EM fields and

nanosystems for
biomedical
applications
Liberti
Micaela
&
Francesca Apollonio
Department of Information Engineering,
Electronics and Telecommunications
Sapienza, University of Rome

(liberti@diet.uniroma1.it)

Electromagnetic problem
time domain

frequency domain

Maxwell Equations

B
E = + J mi
t
D
H =
+ J + Ji
t

E = j B+J mi
H = j D + J+J i
Constitutive Relations

D ( r,t ) = t t ' E r,t ' dt '

D ( r, ) = ( ) E ( r, )

D ( r,t ) = (t ) E (t )

c ( ) = ( ) +
j

) ( )

J (r, t ) = E (r, t )

Boundary Conditions
(

n D+ D = S
normal component of Induc0on eld
has a discon0nuity equal to
supercial charge density

n B B
+

)= 0

Ht Ht = J S n
+

Et Et = 0

normal component of Magne0c Induc0on eld is

con0nuos
tangen0al magne0c eld has a discon0nuity equal
to supercial current density /2 rotated on the
tangent plane
tangen0al component of Electric eld is con0nuos

Discontinuity surface: lamina

h

q = d = limh >0

dhdS = limh >0 dh dS = S dS

S h
S
h
S

i = J d = limh >0

J dhdS = limh >0 J dh dS = J S dS

S h
S
h
S

h >0

J dhdS = limh >0 J dh dS = J S dS

h

h

J dS = lim

ABCD

S = limh >0 dh
h

supercial charge density (C/m2)

i=

J S = limh >0 J dh

supercial current density (A/m)

and J should be zero

outside of the foil and
innitely intense in the foil .
This trend can be represented
as a -Dirac -> we introduce
the S and JS

Condizioni al Contorno
Condizioni al Contorno
Condizioni al Contorno

Boundary Conditions
1

0
0

0
N-1
0 0

0.5

1
N n1
1

0.5

N-1
r

=
N-1

N nPEC
n
1 Nx
1

0.5

For Dirichlet boundary condi0ons, the value of F is specied on the

Condizioni
di Dirichlet
boundary. We therefore sa0sfy
the equa0on
at all internal points only.
Condizioni di Dirichlet
Condizioni di Dirichlet

i 0

0, i 1

i 0
i 0

0, i 1
0
0, i 1I 0 0

0, I N

I 0
0 I 0

0, I N
0, I N

Condizioni di Neumann
For
Neumann, di
the
boundary condi0on
is on di
the
deriva0ve of F. This makes
it acos
liDle x dicult. 2T x, y
T x, y
Condizioni
Neumann
Condizioni
Neumann
t

di
dx

x 0

di
0,
dx

di
di
0dx x
dx

x 1

0
x 0

di
0,
di
0, Idx1 x
dx

0
I 0 0, I N

I 1
I 0 , I N 1
1 I I1 N I 0 , I N 1

x 1

Condizione di (i,j+1)
Impedenza (per es., sul carico di una linea di trasmi
(i,j)
Condizione di Impedenza (per
Condizione
es., sul carico
di Impedenza
di una linea
(per di
es.,
trasmissione)
sul carico di una linea di trasm

v 1
i 1

v 1
f
= 0 at1 (di,j
11
i )i v(i-1,j)
1
ZxL
Yp dx

i i11

ZL
ZZLL

x 1

interior

I N 1
1 I N
1 di
1
(i+1,j)
I N 1 IZN
1 1 1 I N
I N 1
L
1
1
d
i
Yp
Yp dx x 1 i 1
ZL h
Z L unknown!
h
Yp Y
dxp x 1 i 1 h
I NYp
(i,j-1)
exterior

Solving EM Field Problems

Find electromagnetic field and/or source
functions such that they
obey Maxwells equations,
satisfy all boundary conditions,
satisfy all interface and material
conditions,
satisfy all excitation conditions.
(in both time and space, or at one freq. in space)

The solution is generally unique (Analysis)

Analitical Solutions o Numerical Solutions?

Soluzioni Analitiche o Soluzioni Numeriche?

T x, y

cos

PEC

2
t

T x, y

k t2T x , y

Classic Electromagnetic Solution

Mathematical
Formulation

Maxwells
Equations

Boundary
Conditions

Analytical
Preprocessing

Analytical
Models

Discretization

Computer
Programs

Material
Properties

Problem-dependent

User Data
Computation
User Interface
Postprocessing

Results

Examples:
Closed-form expressions for microstrip;
spectral domain program for
coplanar waveguides;
mode-matching program for
waveguide filters;

Numerical Electromagnetic Solution

FD-TD
Numerical
Formulation
and
Discretization

Maxwells
Equations

TLM
Huygens
Principle
Numerical
Model
Computer
Programs

FEM
Variational
Principle

Problem-independent

Boundary Conditions
Material Properties

Computation
Examples:
Finite Element Frequency

User Interface
Postprocessing

Results

Domain Solver;
TLM or FDTD EM Simulator

Numerical Electromagne0cs

Theoretical
Analysis

Laboratory
Experiments

Numerical
Modeling

The Three Pillars of

Electromagnetic
Microwave and
Lightwave
Engineering 10

Electromagnetic Simulators
An Electromagnetic Simulator is a field
modeling tool that:
solves electromagnetic field problems by numerical
analysis;
extracts engineering parameters from the field
solution and visualize fields and parameters;
allows design by means of analysis combined with
optimization.

The field solver engine employs one or

several numerical methods obtained
through the practice of Numerical
Electromagnetics.

Field-Solving Methods
Methods for solving Maxwells Equations:
Analytical Methods
Exact explicit solutions (only a few ideal cases)

Semi-Analytical Methods
Explicit solutions requiring final numerical evaluation

Numerical Methods
Differential or integral equations are transformed into
matrix equations by projective approximations and solved
iteratively or by matrix inversion

IntegrazioneIntegration
Numerica 1/2
Numerical
f x

f a

Everyintegrale
integraldefinito
operation
on [a,b]
can bericondotto
reformulated
tointegrale
an
Ogni
su [a,b]
pu essere
ad un
integral on [0,1]
variable
position
sullintervallo
[0,1]with
con proper
un opportuno
cambiamento
di variabili:
b

f x dx
a

f a
0

a d

F
0

Integrazione Numerica 2/2

Numerical Integration
Lintegrale definito interpretabile come area del rettangoloide colorato in figura,
figura
approssimabile mediante il procedimento di esaustione:
1

lk F

k 1

Formula di quadratura
N

lk

lk

(N t che
(Nota
h
k

1)

Allaumentare del numero di punti di quadratura N, aumenta laccuratezza del

Increasing N improve accuracy but worsen the computational cost
risultato ma aumenta il costo computazionale dell
risultato,
delloperazione
operazione (devo calcolare N
volte lintegrando e il codice diventa lento).
the function
to be integrated
constant
enough
Se la Iffunzione
integranda
costante,is basta
un one
solo rectangle
rettangolois per
ottenere il
risultato
corretto.
generale,
pi la funzione
varia
velocemente,
In general
theInfaster
changes,
the bigger
number
of points un maggior
numero di punti richiesto per una buona approssimazione.

quadratura
per for
p
minimizzare
numero di
Lanalisi
numerica
studiastudies
formule
di q
Numerical
Analysis
quadrature
formulas
minimal Niland
punti maximal
N per una
precisione
richiesta del
risultato
(formuleetc)
di Newton-Cotes,
accuracy
(Newton-Cotes,
Gaussian
quadrature
formule di quadratura gaussiana, quadratura di funzioni singolari).

Numerical
DerivazioneDerivative
Numerica
Given
derivative
x0 can
approximated
incremental
D t f(x)
Data
una its
ffunzione
i
f( ) ciiinaspettiamo
f(x),
tti be che
h la
l sua derivata
d i with
t nelthe
l punto
t x0 possa
ratio
between
x0 andtramite
x0 + h il rapporto incrementale fra x0 e x0+h:
essere
approssimata

f x0

f x0

f x0

Lerrore
commesso
essere valutato
tramite
la formula
The error
can be pu
evaluated
with Taylor
at first
orderdi Taylor troncata al
primordine
prim
ordine (ammettendo che la funzione abbia derivate seconde continue):

f x0
f x0

f x0

f x0

h f x0
f x0

x0

x0

h2f

O h ,

per h

Numerical
Derivative
Derivazione Numerica: Formule al Secondo Ordine 1/2
f x

f x

f x

O h ,

per h

How to improve O(h) estimation?

Come ottenere stime pi accurate di O h ?

f x

f x

f x

f x

hf x

h2
f
2

O h3

hf x

h2
f
2

O h3

Sottraendo:

f x

f x
2h

f x

O h2 ,

per h

Formula al secondo ordine per la derivata nel punto x intermedio fra xh e x+h.

Numerical
Derivative
Derivazione
Numerica: Formule
al Secondo Ordine 2/2
f x

f x

f x

f x

f x

h
h

f x

O h

2h

O h2

Esatta
O(h)

O(h2)

h x x

Analogamente, per la derivata seconda vale la seguente:

f x

2f x
h

f x

O h2 ,

per h

What have all Methods in Common ?

1. In all methods, the unknown solution is expressed as a
sum of known functions (expansion functions or basis
functions).
2. The weight (coefficient) of each expansion function is
determined for best fit.

What distinguishes them?

the electromagnetic quantity approximated, the

expansion functions used,

the strategy employed for determining the

coefficients of the expansion functions.

Options when choosing the simulation method

C 621 Lecture 1

Fall 2005

Entire domain expansion

versus
Space discrete methods

Full-Domain Expansion Functions

1.5

Function to be approximated
Approximation

1
0.5

Mode Matching,
Spectral Domain

Expansion functions

0
-0.5
-1
-1.5

Proposed by Joseph Fourier (1768-1830)

Options when choosing the simulation method

Entire domain expansion
versus
Space discrete methods

Boundary discretization (BEM)

versus
Volume discretization

Why space-discrete ?
It is universal in handling complicated
geometries and complicated inhomogeneous
and arbitrarily shaped materials (limits to
minimum user involvement)
It allows to differentiate easily dissipated power
and physical parameters in various spots of the
structure
Conclusion: practically all commercial software
packages available now use discrete methods,
e.g. FEM, FDTD,TLM or BEM (with

MoM formulation)

Subsectional
Expansion
Functions
Subsectional Expansion Functions
F(x)

Finite Differences

F(x)

Finite Elements

X
F(x)

Point Matching

F(x)

Rooftop Functions

x
Wolfgang J. R. Hoefer

ELEC 621 Numerical Techniques in Electromagnetics

x
Lecture 1 - 23

Options when choosing the simulation

method
Frequency domain analysis
versus
Time domain analysis

Note: Direct discretization of time gives simple explicit algebraic

relations between fields in consecutive instants. In frequency domain
more sophisticated methods (e.g. Method of Moments, Perturbation
Method, or Spectral Domain Method) are needed to formulate linear
implicit algebraic problem (requiring inversion of a large matrix).

Options when choosing the simulation

Time or frequency domain?
method
Time domain
Time
domainanalysis:
analysis:
explicit algebraic
relations between fields in
consecutive instants.

explicit algebraic relations between fields in consecutive instants.

Frequency domain analysis:

implicit algebraic relations between

complex amplitudes of fields.

Frequency domain analysis:

2 2
0

2 2
0

h 4

h 4
1

1
1
h 4

2 2
0

2 2
0

2 2
0

h 4

h 4
1

1
1
2 2
4
0h
1

0
0

1
1
1
1

2 2
0

h 4

h 4
0

1
1
1

1
0
h 4

0
g

2 2
0

2 2
0

2 2
0

2 2
0

h 4

h 4
1

1
h 4

2 2
0

0
0

10

11
12

0
0

implicit algebraic relations between complex amplitudes of fields.

Why model in the frequency domain?

Most microwave engineers are more familiar with
FD concepts than with TD concepts,
Frequency domain simulations are steady- state,
Complex notation is elegant and efficient,
Specifications are traditionally formulated in the
FD (S-Parameters, loss tangent, dispersion),
Time domain information can be obtained by
inverse Fourier Transform (Causality issues!),
Dispersive materials and boundaries are easily
described by frequency-dependent parameters.

Why model in the time domain?

Time domain simulations are life-like
Virtual experiments are set up as in the lab
(Source, reference planes, output probes)
Cause and effect can be distinguished
One simulation can cover a wide bandwidth
Transient phenomena can be simulated
Nonlinear behavior is modeled naturally
Dispersive materials and boundaries are
modeled in a more physical manner

time vs frequency

Panorama sui Metodi Numerici

Objective:
discretize
the integro-differential
equations
toi un
Obi i
Obiettivo:
di
discretizzare
i
l equazioni
le
i i integro-differenziali
i
diff
i li per trasformarle
f
l in
transform
them into
a system
of (usualmente...)
algebraic equations,
sistema di equazioni
algebrico,
risolubile
con metodi solvable
standard.
(usually) by standard methods
Formulazioni differenziali

Formulazioni integrali
Metodo dei Momenti
nel dominio del tempo

Tempo

FDTD
(Finite Differences Time Domain)

Differenze Finite
Frequenza
FEM, Elementi finiti
FEM
(Ansoft HFSS)
(Comsol Multiphysics)

FIT, Integrali finiti

nel dominio del tempo
(CST Microwave Studio)
Metodo dei Momenti
(Ansoft Designer, FEKO)
Metodo di Rayleigh-Ritz
Integrali finiti
nel dominio della frequenza

classification
Classification of numerical
methods for
electromagnetic modeling
Space
Time

Continuous
monochromatic
frequency domain (FD)
Discrete
probed pulses
time domain (TD)

Continuous:
expansion into
modes

MoM with mode

expansion

Discrete:
expansion into cells

BEM with MoM

FEM
FDFD

FDTD
MoMTD

TLM
FETD

Classification

Classification of Methods 2
1D Methods:

Fields and voltage/current vary in one space

dimension (Transmission Line Problems)
(

2D Methods:

Touchstone, Supercompact, SPICE

. )

Fields and currents vary in two space

dimensions (Cross-section problems,
TEn0 waveguide problems)
(

.FEM-2D, MEFiSTo-2D

.)

2 1/2 D Methods: Fields vary in three space dimensions,

currents vary in two space dimensions
(Planar multilayer circuits)
(

3D Methods:

.) frequency domain

Fields and currents vary in three space

dimensions (General propagation, scattering and radiation problems)
(

Wolfgang J. R. Hoefer

Sonnet, Momentum, Ensemble

HFSS, FEGS, MEFiSTo-3D, CST, Quickwave..)

ELEC 621 Numerical Techniques in Electromagnetics

Lecture 1 - 19

FDTD Method

FDTD Method (finite difference time domain)

Time Domain
traditional method for RF dosimetry

Solving Scheme
1. Discretization of the domain of analysis in
a grid of nodes (space, time)
2. Discretization of Maxwell's equations in
the time domain (from differential
equations to finite difference equations)
3. Solving a system of difference equations
on the discretized analysis domain

Procedure: step 1
E
H = E +
t
H
E =
t

curl of E along x

H x 1 # E y Ez &
= %

(
t
$ z y '

time and space discretization

Procedure: step 2 (discretization)

IN SPACE
BASIC CELL
Ez

Hx

Hy

It contains information
on the location, on the
dielectric constant and
conductivity

IN TIME
n

Ey

(i,j,k)
Ex

Hz

t = n t

Hn-1/2 En Hn+1/2 En+1

Derivazione
Numerica:
al Secondo
Ordine 2/2
Procedure:
stepFormule
3 (finite
differences)

f x

f x

f x

f x

f x

h
h

f x

O h

2h

O h2

Esatta
O(h)

O(h2)

h x x

Analogamente, per la derivata seconda vale la seguente:

f x

2f x
h

f x

O h2 ,

per h

discretized expression: Hx
H x 1 & E y E z #
!!
= $$

t
% z
y "

y
x
Ez

Hx at instant n + 1/2 depends on Hx in the

same cell at the time n-1/2 and on Ey and Ez
in the instant n in the same cell and in
adjacent cells.
The system is solved iteratively; when a
stationary state is observed for E and H in all
cells, the analysis is complete.

E y(i,j+1/2,k+1)

(i,j,k+1)

Hx

Hy

Ey

(i,j,k)
Ex

Hz

H xn +1/ 2 (i, j + 1 / 2, k + 1 / 2) H xn 1/ 2 (i, j + 1 / 2, k + 1 / 2)

=
t
n
n
1 & E y (i, j + 1 / 2, k + 1) E y (i, j + 1 / 2, k ) E zn (i, j + 1, k + 1 / 2) E zn (i, j , k + 1 / 2) #
=

$
!
0 $%
z
y
!"

Stability and accuracy

Is there a solution?
stability: the solution does not
diverge
Is what we are looking for?
accuracy: how much the solution
comes close to the real one

Stability
Algorith must converge:
If: n = errore al passo n
then:

| n+1 | | n |

STABILITY CONDITION 4 YEE ALGORITHM

1
t
c

1
1
1
1
+ 2+ 2
2
x
y
z

Accuracy

Accuracy indicates how

the found solution is close to the real one
= max (x, y, z) <<

FDTD meshes & boundary approximation

Stair-case

Nonorthogonal

Variable mesh

Conformal

Subgridding

Basic errors of FDTD method

1. Errors of modelling wave
propagation over large,
homogeneous subregions:
-

dispersion errors on uniform meshes,

numerical reflection from mesh nonuniformity.

2. Errors of shape approximation:

-

dielectric interfaces,

metal boundaries.

3. Errors of finite computing time

Pro and contr

Low memory occupation in the PC (linear with
cell number);
It is not necessary to invert matrices;
Opportunity to study complex scattering elements;
Ability to consider both sources in the far field in
the near field.
- Closing of the domain;
- Limitations in frequency.

Difficulties in FDTD and how they are fought

Explicit algorithm makes it less flexible in meshing

remedy: conformal meshing &/or subgridding.
Difficult to apply with dispersive media
remedy: use special dispersive media models
Slowly converging with high-Q structures
remedy: use special signal processing
techniques (e.g. Prony method)