in Carte-

sian coords coincide with Poisson Eq.
for potential ϕ, so can repeat verbatim
what we did for electrostatics
Green function & particular solution
Equation in Coulomb gauge is similar to Poisson, so can use results which we got
earlier:
∆A = −µ0⃗j ∇·A=0
ˆ
µ0 j (r ′ )
⇒ Apart = d 3r ′ ,
4π |r − r ′ |
ˆ  
µ0 3 ′ ′
1
∇·A= d r j r · ∇r ′ =
4π |r − r ′ |
ˆ
µ0 1
d 3r ′ ∇r ′ · j r ′ = 0


=−
4π |r − r ′ | | {z }
=0

⇒ equations are compatible in view of charge conservation. So now we can literally

repeat what we did earlier for electrostatics:
 e.g. Green function for inhomogeneous part (no border conditions):
∗ ′  ℓ


1 1 X Yℓm (r̂ ) Yℓm r̂ r<
G =− = −
4π|r − r ′ | r> 2ℓ + 1 r>
ℓm

where r< = min(r , r ′ ), r> = max(r , r ′ ). Recall that we can add solution of
∆Ghomog (r , r ′ ) = 0 in order 
to meet boundary
  conditions,
 so the full result is
′
X γℓ δℓ ∗
ℓ ℓ
r̂ ′


G (r , r ) = Cℓ r> + ℓ+1 r< + ℓ+1 Yℓm (r̂ ) Yℓm
ℓm
r> r<
 ˆ
µ0 I
Aφ = r ′2 dr ′ sin θ′ dθ′ dφ′ ×
Example
. problem 4πa
δ(r ′ − a)
Evaluate the magnetic field created by a × sin θ′ cos φ′ δ(cos θ′ )
|r − r ′ |
current I flowing inside a circular ring of
radius a, as shown in the Figure.
r − r ′ 2 = r 2 + r ′2 −
 

− 2rr ′ (cos θ cos θ′ − sin θ sin θ′ cos(φ − φ′ ))

Use δ-function properties:

ˆ
µ0 aI
Aφ = dϕ′ cos φ′ ×
4π
−1/2
× r 2 + a2 − 2ar sin θ cos φ′

First evaluate potential A
Easier to work in spherical coordinates:
(Mathematica: lengthy expression, in-
 The only nonzero component j φ ⇒only cludes elliptic functions)
Aφ ̸= 0 Magnetic field: B = rotA
1 ∂
j φ (r ) = Jφ (− sin φx̂ + cos φŷ ) Br = (sin θAφ ) ,
r sin θ ∂θ
I 1 ∂
Jφ = δ(r − a)δ (θ − π/2) = Bθ = − (r Aφ ) , Bφ = 0
a r ∂r
I
= sin θδ(r − a)δ (cos θ) This gives formal solution for A, B.
a
Example problem  Useful expression:
Now we will try to find an approximate  
result for the case when a ≪ r . cos ϕ′ ≡ e iφ + e −iφ /2
Main tool: Green function expansion ˆ
over spherical harmonics: ⇒ dφ′ → π (δm,1 + δm,−1 )

1
G =− = Aφ = 2πµ0 aI × (1)
4π|r − r ′ |
∗ π ℓ


∗ ′ 

ℓ 1 Yℓ,1 (r̂ ) Yℓ,1 , 0 r<
1 X Yℓm (r̂ ) Yℓm r̂
X
r< × 2
=− r> 2ℓ + 1 r>
r> 2ℓ + 1 r> ℓ
ℓm
∗ π


where r< = min(r , a), r> = max(r , a) Yℓ,1 2
, 0 is just some number,
ˆ
0,

µ0 I ℓ = even
Aφ = r ′2 dr ′ sin θ′ dθ′ dφ′ ×

π n+1 3
Γ(n+ 2 )
  q
a ∗ 2ℓ+1 (−1)
Yℓ,1 ,0 = 4πℓ(ℓ+1) Γ(n+1)Γ( 3 )
,
2
× sin θ′ cos φ′ δ(cos θ′ )δ(r ′ − a) 2


for ℓ = 2n + 1
∗ ′  ℓ


1 X Yℓm (r̂ ) Yℓm r̂ r<
× = and Yℓ,1 (r̂ ) ∼ Pℓ1 (cos θ) e iφ .
r> 2ℓ + 1 r>
ℓm
ˆ In (1) the dominant contribution comes
= µ0 aI dφ′ cos φ′ × from ℓ = 1, so Aφ ∼ 1/r 2 .
Now we’ll revisit general multipole ex-
∗ pansion for magnetic fields and try to un-
π

 ℓ
1 X Yℓm (r̂ ) Yℓm 2 , 0 r<
× derstand why there is no ℓ = 0 term
r> 2ℓ + 1 r>
Multipolar expansion ´
 Derivation similar to electrostatic case Lemma: d 3 r 1 j (r 1 ) = 0
(just replace ρ → j ):
ˆ (obvious for closed current loop)
µ0 j (r 1 ) general proof: let’s choose some smooth
Apart (r ) = − d 3r 1 ,
4π |r − r 1 | function f (r ) and assume j (r ) vanishes at
∞:
r ˆ ˛
r2 r1 · r
|r − r 1 | = r 1 + 12 − 2 2 d 3 r ∇ · (f j ) = f j · dS = 0
r r
r12 (r 1 · r )2
 
r1 · r ∇ · (f j ) = f ∇ · j + (j · ∇) f
≈r 1− 2 + 2 −
r 2r 2r 4
∇ · j = 0 static charge conservation
ˆ
ˆ
µ0 ⇒ d 3 r (j · ∇) f (r ) = 0
Apart (r ) = − d 3 r 1 j (r 1 ) ×
4πr
 Let’s choose f (r ) = ri (one of the com-
r2 3(r 1 · r )2
 
r1 · r
× 1 + 2 − 12 + + ... ponents in Cartesian ref. frame, e.g. x, y
r 2r 2r 4 or z) ˆ
⇒ d 3 r ji (r ) = 0

for all components i, Q.E.D.

⇒there is no monopole term contribution
in Apart
Multipolar expansion  Let’s choose f (r ) = ri , g (r ) = rk (one
ˆ of the components in Cartesian ref. frame,
µ0
Apart (r ) = d 3 r 1 j (r 1 ) × e.g. x, y or z)
4πr ˆ
r2 3(r 1 · r )2
 
r1 · r ⇒ d 3 r [ri jk (r ) + rk ji (r )] = 0 (1)
× 1 + 2 − 12 + + ...
r 2r 2r 4

Lemma for all components i, k

ˆ ⇒if we have external vector a, then
d 3 r 1 j (r 1 ) (r 1 · r ) = −r × m, ˆ
ˆ ⇒ d 3 r [(a · r ) j + r (a · j )] = 0 (1′ )
1
m≡ d 3 r 1 r 1 × j (r 1 )
2
ˆ
1
Proof: let’s choose some smooth func- r ×m = d 3 r 1 r × [r 1 × j (r 1 )] =
2
tions f (r ), g (r ) and assume j(r ) vanishes ˆ
1
at ˆ
∞: ˛ = d 3 r 1 [r 1 (r · j (r 1 )) −
2
d 3 r ∇ · (f g j ) = f g j · dS = 0
− j (r 1 ) (r 1 · r )]
∇ · (f g j ) =f g ∇ · j +
Now apply (1’) to the first term:
+g (j · ∇) f + f (j · ∇) g ˆ
∇ · j = 0 charge conservation ⇒ r × m = − d 3 r 1 [j (r 1 ) (r · r 1 )]
ˆ
3
d r [g (j · ∇) f (r ) + f (j · ∇) g (r )] = 0 Q.E.D.
Magnetic dipole contribution

µ0 µ0
A(r ) = m×r = m × r̂ , Demonstrate that for a closed plain
4πr 3 4πr 3
r contour with infinitely thin wire with
r̂ ≡ current I inside, the magnetic moment is
r ˆ
1 given by m = I × area
m≡ d 3 r 1 r 1 × j (r 1 )
2
µ0 Please let me know you ideas how we
B = rotA = [3r̂ (r̂ · m) − m] can prove this?
4πr 3
Similarly we can analyze higher order
terms in multipolar expansion (quadrupole
etc). Usually those terms are negligibly
small.
Magnetic dipole contribution More formal proof:
Let’s evaluate m ·a where a is constant
Demonstrate that for a closed plain fixed vector. ˆ
contour with inifinitely thin wire with 1
m·a ≡ d 3 r 1 a · [r 1 × j (r 1 )] =
current I inside it the magnetic moment 2
˛
is given by m = I × area I
ˆ = dℓ1 · [a × ℓ1 ]
1 2
m≡ d 3 r 1 r 1 × j (r 1 ) =
2 Can apply Stokes’ theorem for circula-
˛
=
I
ℓ1 × dℓ1 ˆ
tion of vector field
2 I
m·a = dS · rot [a × ℓ1 ]
1 2
Geometry: ℓ1 × dℓ1 = da n ⊥
2 rot [a × ℓ1 ] = ∇ℓ × [a × ℓ1 ] =
= a (∇ℓ · ℓ1 ) − (a · ∇ℓ ) ℓ = 2a
| {z } | {z }
3 a
ˆ
m · a = I dS · a


´ Since a = const and contour is plain,
dS · a = S n ⊥ · a
where da is the area of the triangle; n ⊥ is m · a = I S n⊥ · a
normal to the plane of the contour
⇒Full integration yields the area inside the Since a is arbitrary, conclude that m =
whole contour, Q.E.D IS n ⊥ , Q.E.D.
Magnetic dipole contribution

Please let me know your ideas ...

ˆ
1
m≡ d 3 r 1 r 1 × j (r 1 )
2

Demonstrate that for a pointlike particle

of mass µ moving along some trajectory
r 0 (t) the magnetic dipole moment
q
m = 2µ L where L is angular momentum.
Can we generalize this result for a
many-body system ?
Magnetic dipole contribution  Generalization for multiparticle sys-
 Single particle of mass µ: tem:
Angular momentum Angular momentum
X
L= µi r 0i (t) × v i (t),
L = r 0 (t) × p = µ r 0 (t) × v (t), i

Current density: Current density:

X
j (r , t) = qi v i (t) δ 3 (r − r 0i (t)) ,
j (r , t) = q v (t) δ 3 (r − r 0 (t)) ,
i
v (t) = ṙ 0 (t) v i (t) = ṙ 0i (t)
Magnetic moment:
Magnetic moment:

1 1X
m= q r 0 (t) × v (t) = m= qi r 0i (t) × v i (t) =
2 2
i
q q X qi
= µr 0 (t) × v (t) = L = µi r 0i (t) × v i (t)
2µ 2µ 2µi
i

⇒m ∼ L if & only if qi /µi is the same

for all particles
Magnetic dipole contribution

Please let me know your ideas how to

ˆ
1 prove this
m≡ d 3 r 1 r 1 × j (r 1 )
2

Demonstrate that for a localized

distrbution of currents magnetic moment
is related to integral of magnetic field B
as ˆ
2µ0
m= BdV
3 r <R

where integral is taken over the whole

volume which encloses the currents j
Magnetic dipole contribution Now use decomposition of Green function:
Assume we
ˆ have a sphere
ˆ of radius R 1
=
dV B = dV ∇ × A 4π |r − r 1 |
∗ ′  ℓ


1 X Yℓm (n̂) Yℓm r̂ r<
Let’s multiply both parts by arbitrary con- =
r> 2ℓ + 1 r>
ˆ vector a
stant ˆ ℓm

dV B · a = − dV ∇ · (a × A) = (obviously, r̂ ≡ n)
˛ Let’s focus on surface ´ integral over ori-
= − dS · (a × A) = entation of n̂, c ℓm = dΩ Yℓm (n) n
˛ Components of n =
= dS a · (n × A) (sin θ cos ϕ, sin θ sin ϕ, cos θ), linear
combinations of Y1m :
where dS = dS n
Since aˆ is arbitrary,
˛ conclude that
dV B = dS (n × A)
¸ ´
R 2 dΩ ˆ
dS ≡
µ0 j (r 1 )
A(r ) = d 3r 1 ,
4π |r − r 1 |
ˆ ˛ ˆ
µ0 j (r 1 ) -In view of orthonormality only ℓ = 1
dV B = dS n × d 3 r 1 might contribute.
4π |r − r 1 |
Magnetic dipole contribution
Evaluating coefficients c ℓm with Mathematica & resumming, get
∗
r̂ ′ c ℓm


X Yℓm
∼ r̂ ′
2ℓ + 1
ℓm

′
Inside the sphere r< =ˆ r , r> = R ˆ
4µ0 2µ0
dV B = d 3 r 1 r 1 × j (r 1 ) = m
3 3
Q.E.D.
 Note that if we used “direct” magnetic dipole contribution, we’d get different result:

µ0
B dipole = [3r̂ (r̂ · m) − m]
4πr 3

3r̂ (r̂ · m) − m ∼ P2 (cos Θ) ∼ 3 cos2 Θ − 1

where Θ is angle between r̂ and m⇒after angular averaging get zero (!)
Need to correct our result,
µ0 2µ0
B dipole = [3r̂ (r̂ · m) − m] + mδ(r )
4πr 3 3
The last term does not contribute for r ̸= 0
Magnetostatics in medium

. Please explain, how the expressions

shown in the left Figure change in the
medium? What border conditions should
we impose at the border of two materi-
als?
Magnetostatics in medium µr is very close to 1 for dia/paramagentics:
 Assume linear medium, weak field

⃗ = µr µ0⃗j
rotB

⃗ ≡∇·B =0
divB

µr is called relaltive magnetic permeabil- −In ferromagnetics have histeresis: rela-

ity tion between M and B depends on history.
⋆χm = µr − 1 is called magnetic suscep-
tibility
⋆Introduce vector H ⃗

⃗ = 1 B
H ⃗ −M⃗
µ0
⋆Magnetization M:⃗ magnetic dipole mo-
ment per unit volume; M⃗ ∼ B
⃗ for linear
dia/paramagnetics
−will not consider them here ...
−In isotropic dia/paramagnetic materials
M⃗ ∼ B, ⃗ so µr µ0 H⃗ = B⃗ (differ by a
const)
Border conditions . Evaluate the circulation
around a small contour
of length ℓ and negligible
Ampere’s law
width
⃗ = 1 B⃗ −M ⃗ ˛
H  
µ0 rotH ⃗ ⃗l ≈ H (1) − H (2) ℓ
⃗ = 0 ⇒ 0 = H·d
|| ||
⇒∇×H ⃗ = ⃗J f
(1) (2)
⃗ =0
divB H|| = H||
NOTE: For magentized materials, need
⃗f ,
to add ficticious surface current K
Assume that we consider a border of
two diamagentic materials. Write out the (1) (2) ⃗f × n
H|| − H|| = K ⃗
border conditions for the magentic field.
Assume there is no “free” currents. . Evaluate the flux of mag-
netic field through the rect-
angular parallelepiped
of area S and negligible height.
˛  
divB⃗ = 0 ⇒ 0 = B·d ⃗ S⃗ ≈ B (1) − B (2) S
⊥ ⊥

(1) (2)
B⊥ = B⊥
Magnetic fields
Magnetostatics in the medium Control question
⃗ = ⃗j
rotH Assume that inside paramagnetic
material there is a homogeneous
magnetic field B ⃗ 0 . There are two cavities
⃗ ≡∇·B =0
divB inside it:
⃗ = εr ε0 E
⋆“electric displacement field” D ⃗ (a) inifitely long needle-shaped cavity
whose axis is collinear to the direction of
the field.
Border conditions
(b) inifitely thin wafer-shaped cavity
normal to the direction of the field.
(1) (2) (1) (2)
H|| = H|| ⇔ B|| /µ(1)
r = B|| /µr(2) Evaluate the electric field inside the
cavities
(1) (2)
B⊥ = B⊥

Caution
Note that in general ∇ · H ⃗ ̸= 0 near the
border of two magnetic materials
(µr ̸= const). For this reason we can’t
⃗ →H
replace B ⃗ everywhere.
Scalar magnetic potential
 In absence of charges/currents electro-
and magnetostatics practically coincide:
A hollow sphere with internal radius a
divE = 0, rotE = 0, (1) and external radius b is made of
divB = 0, rotB = 0. (2) paramagentic material with (relative)
permeability µ > 1. The sphere is
Potential ϕ: reduces (1) to very simple immered into the homogeneous magentic
Laplace equation field B 0 . Evaluate the magnetic field
 (1, 2) have similar form⇒ can try to inside and outside the sphere. Note that
introduce ΦM , B ≡ ∇ΦM if µ ≫ 1, then field inside is significantly
smaller than outside (“magentic
⇒ divB ≡ ∆ΦM = 0 shielding”)
 ΦM is unambigously defined only in ab-
sence of currents
ΦM =static magnetic potential
nevetheless helps significantly in some
problems
Caution: We should use either A or ΦM
in all parts of the problem, but not both!
Scalar magnetic potential  Expect no φ-dependence, so for m ̸= 0
. all aℓm , bℓm = 0
 At r → ∞ solution
 should reduce
 to
B 0 = B0 ẑ = B0 r̂ cos θ − θ̂ sin θ
 At r → 0 solution should remain finite

 P αℓ
−B0 r cos θ + ℓ ℓ+1 Pℓ (cos θ), r > b
r


P  γℓ

ΦM = ℓ
ℓ β ℓ r + Pℓ (cos θ), r ∈ (a, b)
 r ℓ+1
 P ℓ
ℓ δℓ r Pℓ (cos θ), r <a


Can introduce both ΦM and A since have For border conditions take into account
no currents that
Easier to use scalar magnetic potential ∂ΦM
ΦM instead of vector potential A (just sin- B⊥ ≡ Br = ,
∂r
gle function)
1 ∂ΦM
Assume axis z points in direction of B 0 , B|| ≡ Bθ = ,
r ∂θ
use spherical coords.
General solution of Laplace: Each Pℓ is independent, so need to im-
X ℓ bℓm

pose border conditions separately on each
ΦM = aℓm r + ℓ+1 Yℓm (θ, φ) P
r term in the sum ℓ .
ℓm
|m|
Yℓm ∼ Pℓ (cos θ) e imφ

( aℓm , bℓm might differ inside and outside)

ℓ ̸= 1:
αℓ
⊥: − (ℓ + 1) b 3 B 0 µ2 + µ − 2 b 3 − a 3



=
b ℓ+2 α1 = 3 ,

γℓ (ℓ + 1)
 b (2µ2 + 5µ + 2) − 2a3 (µ − 1)2
= βℓ ℓb ℓ−1 − ,
b ℓ+2 3b 3 B0 (µ + 2)
β1 = ,

γℓ (ℓ + 1)
 b3 (2µ2 + 5µ + 2) − 2a3 (µ − 1)2
ℓδℓ aℓ−1 = βℓ ℓaℓ−1 −
aℓ+2 3a3 b 3 B0 (µ − 1)
αℓ γℓ  γ1 = ,
2a3 (µ − 1)2 − b 3 (2µ2 + 5µ + 2)

ℓ−1
|| : ℓ+2
= µ βℓ b + ℓ+2 ,
b b 9b 3 B0 µ
 γℓ  δ1 =
δℓ aℓ−1 =µ βℓ aℓ−1 + ℓ+2 b3 (2µ2 + 5µ + 2) − 2a3 (µ − 1)2
a
 linear homogeneous system for coeffi- Inside the sphere the field is controlled by
cients αℓ , βℓ , γℓ , δℓ δ1 ; for µ → ∞ see that
 In general system not degenerate 9B
⇒ ∃! solution αℓ , βℓ , γℓ , δℓ = 0 δ1 ≈  0  ≪ B0
a3
ℓ = 1:   2µ 1 − b3
2α1 2γ1
⊥: B0 − 3 = β1 − 3 ,
b b We have seen that use of scalar magnetic
potential ΦM permitted to get analyitcal
 
2γ1
δ1 = β1 − 3 solution
a
α1  γ1  In Jackson, Section 5, You may find
|| : B0 + 3 = µ β1 + 3 ,
b  b more examples how ΦM could be used
γ1 
δ1 =µ β1 + 3
a