Optical field calculation using

the Fourier Modal Method

In this exercise we will compute the scattering of light due to periodic grating structure in the
2D x-z plane. This corresponds to light propagating in the x-z plane and being scattered by an
object of infinite extent along the y axis. For light polarized along the y axis, the problem can
be treated using the 2D scalar Helmholtz equation

 ∂2 ∂2 
 x2 + 2
+ k02ε ( x, z )  E y ( x, z ) =
0. (1)
∂ ∂z 
In the following we will mostly skip the subindex y for simplicity. As usual we assume a time
dependence e − iωt with= k0 ω= / c 2π / λ . The complete geometry is illustrated below. It
consists of a glass substrate which has been etched to form a grating periodic along the x axis.
The geometry can be divided into three parts, layer 1 is air, layer 2 is the grating and layer 3 is
the substrate.
z = ztop
3 Λ a ε1

z=L ε2

2 L

z=0 z

1
θ
x
z = zbot x=0
Fig. 1: Grating structure illuminated by a plane wave. The filling factor f is defined as f = a/Λ.

The grating is illuminated with by an incoming plane wave from below of the form
Ein ( x, z ) exp(ik1 ( z cos θ + x sin θ )) ,
= (2)

where k1 = ε11/2k0 , and the light is partially reflected and partially transmitted by the grating.

We will use the Fourier Modal Method (FMM) to compute the scattered field. This method is
particularly suitable for geometries featuring periodicity in a lateral x-y plane and piecewise
uniformity along a propagation z axis. In the FMM, the field is expanded not on grid points as
in FDTD and FEM, but instead on the eigenmodes of layers featuring uniformity along the z
axis. The fields at each side of an interface between layers are connected using the scattering
matrix formalism. The FMM thus features three basic steps:
1) Calculation of eigenmodes in each layer assuming uniformity along the z axis.
2) Determination of reflection and transmission matrices for each interface assuming that
no other interfaces exist.
3) Iterative calculation of full scattering reflection and transmission matrices describing
the full structure under study.
Part 1. Eigenmode calculation

Due to the periodicity along the x axis, we have the following boundary condition
E ( x + Λ, z ) = E ( x, z )eiα , (3)

where the parameter α depends on the illumination conditions. For the incoming plane wave
illustrated in Fig. 1, we have
= α k1 sin (θ ) Λ ≡ k x Λ .

With the boundary condition (3), we only need to consider the computational domain 0 < x <
Λ. The field for x < 0 and x > Λ is obtained using (3).

The Helmholtz equation describing the propagation of light in the 2D geometry is given by
∇2 E ( x, z ) + ε ( x, z )k02 E ( x, z ) =
0. (4)

The dielectric constant ε ( x, z ) for the full geometry depends on both the x and the z
coordinates. However, we can split the geometry into layers uniform along the z axis:

Layer 1: ε (1) ( x ) = ε1
Layer 2: ε (2) ( x ) = ε 2 (0 < x < a) ε (2) ( x ) = ε1 (a < x < Λ)
Layer 3: ε (3) ( x ) = ε 2

We will now solve the Helmholtz equation for the dielectric profiles above, that is
∇2 E ( x, z ) + ε ( q ) ( x )k02 E ( x, z ) =
0, (5)

where q equals either 1, 2 or 3.

Due to the uniformity along the z axis, one may use separation of variables and define a trial
solution of the form E ( x, z ) = E ( x )e ± iβ z . Insertion of this solution into (5) gives

∂2
E ( x ) + ε ( q ) ( x )k02 E ( x ) =
β 2 E ( x) . (6)
∂x 2
This is an eigenvalue problem with eigenmodes E ( x ) and eigenvalues β 2 . The eigenvalues
are the squares of the propagation constants ± β, where the ± indicates two possible directions
of propagation either along the positive ( + β ) or negative ( − β ) z axis.
In layers 1 and 3, the solutions to (6) are simply plane waves. We may now describe our
eigenmodes using the index m taking integer values between 1 and the total number M of
plane waves in the calculation. The electric field profile for the m’th mode is then given by

1;m ( x )
E= 3;m ( x )
E= exp(ik x ,n x ) (7)
β1;m =
± ε1k02 − k x2,n (8)

β 3;m =
± ε 2k02 − k x2,n , (9)

where k x ,=
n k x + nG , n being a positive or negative integer related to m as discussed
subsequently and
= G (2π ) / Λ . In a numerically exact calculation, the index n should run
 from –∞ to +∞, however to store the field in computer memory we limit the number of plane
waves to the interval –N to N, such that M ≡ 2N+1 plane waves are included in the
calculation. In that case, we chose n = m − 1 − N in Eqs. (7)-(9).
In anticipation of the subsequent plane wave expansion of the eigenmodes, we may write (7)
as
n=N
E1;m ( x ) = ∑u
n= − N
1;nm exp(ik x ,n x ) , (10)

where u1;nm are plane wave expansion coefficients of the m’th mode given in layers 1 and 3
simply by u1;nm = δ n ,m −1− N .
In layer 2, the solution to (6) is computed by performing a plane-wave expansion of the
electric field and the dielectric constant as
n=N
E2;m ( x ) = ∑u
n= − N
2;nm exp(ik x ,n x ) (11)
n =2 N
ε (2) ( x ) = ∑
n = −2 N
ε n(2) exp(inGx ) . (12)

Here the expansion coefficients ε n(2) for the dielectric constant are given by
x= Λ
1
ε ∫ε
(2) (2)
= n ( x ) exp( −inGx )dx , (13)
Λ x =0

and the electric field expansion u2;nm coefficients for the m’th mode remain to be determined.
In the following we sometimes skip the layer indices (2) and 2 for simplicity. We now insert
(11)-(12) into (6) leading to
n=N n ' N=
= ,n '' N
− ∑ k x2,n unm exp ( i ( nG + k x ) x ) + k02 ∑ ε n ''un ' m exp ( i ( n '+ n '')Gx + ik x x )
n=
−N n '=
− N ,n '' =
−N
n=N
, (14)
= β 2
∑u
n= − N
nm exp ( i ( nG + k x ) x )

which can only be fulfilled if

n '= N
−k x2,n unm + k02 ∑ε
n '= − N
u
n −n ' n ' m β 2unm .
= (15)

We now introduce the vector for the expansion coefficients of the m’th mode
T
um = u− N ,m u− N +1,m ... uN −1,m uN ,m  , (16)

where the coefficients u1;nm are indexed from -N to N, such that the first (last) element in the
vector (16) corresponds to the index -N (N). This choice of indexing of the vector elements is
motivated by the index range in the eigenmode expansions (10) and (11). Furthermore, we
define
 ( − NG + k x ) 2 0 0 0 0 
 2 
 0 (( − N + 1)G + k x ) 0 0 0 
A=  0 0 ... 0 0  (17)
 
 0 0 0 (( N − 1)G + k x ) 2 0 
 0 0 0 0 2
( NG + k x ) 

and

ε 0(2) ε −(2)
1 ε −(2)2 ... ε −(2)2 N 
 (2) 
ε 1 ε 0(2) ε −(2)1 ... ε −(2)
2 N +1 

B = ε 2(2) ε1(2) ε 0(2) ... ε −(2)

2 N +2
. (18)
 
 ... ... ... ... ... 
ε 2 N ε 2 N −1 ε 2 N −2
(2) (2) (2)
... ε 0(2) 

This special type of matrix (18) with matrix elements given by Bmn = ε m(2)−n is known as a
Toeplitz matrix. Using this notation, (15) can be written as

(− A + k B) u
2
0 m =β m2 um . (19)

Equation (19) is a standard eigenvalue problem which can be solved using the EIG command
in Matlab. When given an M * M size matrix as input, the EIG command will return M
eigenvectors and M eigenvalues. Each eigenvector is of the form (16) and represents an
eigenmode profile as given by (11). The associated eigenvalue is the propagation constant
β m2 . We see that the number of plane waves MPW used in the expansion and the number of
eigenmodes MEM returned by EIG are identical. In principle one can choose to retain a smaller
set of eigenmodes such that MEM < MPW. However, for the following calculations of reflection
and transmission matrices, it will be convenient to work with MEM = MPW. Usually, one
manually sorts the eigenmodes and eigenvalues given by the EIG command according to
decreasing Re( β 2 ) .
The eigenvectors returned by the EIG command are unnormalized. It is usually convenient to
work with normalized eigenmodes. There are many ways of normalizing the eigenmodes.
Often we use “power normalization”, where the eigenmode ( Em , H m ) is normalized
according to 1

ωµ0 ∫ ( E*m (r⊥ ) × H m (r⊥ ) ) ⋅ zdr⊥ =

1, (20)

where z is the unit vector along the z axis. It is convenient to introduce the scaled magnetic
field H'm defined by

H'm (r ) = ωµ0H m (r ) (21)

such that the LHS of (20) becomes

1
The power normalization does not fix the phase of the field. It is thus possible to multiply the normalized
eigenmode E and H fields with an arbitrary phase eiφ. You should keep this in mind when comparing your
calculated field profile to that in Fig. 2.
 ∫ (E (r⊥ ) × H'm (r⊥ ) ) ⋅ zdr⊥ .
*
m (22)

To normalize the eigenmodes, we thus need to compute the magnetic field. This can be
accomplished by using Maxwell’s equation for non-magnetic materials
∇ × Em (r ) = iωµ0H m (r ) . (23)

For our choice of electric field polarization along the y axis, (21) and (23) lead to
H x' ,m = − β m E y ,m (24)
−i β m ∂ x E y , m .
H z' ,m = (25)

Only the x component of the magnetic field is needed to evaluate (22) which becomes
x=Λ x=Λ
2
∫ ∫ β E ( x ) dx ,
* '
y ,m x ,m m y ,m − E ( x)H ( x )dx = (26)
=x 0= x 0

where we have omitted the integration along the y axis due to the 2D nature of the geometry.
Insertion of (11) into (26) and carrying out the integration leads to
x= Λ
β m ∑ unm Λ ≡ C .
2
∫ E *y ,m ( x ) H=
'
− x ,m ( x ) dx (27)
x =0 n

The normalization of the eigenvector um is then performed by performing the substitution

1
um → um . (28)
C
Calculate the mode profile for the first eigenmode and the propagation coefficients for the
three first eigenmodes in layer 2, that is those having the three largest Re( β 2 ) values. Use Λ
= 400 nm, f = 0.32, λ = 633 nm, θ = 0°, ε1 = 1 and ε 2 = 1.5. The filling factor f is defined
in the figure caption of Fig. 1.

The normalized mode profile for the first mode and the three first propagation constants are
given below.
β 2 = 148.1911 /µm
Table 1
2
1.6
Refractive index
1.4 Re(Ey) (a.u.)
1.2 Im(Ey) (a.u.) Mode β 2 (/µm2)
1

0.8
1 148.19
0.6

0.4
2 -99.07
0.2

-0.2
3 -124.72
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
x (µm)

Fig. 2: Mode profile of the first eigenmode.

 Part 2. Reflection and transmission at an interface

Having computed the eigenmodes and propagation constants in the three layers, we now
consider the two-layer geometry illustrated in Fig. 3. The interface is illuminated from below,
and light is partly reflected and partly transmitted at the interface.

The field in this geometry is given by

m M=m M
=E ( x, z ) ∑a
m 1=
u ( x ) exp(i β ( z − z )) +
1;m 1;m 1;m bot
m 1
∑b u ( x ) exp( −i β1;m z )
1;m 1;m (z < 0) (29)
m=M
E ( x, z ) = ∑a
m =1
u ( x ) exp(i β 2;m z ) ,
2;m 2;m (z > 0) (30)

z = ztop

2 Λ a
a2 t12
z=0
a1́ b1
1 θ r12
z = zbot a1
x=0
Fig. 3: The two-layer interface illuminated by a plane wave from below.

where a1;m , b1;m and a2;m are the expansion coefficients of the incoming, reflected and
'
transmitted fields respectively. For the layer 1 we define the coefficients a1;m by
a1;' m = p1;m a1;m , where p=
1;m exp( −i β1;m z bot ) . The coefficients a1; j describes the field at z =
'
zbot, whereas the coefficient a1;m is ''facing the interface'' and describes the field at z = 0. This
notation is introduced to ensure numerical stability when plotting the field. The coordinates
zbot and ztop refer to the bottom and the top of the domain chosen for the field plot.

We now choose to illuminate the interface using a single eigenmode of index l by setting
a1;m = δ lm , and our goal is to compute the coefficients b1;m and a2;m . To this purpose we define
the reflection and transmission matrices r12 and t12 defined by b1;m = r12;mn a1;' n and
a2;m = t12;mn a1;' n . The reflection and transmission matrices can be computed using the boundary
conditions at the interface stating that the tangential parts of the electric and magnetic fields
should be continuous. Continuity of the electric field leads to the requirement
=m M=m M

1;l
=
E ( x) +
12;ml 1;m
m 1=
∑r
m 1
12;ml ∑ t E2;m ( x ) .
E ( x) = (31)

Now, for this particular 2D geometry illuminated with S-polarized light, continuity of the
magnetic field leads to the requirement of continuity of the derivative with respect to z of the
electric field such that
 =m M=m M

1;l 1;l
=
β E ( x) − ∑r
12;ml 1;m 1;m
m 1= m 1
12;ml ∑ t β 2;m E2;m ( x ) .
β E ( x) = (32)

In the FMM, we can represent the eigenmode profiles simply using their expansion vectors,
such that (31)-(32) become
=m M=m M

=
1;l u +
m 1=
∑u
1;m 12;ml r
m 1
∑u t
=
2;m 12;ml (33)
= m M= m M

1;l l
=
u β
(1)
−
m 1=
∑u β r
1;m 1;m 12;ml
m 1
∑u β t .
=
2;m 2;m 12;ml (34)

The matrix form of these equations is

u1 + u1 r12 =
u2 t12 (35)
u1 β1 − u1 β1 r12 =
u2 β 2 t12 , (36)

where uq is the matrix with columns uq;m and β q is the diagonal matrix having the
propagation constants β q;m along its diagonal. Solving (35)-(36) for r12 and t12 leads to

( ( ) ( ))
−1
=t12 2 u1 \ u2 + u1 β1 \ u2 β 2 (37)

u \ u − (u β ) \ (u β )) t
2(
1
=r12 1 2 1 1 2 2 12 , (38)

()
−1
where \ denotes matrix division, that is A \ B = A B.

Using the same parameters from the previous part, calculate and plot the absolute value of
the field generated by launching light with normal incidence θ = 0° towards the interface
sketched in Fig. 3. The field is shown below.

Fig. 4: Field generated by illuminating the interface with normal incidence light from below.

The reflection and transmission matrices r21 and t21 corresponding to illumination from
above may be computed simply by inverting the indices in (37)-(38).
 Part 3. Full three-layer geometry

We are now ready to treat the full grating geometry illustrated in Fig. 1. As before we set
a1;m = δ lm and the fields are then given by
m=M
=E ( x, z ) a1;m E1;m ( x ) exp(i β1;m ( z − zbot )) + ∑b
m =1
1;m E1;m ( x ) exp( −i β1;m z ) (z < 0) (39)
=m M=m M

=
=E ( x, z )
m 1=
2;m 2;m 2;m
m 1
∑a E ( x ) exp(i β z) + ∑b 2;m E2;m ( x ) exp( −i β 2;m ( z − L)) (0 < z < L) (40)
m=M
=E ( x, z ) ∑a
m =1
3;m E3;m ( x ) exp(i β 3;m ( z − L)) (L < z) (41)

' '
For the layer 2 we define coefficients a2;m and b2;m by a2;' m = p2;m a2;m and b2;' m = p2;mb2;m ,
where p2;m = exp(i β 2;m L) . The coefficients a2; j and b2; j can be thought of as ''having their
' '
back against the interface,'' where the coefficients a2;m and b2;m are ''facing the interface.''

Armed with the formalism in Part 2, we can, under the assumption that the other interface
doesn’t exist, calculate the reflection and transmission matrices for the interface at z = 0 and
for the interface at z = L. However, these matrices are not sufficient to describe light
propagation in the three layer geometry. To do this, we also need the scattering reflection and
transmission matrices for the entire structure.

Referring to Fig. 5, we now consider the reflection of light by the three-layer geometry. The
light of mode l travels forward in layer 1. At the interface it is partly reflected back into layer
1 with reflection described by the matrix r12 . However, part of the light is also transmitted
into layer 2. It propagates towards the interface between layer 2 and 3 and is partly reflected.
It then travels backwards towards the first interface and is partly transmitted at the interface
into layer 1. This journey can be described by the matrix t21 p2 r23 p2 t12 , where p2 is the
propagation matrix for the layer 2; a diagonal matrix consisting of the elements p2;m .
This journey consisting of one round-trip inside layer 2 adds an additional reflection to that
described by r12 . There is, however, also contributions from light making two, three etc.

Fig. 5: The reflection and transmission at the two interfaces and the multiple scattering of light inside the
central layer are illustrated.
round-trips in layer 2. The scattering reflection matrix r13 describing reflection at the
interface between layer 1 and 2 taking into account multiple reflections inside layer 2 is thus
given by

( ) ( )
2
r12 + t21 p2 r23 p2 t12 + t21 p2 r23 p2 r21 p2 r23 p2 t12 + t21 p2 r23 p2 r21 p2 r23 p2 t12 + ...,
r13 = (42)

where the product r21 p2 r23 p2 describes a round trip in layer 2. We now employ the relation
for a geometric series I + a + a 2 + ... = ( I − a ) −1 , where I is the identity matrix. This
relation holds for matrices with absolute eigenvalues below unity. The defining relation
becomes

( )
−1
r13 =
r12 + t21 p2 r23 p2 I − r21 p2 r23 p2 t12 . (43)

Again considering the multiple reflections in layer 2, one can in a similar manner derive the
scattering transmission matrix t13 describing transmission of light from layer 1 to layer 3. It
is given by

( )
−1
=t13 t23 p2 I − r21 p2 r23 p2 t12 (44)

The relations b1 = r13 a1' and a3 = t13 a1' give the fields in layers 1 and 3. In the central layer 2,
the field is determined from the expansion vectors a2 and b2 describing the forward and
backward traveling parts of the field. The vector a2 includes the contribution t12 a1' and its
complete expression becomes

( ) (t a ) ,
−1
'
a=
2 I − r21 p2 r23 p2 12 1 (45)

where the inverse matrix in a similar way as in (43) takes into account the effects of multiple
round trips of light inside layer 2. The corresponding expression for b2 becomes

( ) (r )
−1
b=
2 I − r23 p2 r21 p2 23 p2 t12 a1' , (46)

where the multiple reflections are again taken into account. The scattering matrices (43)-(44)
and (45)-(46) for the fields in the central layer completely describe the field in the entire
three-layer geometry.

Calculate the field generated by launching light with normal incidence θ = 0° towards the full
grating geometry interface from Fig. 1 using the parameters from the previous part and L =
385 nm. Plot the absolute value of the fields for two cases, Λ = 400 nm and Λ = 800 nm while
keeping the filling fraction f = 0.32 fixed.

The field in the two cases is shown below.

 Fig. 6: Field generated by illuminating the grating with normal incidence light from below.

How do the transmitted fields differ in the two cases?

Finally, we will consider the reflection of the incoming wave. The modal reflectivity of the
2
first mode is given by r13;11 , that is the absolute square of the (1,1) element of the total
reflection matrix r13 .

Calculate the modal reflectivity for values of L ranging from 0 to 600 nm for Λ = 400 nm and
Λ = 800 nm using f = 0.32. Compare your result with Fig. 7.

Fig. 7: Reflectivity of the grating for f = 0.32.

What is the reflectivity for L = 385 nm and Λ = 400 nm?

The reflectivity at L = 385 nm is lower for Λ = 400 nm than for Λ = 800 nm. How can this be
observed in the field plots in Fig. 6?