zyxwvu
zyxwvutsrqponm
zyxwvuts
zyxwvu
zy
zyxwvutsrqponmlkjihgf
IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, VOL. 12, NO. IO, OCTOBER 1993
1557
Numerical Simulation of Optical Devices
Maria Cristina Vecchi, Massimo Rudan, Senior Member, IEEE, and Giovanni Soncini
Abstract-The description of the optical-generation phenomena has been incorporated in the semiconductor-deviceanalysis
program H F I E L D S, keeping the structural and geometrical
flexibility of the original code. To this purpose a substantial
effort has been devoted to the software implementation. The
latter required the introduction of a number of optical windows
and interleaved material layers through which a radiation with
arbitrary spectrum, incidence angle, and polarization state enters the crystal. The corresponding generation rate at each node
of the discretization grid is then evaluated by adding up the
contribution of each monochromatic component of the impinging radiation. A preliminary report has been given in [l].The
code equipped with this new capability makes the description
of realistic semiconductor optical sensors feasible, as demonstrated by several examples given in the paper. As an additional
feature, the black-body radiation has been incorporated into
the code to allow for the simulation of solar cells.
After a brief review of the underlying physics, given in Section I, the software implementation is described in Section I1
and a number of examples are illustrated in Section 111.
with
p =
q(p -n
+ NA
-
NA),
P = 4Nom
(1.6)
(1.7)
+ qD, grad n ,
(1.4)
respectively. The remaining equations hold in the semiconductor domains only. The symbols in (1 .1)-( 1.7) have
the customary meaning of the semiconductor-device theory. Hence, only the form of the optical terms G,OP',Gip'
will be discussed here in some detail. It is worth noting
that, for the sake of completeness, different symbols for
the electron- and hole-generation/recombination terms
have been used in the continuity equations. In fact, if the
concentration of traps in the band gap is large and a transient condition holds, it would turn out G:P' # Gip' and
U,, f U,. As a consequence, it would be necessary to add
to (1.2), (1.3) a continuity equation for the traps, and
modify the RHS (1.6) of Poisson's equation accordingly.
However, in the practical cases the concentration of traps
is low enough to avoid these complicacies (see also the
discussion in Section 1.3).
The numerical solution of system (1. l ) , (1.2), (1.3) is
carried out in H F I E L D S by discretizing the equations
on a two-dimensional, triangular grid [2]. This transforms
the system of PDE's in a (non-linear) algebraic system
whose unknowns are the values of cp, n , and p at the grid
nodes. To incorporate the optical-generation effects it is
then necessary to evaluate GZpt, Gip' at each grid node
belonging to the semiconductor. This is done in two steps:
in the first one, the intensity and the direction of the electromagnetic wave which actually enters the semiconductor crystal are evaluated. This calculation accounts for the
characteristics of the impinging radiation and for the geometrical structure and physical nature of the material layers interleaved between the radiation source and the crystal. In the second step, the carrier generation produced by
radiation at each grid node is evaluated.
Jp = - q p pP grad P - 4Dp grad P ,
(1.5)
1.2. Treatment of the Electro-Magnetic Radiation
zyxwvuts
zyxwvutsrq
zyxwvu
I. PHYSICAL
ASPECTS
1.I. Basic Equations
HE SYSTEM of (non-linear) PDE's constituting the
semiconductor-device model and incorporating the
optical-generation effects is made of Poisson's equation
T
-div
(E
grad p)
= p
(1.1)
and of the electron- and hole-continuity equations
an
--
at
1
-
div J , = GZP' - U,,,
4
supplemented by the drift-diffusion expression for the
electron and hole current densities
J , = -qp,,n grad p
and by the expression for the charge density. Eq. (1.1)
holds in the semiconductor and in the insulator domains
Manuscript received August 11, 1992; revised March 15, 1993. This
paper was recommended by Associate Editor J . White.
M. C. Vecchi and M. Rudan are with the Dipartimento di Elettronica,
Informatica e Sistemistica, Universita di Bologna, 40136 Bologna, Italy.
G. Soncini is with the Dipartimento di Ingegneria dei Materiali, Universita di Trento, Mesiano (Trento), Italy and IRST-Divisione Microelettronica, 38050 Povo (Trento), Italy.
IEEE Log Number 92 12217.
The material layers mentioned above are modeled using
the following simplifying assumptions. A stack of m - 2
layers of different materials is considered, separated by
plane and parallel interfaces. The stack is placed between
medium 1 (initial medium), in which the radiation'source
is located, and medium m (final medium), coinciding with
a semiconductor zone (Fig. 1). Further, it is assumed that
all the m media are optically linear, isotropic, and
homogeneous,
0278-0070/93$03.00 0 1993 IEEE
zyxwvutsrqponm
1 (initialmedium)
case) is taken into account by the complex refraction index ncm.
The calculation of the transmission coefficient 3 is carried out by introducing the transmission coefficients ~ T
and 3 T M associated with the transverse electric (TE) and
transverse magnetic (TM) waves, and expressing 3 as a
linear combination of the above [3]:
2
3
3
=
+
a 2 3 ~ b2&E.
~
E
(1.13)
The coefficients in (1.13) turn out to be
m-2
a 2
=
b2
=
cos2 P
1
++BB ~~ / / sin2
A ~P
A ~
(1.14)
9
zyxwvutsrqp
m-1
sin2 P
+ B ~ / cos2
A ~P
1+B ~ / A ~
(1.15)
zyxwvutsrqp
zyxwvut
zyxwvut
zyxwvutsrqpon
zyxwvutsrq
m (finalmedium)
Fig. 1 . Stack of rn material layers separated by plane and parallel
interfaces.
the initial and final media have an infinite extension
in the direction z normal to the interfaces, and
the electromagnetic wave is described by a superposition of plane and monochromatic waves.
Thanks to the above hypotheses, the refraction angle
0; in the final medium and the transmission coefficient 3
of the whole structure (i.e., the fraction of the electromagnetic intensity transmitted to the final medium) are
easily calculated for each monochromatic component of
the impinging radiation. Quantities 0 ; and 3 turn out to
depend on the frequency v of the monochromatic component, on the initial angle of incidence O1 and polarization state of the impinging radiation, on the layers' thicknesses di, i = 2,
, m - 1 and on the (typically
complex) refraction indices of all the media involved, nck
= nk(1 + j l k ) , k = 1, * * * , m. Their expressions are
easily derived following e.g., [3]. The derivation is summarized in the following, starting with the final refraction
angle 0; whose expression is
9
for a specific polarization state of the incident wave, or
a 2 = b2 = 1/2
foranon-polarized wave. In (1.14), (1.15), 0 5 B / A 5
1 is the ratio between the axes of the polarization ellipsis
of the incident wave, and 0 5 P I7r/2 is the angle between one of the axes and the incidence plane. The transmission coefficients &,, 3 T M , in turn, are given by
3TE
=
vl Re { 1/vcm(sin e,
. ItTE12,
1
~ T M= - Re
TI
{q,(sin
sin
8, sin 0 ;
e; + cos e,
=
qn, (cos 6 - 3;, sin 6)/Q,
sin 0; = nl sin e,/Q.
(1.8)
(1.9)
In (1.8) and (1.9), nl is the real part of the refraction
index of the initial medium, while nm, S;, are the real part
and the imaginary term of the refraction index of the final
medium. Quantities q, 6, and Q in (1.8) are, in turn, expressed by means of the relationships
q exp ( j 6) = cos
e,
= J1
- sin2 e,,
(1.10)
el,
(1.11)
=
e;)}
+ COS 8,
COS
e;)}
(1.18)
The real quantity q l and the complex one vcmin (1.17),
(1.18) are the intrinsic impedances of the initial and final
medium, respectively; in turn, tTE is the (complex) ratio
between the amplitude of the electric field transmitted to
the final medium and that of the incident one for a TE
wave, and tTM is the corresponding quantity for a TM
wave. Finally, using the speed of light in vacuum and the
magnetic permeabilities of the media to define the quantities
n1
PI = -cos
el,
(1.19)
CP 1
(1.20)
zyxwvutsrq
sin 0, = n l / n c , sin
and
COS
(1.17)
- -
cos 0 ;
(1.16)
n: sin2 el
+ q2ni(cos 6 - rrnsin s ) ~ .
(1.21)
qi
(1.12)
It is seen that angle 0 ; depends only on the electromagnetic properties of the initial and final medium, and
that the absorption in the final medium (silicon in this
CPi
= - cos
ei,
(1.22)
nci
ei = Ji'TZ&,
sin ei = n l / n c i sin el,
cos
(1.23)
(1.24)
zyxwvutsrqponmlk
zyxwvutsrqponmlkj
zyxwvutsrqponm
zyxwvutsrqpo
zyxwvutsrqp
1559
VECCHI er al.: NUMERICAL SIMULATION OF OPTICAL DEVICES
i = 2,
, m, the expressions for q l , qcm,tTE,
found to be
Ill
=
tTM
CP1
are
(1.25)
-3
nl
(1.26)
2Pl
~ T E=
(mil
+ m 1 2 ~ r n ) ~+l
+ m22~m)'
(m21
24 1
~ T M=
(mil
+ mi2qrn)ql +
+
(mil
(1.27)
(1.28)
mhqm)'
zyxwvuts
zyxwvut
The quantities that are now left to define in (1.27) are
the elements mil, mI2,m21,m22of the 2 X 2 characteristic
matrix % associated to a stack of homogeneous layers for
a TE wave [3]. Similarly, in (1.28), for the elements
m i l , mi2, m i l , miz of the corresponding TM-wave characteristic matrix %'. By remembering that d;, i = 2,
*
* , m - 1 are the thicknesses of the intermediate layers, the expressions of %, 3X'are obtained as the product
of the TE and TM characteristic matrices of the individual
layers
-
-
311. = %,(d,)
zyxwvutsrqponmlkjihgfed
%3(d3)
- *
*
*
Xm-l(dm-,)
(1.29)
%'
=
%$(d,)
*
Fig. 2. Transmission coefficient for an Air-Si02-Si structure. The curves
correspond to four different oxide thicknesses.
%$(d3)
*
*
%;-l(dm-l),
(1.30)
where
cos (yidj)
-j/pj sin (yidi)
where
3Z'" are the intensities of the incident and
transmitted wave, respectively, and hv is the photon energy. Different attenuation mechanisms are possible for
an electro-magnetic wave crossing a doped semiconductor, namely
1) excitation of lattice vibrations,
2) excitation of free electrons (holes) between states
belonging to the same band,
3) excitation of free electrons (holes) between different
conduction (valence) bands,
4) excitation of electrons from the valence band to
states belonging to the band-gap, or from the latters
to the conduction band, and
5) excitation of electrons from the valence to the conduction band.
Of the above mechanism, only the last two directly
contribute to the generation terms in (1.2), (1.3), hence
to the photocurrent. The other mechanisms contribute to
cos (yidi) - j / q i sin (yidi)
reduce the intensity of the electromagnetic wave, hence
3 2 ; (d;) =
to
decrease the number of photons available for generat- j q j sin (yidj)
cos (yidj)
ing free carriers in the semiconductor. Due to this, they
and
have an indirect influence on the generation terms as well.
However,
theoretical calculations and experimental evi2av
y; = -nci cos Oi.
(1.33) dence indicate that, for non-degenerate semiconductors in
C
the visible range the last mechanism, i.e., the excitation
An example of application of (1.13) is given in Fig. 2, of electrons from the valence to the conduction band,
where the transmission coefficient 3 of an air-Si02-Si largely dominates [4]. For degenerate semiconductors, on
structure is plotted as a function of the wavelength at four the contrary, the excitation of free electrons or holes (see
different oxide thicknesses, for a non-polarized wave. It point 3 above) may become significant. This mechanism,
is seen that the behavior of 3 is such that the distance in turn, makes the absorption to depend also on the carrier
between the peaks decreases as the oxide thickness in- spatial distribution, hence on the doping distribution itcreases.
self. From the viewpoint of the semiconductor-device
model, the equations can still be solved at the additional
1.3. Optical-Generation Rate
cost of taking into account the dependence of the opticalThe expression for 3 determined in the previous section generation term on the carrier concentration. For the sake
allows one to express the photon flux at the semiconductor of simplicity, heavy degeneracy has not been considered
in this paper. From this discussion it follows that one may
surface and for each monochromatic component as
adopt the simplifying assumption of keeping only the last
(1.34) mechanism (i.e., item no. 5 in the above list) in the computation, and neglecting the others. Hence, each photon
%;(dJ =
-jpi sin (yidi)
cos (yidi)
zyxwvuts
1560
zyxwvutsrqpo
zyxwvutsrq
zyxwvutsrqp
zyxwvuts
zyxwvutsrq
zyx
IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, VOL. 12, NO. 10, OCTOBER 1993
absorption produces one electron-hole pair, and the
expression of
GZP' = GpOPt =
GoPt
let07
SiliconalT=300K -
(1.35)
holds for each monochromatic component and for the total radiation as well. Furthermore, assuming that the excitation of an electron by a photon in an elementary layer
of thickness d t takes place independently of what happens
to all the other electrons, it follows that the decrease d+
in the photon flux across the elementary layer is proportional to the flux entering the layer and to d t . Hence
+
-d+(O = a+(t)d t ,
(1.36)
where a is the absorption coefficient for the band-to-band
transitions. Thanks to the simplifying assumption introduced above, the effects of the non-uniform doping and
free carriers can be neglected, and the absorption coefficient becomes independent of position and equals the intrinsic one. Then, for a given semiconductor, a turns out
to be a function of the radiation frequency v and lattice
temperature TL only [5]. The dependency of a on the frequency, in silicon at room temperature, is shown in Fig.
3 by way of example.
Due to (1.36), the photon absorption is expressed by
means of a Poisson statistical distribution in which the
mean penetration length is A = l / a . For each monochromatic component the photon flux at a given depth 4
in the semiconductor along the propagation direction turns
out to be
W
0
I=
($)2
27rh2
exp (hc/k,Th) - 1
dh
(1.42)
zyxwvutsrqp
zyxwvut
+"(E)
=
exp (--at),a = a ( v , TL). (1.37)
Since each photon absorption produces one electron-hole
pair, it follows
CP'(t) d t = -d+,(t),
whence
Fig. 3. Silicon absorption coefficient as a function of wavelength at
T = 300 K.
G:P' (4) = a+ovexp
(-at).
(1.38)
(1.39)
The total optical generation rate Gop'(t) is then found by
adding up the monochromatic contributions (1.39), viz
Gop'(t) =
GZp'(4).
(1.40)
As a final remark, we quote the relevant expressions describing a black-body radiation, which will be used in the
next paragraphs. The energy density per unit wavelength
of the black body is given by
u ( h ) = 87rhc
exp (hc/k,Th) - 1 '
(1.41)
where Tis the black-body temperature and the other symbols have the usual meaning. The solar spectrum can be
approximated by that of a black body at T = 5800 K. It
follows that the solar constant, i.e., the power impinging
on the unit area outside the earth's atmosphere, is given
where Rs = 7 X lo* m is the radius of the sun, RSE= 1.5
x 10" m is the sun-earth distance, and U = 5.652 X
W m-2 * K - 4 is the Stefan-Boltzmann constant. Letting
T = 5800 K in (1.42), it is found I = 135.3 m * W
cmP2. As mentioned above, this value pertains to the solar spectrum outside the atmosphere, and is referred to as
the "air mass zero" (AMO) condition [9].
11. SOFTWARE
IMPLEMENTATION
2. I . Illumination Windows
Optical generation has been implemented in such a way
as to allow the maximum geometrical and structural flexibility. To this purpose, it is worth mentioning that
H F I E L D S has the capability of managing a number of
different semiconductor zones for a given device, supplemented with the necessary interfaces toward insulating
zones, and different types of contacts, gates, and floating
gates. In this context, a "zone" is a two-dimensional
closed region whose physical properties are fully specified. For instance, in the simple case of an insulating zone
one must provide the insulator's permittivity and the fixed
charge stored in it. A semiconductor zone is specified in
a similar manner, but the physical properties to be provided are of course much more numerous. The maximum
number of semiconductor zones that can be dealt with in
each simulation is ten. The illumination conditions can be
prescribed independently for each of them, since the code
has been the capability of managing up to ten illumination
windows per semiconductor. The illumination window is
defined as the portion of the semiconductor boundary on
which a radiation impinges. Consistently with the hypoteses outlined in the previous section, each window is
a segment laying along the boundary. For its definition it
zyxwvu
VECCHI er al.: NUMERICAL SIMULATION OF OPTICAL DEVICES
zyxwvutsrq
1561
then suffices to provide the coordinates of the two edges,
which need not coincide with any of the grid nodes. Apart
from the coordinates, the parameters to be defined for each
window are the following.
(1.8) and (1.9); the possible occurrence of total refraction is checked at this stage,
evaluating the photon flux (Pop transmitted to the final
medium, using (1.13) and (1.34).
A flag indicating the material of which the final medium is made. This information will be used by
H F I E L D S to fetch the real and imaginary part of
the refraction index, and the absorption coefficient,
at each frequency. It is worth reminding here that the
final medium is assumed semi-infinite.
A similar flag indicating the material of which the
initial medium is made. The initial medium is assumed to be a semi-infinite perfect dielectric.
More flags indicating the existence and nature of
possible intermediate layers (up to six), which may
be absorbent or not. The thickness of each existing
layer must be specified as well (see (1.31) and (1.32).
The incidence angle, i.e., the angle between the direction of the impinging radiation and the normal to
the window.
A flag indicating the verse of propagation along the
prescribed direction.
A flag indicating if a black-body spectral distribution
is present. If this is the case, the black-body temperature must be provided as well.
More flags indicating the existence (besides that of
the black-body radiation) of up to twelve independent spectral components. The properties of the existing spectral components must than be specified as
follows:
The wavelength in vacuum and the intensity.
A flag indicating if the component is polarized or
not. If this is the case, the angle and polarization
ratio must be specified as well (see (1.14) and
(1.15)).
Then, a two-dimensional strip (Fig. 4) is constructed by
drawing two lines starting from the window’s edges and
entering the semiconductor in the O h direction. If node i
falls within the strip, and is placed in the correct half plane
with respect to the verse of the wave propagation then the
contribution of the monochromatic wave to the opticalgeneration rate at node i is evaluated by:
zyxwvutsrqpo
zyxwv
zyx
zyxw
zy
fetching the absorption coefficient of the final medium a ( ~TL)
, at the current frequency v,
determining the distance t j of node i from the window along the direction defined by the final refraction angle Oh,
evaluating the optical-generation rate due to the
monochromatic wave, using (1.39) with 4 = t i .
The procedure depicted above is nested in a double loop
scanning the monochromatic components defined for each
of the existing windows, and results in accumulating the
individual contributions into the total generation rate
G O p t (ti).
Finally, in order to scan all the nodes belonging
to semiconductors, the calculation is repeated for each
semiconductor zone present in the device structure.
2.3. Treatment of the Black-Body Radiation
If a black-body radiation is to be considered, its effects
are taken into account by H F I E L D S using (1.34),
(1.39), and (1.40). The differences between this case and
that related to the individually-defined spectral components, as far as the software implementation is concerned,
are the following: 1) all the relevant quantities are expressed in terms of wavelength instead of frequency, and
2) the intensity I ( ’ ) in (1.34) is not given as an input datum
but is evaluated by the code consistently with (1.42). To
this purpose, a finite wavelength-interval B = [Amin, A,],
is selected and then subdivided into N A smaller intervals
of equal amplitude AA. The band B includes the visible
part of the spectrum. In this way, the physical models and
the simplifying assumptions listed in the previous paragraph hold and, at the same time, most of the intensity
radiating from the sun is taken into account. The value of
NA,in turn, is taken large enough to attain a sufficient
accuracy in the calculation of the integral in (1.42). The
continuous spectrum in B is then replaced by a finite num,
ber of wavelengths A, = Amin (s - l)AA, s = 1,
N A , whose intensity is defined as
zyxwvutsrqpo
2.2 Evaluation of the Optical Generation
The effect of the user-defined wavelength, mentioned
in the previous section, on the total optical-generation rate
is found by adding up the monochromatic contributions.
In addition to that, a further contribution comes from the
definition of a black-body radiation, whose temperature is
imposed by the user. For each window, the black-body
radiation is then automatically partitioned into a number
of monochromatic components whose effects are then
added to those of the individually-defined wavelengths.
Let us now focus on the evaluation of the generation
rate at a specific node, say node i , in a given semiconductor zone. For each window defined over the zone, and
for each monochromatic wave associated to the window,
the number of photons entering the zone and the final refraction angle are determined. This is done by:
0
zyxwvuts
fetching from the database the refraction index of all
the media crossed by the wave.
assembling the characteristic matrices (1.3 1) and
(1.32) associated to the intermediate layers, and
evaluating the final refraction angle Oh, given by
+
A y5
AA.
exp (hc/k,TA,) - 1
-
(2.1)
Using the above expression to replace
in (1.34), and
introducing (1.39) recast in terms of wavelength, yields
the black-body optical generation
1562
zyxwvut
zyx
zyxwvutsrqponmlkjihg
IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, VOL. 12, NO. 10, OCTOBER 1993
\
>
:
*
Y
*
#
*
*
Optical Window
(200 urn)
i
*
* :
silicn dioxide
9
i
zyxw
It
zyxwvutsrq
zyxwvutsrq
zyxwvutsrqpon
pdimlsi0Il~
junction depth
well depth
n-well
\
\
X
paubstrate
Fig. 4 , The optical generation effect, for each monochromatic component
of the impinging radiation, occurs in the two-dimensional strip defined by
the optical window and by the verse of wave propagation.
where as = a (As), gs = 3 (As). Here 3$ is calculated using
the coefficients for non-polarized light in (1.13). Remembering (1.40), the total optical generation is finally found
to be
Gop'(E) =
Fig. 5. Structure of the two diodes described in Sections (3.1) and (3.2).
lected in order to have x j p = 0.56 pm, so that xi,, - x j p
= 7 pm. The applied biases are selected as follows: both
junctions are reversely biased and, for a given bias V, of
the lower junction, that of the upper one ( V u ) is tuned in
such a way as to deplete the n-well region completely. In
this way, for each pair (Vu, V,) the inversion point of the
electric field occurs at a different position in the n-well
region. By way of example, the electric potential calculated by H F I E L D S along a cross section of the device
at Vu = 8 V is shown in Fig. 6. The position of the inversion point of the electric field is about xinv= 4.35 pm
in this case. At a given value of xinv,the upper contact
will collect only the holes generated in the region x I
xinv.In particular, for a monochromatic radiation at frequency v , the amount of collected holes will depend on
the relative magnitude of A = 1/ a (Y) and xinv,thus providing a correlation between current and frequency.
The results of the simulations carried out on the device
described above are shown in Figs. 7 and 8. In the first
one, the current due to the optically-generated holes collected at the upper contact is drawn as function of Vu at
different wavelengths. The current is normalized to its
value at Vu = 15 V which, for the frequency range under
investigation, corresponds to a position of the inversion
point deep enough in the n-well region to have the majority of holes collected by the upper contact. In Fig. 8 the
simulated and experimental relative spectral response r] ,
at a constant reverse bias Vu = 1 V , are shown as a function of the wavelength. Indicating with A , Zuv the junction
area and the current collected at the upper contact at frequency Y , respectively, the relative spectral response is
defined as in [9], namely
zyxwvutsrqp
Gig([) + E' G;pt(E),
V
I
(2.3)
where symbol Cl refers to the individually-defined components only.
111. RESULTS
3.1. Colour-Sensing Photodiode
The physical soundness and software implementation
of the model depicted in the previous paragraphs have
been tested on a number of devices. The results are illustrated in the following. The first set of simulations has
been carried out on a photodiode presented in [6]. This
device is a part of a more complex circuit which works as
a colour sensor. The colour discrimination is obtained
through the frequency dependence of the silicon absorption coefficient. In fact, short-wavelength radiation generates most of the electron-hole pairs within a semiconductor zone close to the surface, while long-wavelength
radiation generates them in a wider region. By varying the
extension of the depleted region it is possible to pinpoint
the wavelength of the impinging radiation. The photodiode has been realized with a pf region diffused in an
n-well which, in turn, was epitaxially grown on a p substrate. More specifically (referring to the structure of Fig.
5 ) , the oxide thickness is 0.2 pm, the p substrate has a
constant acceptor concentration N~~~ = 5 x loi5 cm-j,
the n-well region is made of an epitaxial layer with Nepi
(3.1)
= 7 x 1014(3m-j and xi,, = 7.56 pm, and the pf region
is made of a diffusion whose peak concentration is 5 x In both Figs. 7 and 8, the dots refer to the experiments
10" ~ m - The
~ . standard deviation of the diffusion is se- taken from [ 6 ] ,and the continuous lines refer to the sim-
zyxwvu
zyx
zyxwvutsrqponml
zyxwv
1563
VECCHI er al.: NUMERICAL SIMULATION OF OPTICAL DEVICES
ment between simulation and experimental data resulted
in the whole frequency range.
0
zyxwvutsrqponmlkj
zyxwvutsrqp
L2E-03
60E-04
I BE-03
Device depth ( c m )
Fig. 6. Behaviour of the electric potentials along a cross section of the
colour-sensing photodiode described in Section (3.1).
I
.
zyxwvutsrqpo
Simulation result
. . . . . . Experimental data
(Wolfenbuttel)
04t
zyxwvutsrqponmlkjihg
zyxwvutsrqponmlkjihgfedcbaZYXWV
X = 400,550,630 nm
01
02,
0
;
0
I
i
I
2
I
3
I
4
/
5
-
7
6
A
8
,
!
1
,
1
I
I
_1
1 0 1 1 1 2 1 3 1 4 1 5
9
Reverse Voltage ( V )
Fig. 7. Colour-sensing photodiode (Section 3.1): simulation and experimental Illmax
curves as a function of the reverse voltage of the upper junction for three different values of wavelengths. The upper, intermediate, and
lower curves correspond to X = 480, X = 550, and X = 630 nm, respectively.
zyxwvutsrqponmlkjihgfe
IO,
e,
C
0
3.2. CMOS-Compatible Sensor
The second set of simulations has been carried out on
a specially-designed, CMOS-compatible photodiode fabricated at IRST. The structure under test is structurally
similar to that shown in Fig. 5 , and consists of a 200 X
200 pm2 square photodiode made of a p + region diffised
in and n-well region. The latter, in turn, was diffused in
a p substrate. Two guard rings provide the device isolation with respect to the carriers generated outside the photodiode area. The relatively large dimensions of the photodiode allow for an easier characterization of both the
physical and electro-optical parameters. The peak values
of the doping concentrations are 5 x l O I 9 and 10l6cmP3
in the p + and n-well region, respectively. In the p substrate, the dopant concentration ranges from 4 X l O I 4 to
4 x 10l6 cmP3. The oxide thickness is tOx = 3 pm,
whereas the junction depths are x j p = 0.7 pm and xi,, =
4.7 pm for the p + and the n-well regions, respectively.
Both the above regions are accessed through lateral contacts. The electro-optical measurement setup consists of
a monocromator-illuminator system with a halogen lamp
source and a reference calibrated detector. Orthogonallyimpinging radiation at a wavelength ranging from 400 to
1000 nm has been selected. An HP 4145B parameter analyzer has been used for the measurement of the electrical
parameters.
Figs. 9 and 10 show the results of the numerical simulation and experiments, respectively. In these figures,
the normalized current versus reverse bias is drawn at different wavelengths. Two different terms contribute to the
photocurrent: the first one (essentially diffusive) refers to
the optical generation in the quasi-neutral regions of the
diode, while the second (essentially ohmic) is due to the
optical generation in the depleted region. The obtained
results show that at short wavelengths the photocurrent is
dominated by the diffusive term, and depends weakly on
the bias because the optical generation occurs just beneath
the semiconductor surface, i.e., far from the depleted region of the p + n junction. Conversely, the ohmic term
becomes more important at -long wavelength since the latter reach the depleted layer, and the photocurrent exhibits
a stronger dependence on the applied bias.
The comparison between the experimental and simulation data shows a good agreement, especially at short
wavelengths. As in the previous example, at wavelengths
shorter that 550 nm the effect of surface states becomes
significant: in this case the estimated density of state
ranges from 5 x 10" to 5 x 10" cm-' eV-', with a
capture cross-section still set at
cm2. The lower limit
5 x 10" of the density of states refers to the midgap region, and has independently been determined by preliminary measurements carried out at IRST. The poorer
agreement at long wavelengths is to be ascribed to the
insufficient knowledge of the finer details of the doping
,
,
,
,
I
Simulation result
~
0
400
150
500
550
600
(I50
700
750
800
Wavelength ( n m )
Fig. 8. Colour-sensing photodiode (Section 3.1): simulated and experimental spectral response at constant reverse bias of the upper junction.
ulations . The electron-hole pairs optically generated near
the interface undergo a heavy recombination effect which
has been taken into account in the simulations by introducing a surface-state distribution at the semiconductorinsulator interface. This has been accomplished in
H F I E L D S by means of a distribution of states along the
gap of the form given in [9], in which the capture crosssection was set at
cm2. The other parameters describing the surface states were selected in order to fit the
short-wavelength response, resulting in a distribution
ranging from 5 x 10" cm-2 eV-' to 5 x 1 0 ' ~cm-2 ev-'.
The long-wavelength response was found to be practically
unaffected by the surface states, and a satisfactory agree-
1564
zyxwvut
zyxwvutsrqponmlkjihgf
zyxwvutsrqponmlkjihgfedcbaZYX
zyxwvu
zyxwvutsrqponmlkjihg
IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, VOL. 12, NO. 10, OCTOBER 1993
eo,
S l r n l ~ l a l ~ oresult
n
( w i t h s s.)
__
. I
c
owt
0.
.
.
1-
.
.
.
-
.
:::I
zyxwvutsrqponmlkj
zyxwvuts
zyxwvutsrqpo
. . .
-30
L A
300
20
50
o n o L . - - -
30
IO
0
.
I
.
,
500
400
Slmulalion result
,
600
800
700
'
,
900
,
I000
4 0
Wavelength (nrn)
Ikverse Voltage ( V )
Fig. 9. CMOS-compatible sensor (Section 3.2): simulated I / I m a x curves
as a function of the reverse voltage of the upper junction at different values
of wavelengths. The upper and lower curves correspond to X = 400 and X
= 800 nm, respectively. The wavelength of the intermediate ones is 500,
600, and 700 nm.
Fig. 1 1 . CMOS-compatible sensor (Section 3.2): simulated and experimental spectral response at constant reverse bias of the upper junction.
,
*
,
I
L
+
,
Experimental dala
junction depth
X
O
i
n
l
O
L
L
I O
~-
= 400.. ..800nm
'
LA
20
30
.
1
40
50
Reverse Voltage ( V )
Fig. 10. CMOS-compatible sensor (Section 3.2): experimental l / I m a x
curves as a function of the reverse voltage of the upperjunction at different
values of wavelengths. The wavelengths are the same as in Fig. 9.
Fig. 12. CCD-oriented structures described in Section 3.3.
distribution with the device. This reflects somehow onto
the path of the carriers, especially of those generated at a
larger distance from the interface.
In Fig. 1 1 the simulated and experimental current of
the photodiode vs . wavelength, at constant impinging
power density (90 pW/cm2) and constant reverse bias
(1 V), are shown. The experimental spectrum has been
normalized to the lamp spectral emission. The figure
shows several peaks due to the multiple reflection and interference phenomena resulting from the presence of a
stack of Si02 layers above the semiconductor surface.
Preliminary simulations showed that the number and position of the peaks depends more critically on the stack's
total thickness than on the thickness and composition of
the individual layers, since the latter present similar values of the refraction index. Therefore, in the final simulation a uniform 3.0 pm Si02 has been used.
area. The latter is made of an n+ region diffused in a p
substrate with a doping concentration of Nsub= 1 X lOI4
~ m - The
~ . first diode (buried-channel diode) has a n region with a peak concentration No = 10l6 cmP3, a junction depth xjn = 0.9 pm, and an oxide thickness tox =
2.39 pm. The second one (diffused diode) has a n+ region
with a peak concentration No = lo2' cmP3, a junction
depth x j n = 1.6 pm, and an oxide thickness tox = 2.32
pm. Both photodiodes are reversely biased at 5 V. The
incident radiation has a constant power density of 90
pW/cm2 and a wavelength ranging from 400 to lo00 nm.
The measurement set-up is the same as the one described
above. It is worth noting that, in order to make the comparisons with the experiments easier, the spectral response
3.3 CCD-Oriented Structures
The third set of simulations has been carried out on two
different photodiodes fabricated at IRST with a CCD-oriented process, characterized by three polysilicon levels,
two metallizations and one buried channel. Both structures under test (see Fig. 12) consist of 100 X 100 pm2
square photodiodes with a 90 x 90 pm2 photosensible
has been used in the examples illustrated here. Its relationship with the relative spectral response (3.1) is RS =
q 3 q / ( h v ) . Opposite to the previous example, where the
photocurrent was due to the generation occurring in the
vicinity of the upper junction only, here the contribution
comes from the whole structure. Hence, a much larger
portion of the device in the direction of the light propa-
(3.2)
zyxwvutsrqponm
1565
VECCHI er al.: NUMERICAL SIMULATION OF OPTICAL DEVICES
gation (i.e., vertical in Fig. 12), had to be considered in
the simulation in order to take into account the carriers
photogenerated by long-wavelength radiation. Fig. 13
shows the spectral response of the buried-channel diode.
The high responsivity of the diode at short wavelengths
is due to the small concentration of traps at the semiconductor-insulator interface. In the figure, each curve refers
to a different interface-state concentration, showing the
effect of the latter on the spectral response. The comparison between experimental and simulated data reflects the
high quality of the oxide-silicon interface; the latter, in
fact, is obtained by the same step of the CCD-oriented
process during which the gate oxide is grown, and yields
a global surface-state concentration lower than 10" cm-2.
Fig. 14 shows the spectral response of the diffised
diode. In this case the spectral response at short wavelengths is weakly dependent on the trap density at the interface, since for this diode the junction is deeper than for
the buried-channel one. At long wavelengths the agreement between experimental and simulated data becomes
poorer due to a lower than expected minority-carrier lifetime in the p substrate. The simulated spectral responses
show an inversion in the middle of the considered spectrum with respect to the corresponding experiments. This
is to be ascribed again to some uncertainty in the junction
depth and doping distribution of the two diodes.
06
-
eipcrimcnlal data
simulateddata i ---simulatedd i l l 2 ..-.-
os
-
x)
Fig. 13. Simulated and experimental spectral response of the buried-channe1 diode of Section 3.3. The simulations refer to three different concentrations of trap states.
zyxwvutsrq
zyxwvutsrqpo
zyxwvut
zyxwvut
zyxwvutsrq
zyxwvu
zyxwvutsrqpon
0.6
-
crpcmenlaldata
mulalion data ----
0.5
0.4
0.3
0.2
3.4 CMOS Capacitor
The fourth set of simulations has been carried out on a
CMOS capacitor. Also this device, whose structure is
shown in Fig. 15, has been realized at IRST. Its size is
100 x 100 pm2 with a photosensible area of 90 x 90
pm2. The polysilicon gate illuminated by the impinging
radiation (photogate) is 106 X 106 pm2 and has a thickness tpoly= 0.35 pm. The thickness of the gate oxide is
tpat, = 0.1 pm, while that of the oxide covering the
polysilicon is tOx= 2.39 pm. The capacitor is built on a
n+ region diffised on a p substrate. The doping concentration of the substrate is Nsub= l O I 4 cmP3, while the
peak concentration and the junction depth of the n+ region
are No = 2 X 10l6 cm-3 and xi. = 0.9 pm, respectively.
On this region, another gate is present (output gate) at
0.15 pm from the photogate. Its function is to collect the
charge generated under the photogate and transfer it to a
reversely-biased contact realized on a heavily-doped re~ , = 1.6 pm). In this device the
gion (No = lo2' ~ m - xjn
illuminated region is covered with a layer made of three
materials: two oxide layers and one polysilicon layer. The
presence of a polysilicon layer makes the behavior of the
spectral response quite different from those obtained from
the photodiodes, where only oxide layers were present.
This is due to the absorbent nature of polysilicon. In contrast with the air-oxide-silicon structure, in this case the
transmission coefficient 3 (1.13) is not characterized by
uniform oscillations but shows some evident peaks. As
suggested in [7], the optical behaviour of polysilicon can
be assimilated to that of silicon, obtaining a qualitatively
0.1
400
MO
WO
700
800
WO
x)
Wavelength (nm)
Fig. 14. Simulated and experimental spectral response of the diffused diode
of Section 3.3.
similar transmission coefficient. The description of the
latter can be improved by considering a complex refraction index of polysilicon which depends not only on the
frequency but also on the grain size and doping concentration. In fact, this dependency has been accounted for
in the simulations using the data taken from [8]. The
transmission coefficient for the Si02-polysilicon-Si structure is shown in Fig. 16. The capacitor spectral response
has been obtained applying 1 V to the photogate, 5 V to
the output gate, and 10 V to the output contact, while
keeping the substrate at 0 V. Fig. 17 shows the comparison between the simulated and experimental spectral response: although the peaks in the simulated curve are
steeper and slightly shifted with respect to the experimental ones, the agreement is qualitatively satisfactory. Apart
from the error arising from the incomplete knowledge of
the electromagnetic behavior of polysilicon, the discrepancies in the horizontal position of the peaks, as well as
the differences in their relative amplitudes, are to be ascribed to the inaccuracy in the measurements of the thickness of the three layers.
1566
zyx
zyxwvutsrqponmlkjihgfe
IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, VOL. 12, NO. IO, OCTOBER 1993
silicon dioxide (2.99 urn)
+
,
output contact
Optical Window
Fig. 15. Structure of the CMOS capacitor described in Section 3.4.
0.8
I
Fig. 16. Transmission
0.35
0.4
coefficient for
structure.
zyxwvu
zyxwvut
zyxw
The device is built on a p-type substrate uniformly doped
with NA = 4.5 x 10I6 ~ m - The
~ . S I M S profile reported
in [lo] has been reproduced in H F I E L D S by means of
a double Gaussian diffusion with Nol = 3 X lo2', No2 =
1019 cm-3 peak concentrations, rol = 0, r02 = 50 nm average penetrations, and u1 = 12, u2 = 55 nm standard
deviations, respectively, resulting in an xi,, = 230 nm
junction depth. The dielectric layers deposited on the silicon substrate have been reproduced as well, namely a
5-nm Si02 layer, a 150-nm MgF2 layer, and a 150-nm
ZnS layer. For the Si02 layer a refraction index n = 1.45
has been used. Those reported in [lo] for the other two
materials are in the range 1.35-1.40 and 2.25-2.35, respectively, and the minimum value for both has eventually been used in H F I E L D S since their fluctuations have
shown little effect on the results. The three layers are designed in order to form an antireflection coating. The effect of the latter has been simulated as well and is shown
in Fig. 18, where the transmission coefficient 3 is drawn
as a function of the wavelength. The continuous line represents the transmission coefficient of a pure air-silicon
structure, i.e., before depositing the layers, hence it essentially describes the reflection at the silicon interface.
The dashed line represents a situation, not appreciably different from the previous one, where only the Si02 layer
is present and, finally, the dotted curve shows the combined effect of the three layers.
The simulations have been carried out imposing a blackbody radiation, at different values of the temperature and
impinging power, brought to the cell through an illumination window coinciding with the device's effective area
A, = 2 X 2 cm2. The wavelength-band B introduced in
section 2.3 has been set to [300, 10001 nm. For convenience we shall indicate with Z(B)the value obtained from
(1.42) when the integral is camed out on B . Of course the
value of Z(B) is smaller than that associated with the full
spectrum. For instance, at T = 5800 K it is Z(B) = 96
instead of Z = 135.3 mW * cm2. This figure is close to
zyxwvutsrqp
an
Air-SiO,-PolySi-SiO,-Si
i
Fig. 17. Simulated and experimental spectral response of the CMOS capacitor of Section 3.4.
3.5 Solar Cell
The last example of application consists of a solar cell,
whose structure has been taken from published data [ 101.
zyxw
VECCHI
zyxw
zyxwvutsrqponml
zyxwvutsrqponmlk
zyxwvutsrqponmlkji
1567
al.: NUMERICAL SIMULATION OF OPTICAL DEVICES
et
I
. .
0
-
Air&
Air-SiO2-SI ---Air-ARC-SI -....
-0 02
-0.04
-0.06
-0.OR
-0.1
zyxwvutsrqponmlkjih
zyxwvutsrqponml
-0.12
04
300
400
500
sm
700
My)
WO
lam
Wavelength (nm)
Fig. 18. Solar cell (Section 3.5): transmission coefficient of the antireflection coating compared with those of simpler structures, namely Air-Si and
Air-Si02-Si.
0 ,
I
'
."Id
0
01
0.2
03
04
v (V)
OS
06
07
OR
Fig. 20. Simulated I-V curve of the solar cell at T = 5800 K black-body
temperature. The load characteristic, the maximum-power current, and the
maximum-power voltage are also shown.
I
zyxwvutsrqponmlkjih
I
i
TABLE I
-004
-0 12
OIR
I.
r
'
0
~
........................................
~~~
......... .......--
__I'
~~
Simul.
Exper.
r101
120
636
114
560
63.7
4.9
384
83.5
16.6
142
64 1
145
525
76.1
3.6
400
82.2
18.7
I
01
02
03
04
v (VI
05
06
07
OR
TABLE I1
Fig. 19. Simulated I-Vcurves of the solar cell in the fourth quadrant. The
curves refer to three different values of the black-body temperature.
-
that of [lo], namely I = 100 mW cm2. It should be
remarked, however, that the latter value is to be ascribed
to the "air mass 1.5 condition'' in which the experiments
reported in the above reference have been carried out. The
AM1.5 condition (sun at 45" above the horizon), which
represents a satisfactory energy-weighted average for terrestrial applications, introduces a non-uniform attenuation in the spectral energy density with respect to the AM0
condition. It follows that the spectrum used in our simulations is somewhat different from the experimental one,
despite the near identity of the impinging power densities.
Because of this, the comparisons which will be shown below are of qualitative nature only.
The portion of the simulated I-V curves belonging to
the fourth quadrant are shown in Fig. 19 at three different
black-body temperatures, namely T = 5300, 5800, and
6300 K. It is seen that the effect of a temperature change
is large on the short-circuit current Z,, and less important
on the open-circuit voltage Vac. The intermediate curve,
which corresponds to T = 5800 K, is repeated in Fig. 20,
where it is used for the evaluation of the maximum-power
output P,. The two dashed lines show the maximumpower current (I,) and voltage ( V,), respectively, while
Black-body Temperature (K)
5300
5800
6300
82
626
78
550
42.9
7.0
252
83.6
17.0
120
636
114
560
63.7
4.9
3 84
83.5
16.6
167
644
160
560
89.8
3.5
552
83.5
16.3
zyxwvutsrqp
the thicker line represents the corresponding load characteristic I = - V I R , , R, = V m / I m Letting
.
P , = I, V,,
Pin = A,Z(B), the fill factor and the efficiency of the cell
are derived from the above quantities as [9]
(3.3)
The results of the simulations have been grouped in tabular form: Table I shows the relevant parameters calculated at T= 5800 K and compares them with those taken
from [lo], Table I1 shows a comparison between the simulated parameters calculated at the three different temperatures, and Table I11 shows a comparison between the
zyxwvutsrq
1568
zyxwvutsrqponmlkjih
IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, VOL. 12, NO. 10, OCTOBER 1993
TABLE 111
Concentration Factor
10
1
20
30
40
50
60
~
120
636
114
560
63.7
4.9
384 X 1
83.5
16.6
1203
695
I158
610
707
0.53
384 x 10
84.5
18.4
2410
713
2311
630
1456
0.27
384 x 20
84.7
18.9
3618
723
3470
640
222 1
0.18
384 x 30
84.9
19.3
4829
731
4608
650
2995
0.14
384 X 40
84.9
19.5
6043
736
5817
650
378 1
0.11
384 X 50
85.0
19.7
~
1259
740
7028
650
4568
0.09
384 X 60
84.9
19.8
zyxwvutsrq
zyxwvutsrqpo
zyxwvutsrqpon
zyxwvu
those refemng to r] and V,, as functions of the concentration factor, are also drawn in Figs. 21 and 22. A comparison with published data is possible, in this case, using
those of [ 111 where a device similar to that of [ 101 is presented. Again a fair agreement with the experiments is
found.
IV. CONCLUSIONS
The introduction of the optical-generation effect in a
general-purpose semiconductor-device analysis program
has been described in this paper. The following aspects
have been developed.
165
I
IO
Conccnrnlron Factor
100
A sound physical description of the optical part,
which takes into account the presence of both perfectly dielectric and absorbent media together with
multiple refraction phenomena.
The introduction of the optical-generation effect into
H F I E L D S with the maximum structural and geometrical flexibility, maintaining the generality of the
original simulator.
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPO
zyxwvutsrqponmlkjihgfe
Fig. 21. Simulated efficiency of the solar cell as a function of the concentration factor.
0 76
I
1
0 14
0 72
07
U
0“
Coupling these items, the simulation of realistic optical
devices became possible, profiting by the more detailed
characterization and greater accuracy provided by a numerical-analysis program with respect to the standard analytical approaches. The capabilities of the code have been
illustrated in several examples. The results show that the
effects of absorption, multiple-reflection and interference
phenomena exhibited by the spectral response of the sensors definitely justify the implementation effort required
by the description of the e.m. wave propagation through
the material layers. This becomes particularly relevant for
the many applications where the dependence of the solidstate sensor response on the wavelength of the impinging
radiation is critical.
zyxwvutsrqponmlk
zyxwvutsrqponmlkjihgfedcba
0 68
0 h6
11 (12
Coiicciilrriiuii Factor
100
Fig. 22. Simulated open-circuit voltage of the solar cell as a function of
the concentration factor.
same parameters calculated at different values of the impinging power.
It is seen from Table I that all the simulated parameters
are in the correct range (it is worth adding that no parameter fitting has been attempted here). The purpose of Table I1 is more academic, that is, showing what the trend
of the relevant quantities would be if the temperature of
the illuminating body changed. Finally, Table I11 shows
the effect of a factor C inserted in (1.42) to emulate the
presence of a concentrator. Some data of Table 111, namely
REFERENCES
zyxwvutsrqpo
[l] G. Verzellesi, M. C. Vecchi, M. Zen, and M. Rudan, “Optical Generation in Semiconductor Device Analysis, a General-purpose Implementation,” in Proc. of the Fourrh SISDEP Conf., Zurich, W. Fichtner, D. Aemmer, eds., pp. 57-64, 1991.
[2] G. Baccarani, R. Guerrieri, P. Ciampolini, and M. Rudan,
‘“FIELDS” : A Highly Flexible 2-D Semiconductor-Device Analysis Program, in Proc. of the NASECODE IV Conf., Dublin, J . J. H.
Miller, ed., pp. 3-12, 1985.
zyxwvutsrq
zyxwvutsrqponm
zyxwvutsrqpon
zyxwvutsrqponml
zyxwvutsrqp
zyxwvut
VECCHI et al.: NUMERICAL SIMULATION OF OPTICAL DEVICES
[31 M. Born and E. Wolf, Principle of Optics. New York: Pergamon,
1980.
141 R. H. Bube, Photoconductivity of Solids. New York: Wiley, 1960.
151 H. R. Philipp and E. A. Taft, “Optical Constants of Silicon in the
Region 1 to 10 eV,” Phys. Rev., vol. 120, no. 1, 1960.
R. Wolfenbuttel, Integrated Silicon Colour Sensors. Delft University of Technology, 1988.
171 C. Anagnastopoulos and G. Sadasiv, “Transmittance of Air/SiO,/
Polysilicon/SiO,/Si structures,” IEEE J. Solid-State Circuits, vol. 10,
pp. 177-179, June 1975.
G. Lubberts, B. C. Burkey, F. Moser, and E. A. Trabka, “Optical
Properties of Phosphorus-Doped Polycristalline Silicon Layers,” J.
Appl. Phys., vol. 52, no. 11, 1981.
[9] S. M. Sze, Physics of Semiconductor Devices, 2nd ed. New York:
Wiley, 1981.
[lo] M. A. Green, A. W. Blakers, J. Shi, E. M. Keller, and S . R. Wenham “High-Efficiency Silicon Solar Cells,’’ IEEE Trans. Electron
Devices, vol. ED-31, May 1984.
[ l l ] M. Finetti, P. Ostoja, S. Solmi and G. Soncini, A New Technology
to Fabricate High-Frequency Silicon Concentrator Solar Cells, IEEE
Trans. Electron Devices, vol. ED-27, Apr. 1980.
Maria Cristina Vecchi received the degree in
electrical engineering from the University of Bologna in 1990 and, since then, has been working
at the Dipartimento di Elettronica, Informatica e
Sistemistica (DEIS) of the same University.
At DEIS, she has been engaged in an activity
on the numerical simulation of semiconductor devices, with emphasis on the optimization techniques for process design and on the simulation of
optical sensors. She is presently visiting the IBM
T.J. Watson Research Center of Yorktown
Heights, NY, working on higher-order transport models in semiconductors.
1569
Massimo Rudan (M’80-SM’92) received a degree in electrical engineering in 1973 and a degree
in physics in 1976, both from the University of
Bologna, Bologna, Italy.
After serving as a Naval officer, he joined the
Dipartimento di Elettronica, Informatica e Sistemistica (DEIS) of the University of Bologna in
1975, where he investigated the physical properties of the MOS structures and the problems of
analytical modelling of semiconductor devices.
From 1978 he has been teaching an annual course
of Quantum Electronics in the Faculty of Engineering of the same University, firstly as Lecturer and then as Associate Professor. Since 1983 he has
been working in a group involved in numerical analysis of semiconductor
devices, acting as Task leader in a number of EEC-supported Projects in
the area of CAD for VLSI. In 1986 he was a visiting scientist, on a oneyear assignment, at the IBM Thomas J. Watson Research Center at Yorktown Heights, NY, studying the discretization techniques for the higherorder moments of the Boltzmann Transport Equation. He was again with
IBM later, investigating and implementing impact-ionization models based
on the average energy of the carriers. In 1990, he was appointed Full Professor of Microelectronics at the University of Bologna.
Giovanni Soncini graduated in electronics in 1965
at the School of Engineering of the University of
Bologna, and received the Ph.D. (Libera Docenza) in quantum electronics carrying on research mainly at the Politecnico of Milan and
Stanford University.
Since 1970 he has been responsible for the research activity on MOS physics and technology at
the LAMEL Institute, established in Bologna by
the Italian National Council for Research. In 1984
he became Assistant Professor of Microelectronics at the University of Bologna, and in 1987 became Full Professor of
Aerospace Electronics at the School of Aerospace Engineering of the University of Rome. In 1990 he moved to the chair of Microelectronics Materials and Technologies at the Materials Engineering School of the University of Trento, and acts as Scientist in Charge of the Microelectronics
Division of IRST. He is the author of more than 70 publications in international journals and books on Microelectronics Technology, and Chairman of International Conferences. Workshoo. and Schools. His main interest is now in VLSI CCE/CMOS technolog’ies and development of smart
sensors.