Chapter V

High-injection effects on near-field

radiation mediated
thermophotovoltaics

Contents
V.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

V.2 Numerical solutions of the full and approximate Drift-Diffusion equations . . . . 92

V.2.1 The low-injection approximation and the separation of the minority and ma-
jority carriers (MCS model) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

V.2.2 Questioning the low-injection approximation for NFR-TPV devices . . . . . . . 96

V.2.3 Numerical solution of the full Drift-Diffusion equations (FDD) . . . . . . . . . . 98

V.3 Analysis of low and high-injection regimes in NFR-TPV . . . . . . . . . . . . . . . . 104

V.3.1 Carrier concentrations under illumination . . . . . . . . . . . . . . . . . . . . . . 105

V.3.2 Spatial distribution of the electric field . . . . . . . . . . . . . . . . . . . . . . . . 106

V.3.3 Recombination rate and lifetimes of electrial carriers . . . . . . . . . . . . . . . . 106

V.3.4 Superposition principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

V.3.5 Maximum power output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

V.3.6 Spatial variations of the quasi-Fermi levels . . . . . . . . . . . . . . . . . . . . . . 110

V.4 Optimization of the doping levels and the thickness of the doped regions for
NFR-TPV systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

V.4.1 Optimization of the doping levels of the doped regions . . . . . . . . . . . . . . 111

V.4.2 Optimization of the thickness of the doped regions . . . . . . . . . . . . . . . . . 114

V.5 Toward an experimental configuration: simulations of the performances of an

InSb PIN photodiode in the near field . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

V.6 Conclusions and prospects on the simulations of NFR-TPV devices . . . . . . . . . 119

89

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés
Chap.V: High-injection effects on near-field radiation mediated thermophotovoltaics

V.1 Introduction

The concept that a TPV system could benefit from the modification of radiative heat transfer due
to near-field effects, by decreasing the distance between the radiator and the cell to a subwave-
length distance, was introduced by Whale and Cravalho [19]. These systems are referred to as
near-field radiation mediated thermophotovoltaic (NFR-TPV). The first experimental work on this
topic was conducted by the team of DiMatteo [20, 107, 108]. A silicon radiator and an InAs cell,
placed inside a vacuum chamber, were used. The distance between the radiator and the cell was
controlled by a piezoelectric device. Variations of the short-circuit current were observed when
varying the size of the vacuum gap. More precisely, the authors reported an increase of the short-
circuit current by a factor 5. A dynamic test was performed, where the vacuum gap size was
oscillating. In-phase variations of the short-circuit current with the gap size suggested that the
increase of the short-circuit current was due to near-field effects. The second experimental study
on this subject was conducted by Hanamura [21], and presented only in a scientific conference. A
marginal enhancement of the performances of the system when decreasing the distance between
the radiator and the cell was reported, and the size of the vacuum gap could not be controlled
below 10 µm. Numerical studies of the performances of NFR-TPV systems began with Whale and
Cravalho’s work [19]. The radiator and the cell were considered semi-infinite. The optical prop-
erties of the radiator were described by a Drude model, while the cell was made of an alloy of
InAs and GaAs. Fluctuational Electrodynamics was used to model radiative heat transfer between
the radiator and the cell. A 100% quantum efficiency was considered to avoid modeling transport
of electrical charges inside the PV cell. No thermal effects were accounted for. They reported a
large increase of the electrical power output of the system due to near-field effects, but only a
marginal gain on conversion efficiency. Narayanaswamy and Chen [109] introduced the idea that
surface phonon-polaritons supported by a cBN substrate could enhance the efficiency of NFR-TPV
systems, by considering a cell made a of a fictitious semiconductor, with a bandgap energy cor-
responding to the frequency of the SPhP resonance of cBN. Laroche et al. [110] analyzed the
performances of NFR-TPV systems using a tungsten radiator and a source with an emission spec-
trum concentrated in a narrow spectral band, referred to as "quasi-monochromatic" heat source.
They reported a maximum efficiency of the system of 35 % using the quasi-monochromatic heat
source, and also analyzed the impact of near-field effects on the dark current. Park et al. [27]
introduced in the modeling of NFR-TPV systems an electrical model with a spatial distribution of
the generation rate of electron-hole pairs inside the PV cell. With a tungsten radiator and a cell
made of an alloy of InSb and GaSb, they reported a maximum conversion efficiency of 23 %. The
most complete model to simulate the performances of NFR-TPV systems up to now was presented
by Francoeur et al. [4]. A thermal model was added, allowing the analysis of the performances
of the NFR-TPV systems while accounting for the thermal behavior of the PV device on the per-
formances of the system. By calculating the different heat sources inside the cell, the authors
calculated the temperature of the cell, and evaluated the impact on the maximum power output
of the systems. It was shown that to maintain the cell at ambient temperature, a convective heat
transfer coefficient up to 105 is required due to the important heat sources arising inside the PV
cell. With smaller values of the convective heat transfer coefficient, the cell temperature increases

90

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés
 V.1. Introduction

and the energy conversion performances drop significantly. Attemps of optimizations of NFR-TPV
systems were also performed. Bright et al. [25] studied the impact of a backside gold reflector to
increase the amount of radiative energy absorbed by the cell. They also reported the impact of
the small penetration depth of evanescent waves on the increase of surface recombinations. This
phenomenon was later studied in depth in [7]. The contributions of the propagating, frustrated
and surface modes were analyzed. It was shown that a radiator supporting mostly frustrated
modes generates a stronger electrical power output than a radiator supporting surface modes,
despite the fact that SPhPs were used to enhance radiative heat transfer from the radiator to the
cell in a narrow spectral range. This is mostly due to the fact that optical optimization is not suf-
ficient: electrical and thermal losses have to be taken into account to optimize the system. The
idea of a spectral filter to cut high-energy photons was also introduced: photons of energy much
larger than the bandgap generate electron-hole pairs (EHP) but also heat through thermalization.
Therefore, due to thermal impact, high-energy photons can be detrimental. Other optimizations
were performed using graphene [111, 112, 113, 28] in order to increase the near-field thermal
radiation between the radiator and the cell, or radiators supporting hyperbolic modes with large
penetration depth [114]. Eventually, more exploratory systems, such as solar NFR-TPV systems
were investigated [115], where the radiator is heated by the sunlight radiation.
Up to now, no convincing experimental demonstration proving the efficiency of the concept of
NFR-TPV systems could be performed. The only two experimental works exhibited marginal in-
creases of the energy conversion performances due to near-field effects. This is due to two facts.
First, the control of the distance between the radiator and the cell is a technological challenge.
Successful experiments on the measurement of thermal radiation between objects separated by
subwavelength distances were performed only recently (e.g. [91, 90, 96]). The second fact is
that without optimization of the PV cell, any increase of the performances due near-field effects
is likely to be marginal because of the impact of the radiative, electric and thermal losses in the
PV device. Understanding the different phenomena occurring in the PV cell and optimizations
of the full NFR-TPV system are thus required to design an efficient prototype, that could allow
demonstrating that a TPV system can benefit from near-field effects.
In this chapter, we first present the numerical model that allows simulating properly the trans-
port of the photogenerated electrical charges inside the PV cell. The usually applied low-injection
approximation will be introduced first. This model was already used in many theoretical works
on NFR-TPV systems [27, 4, 25, 7, 28]. However, no one ever questioned the validity of these
approximations in the case of NFR-TPV converters where the injection of carriers is high due to
the near-field radiation enhancement. Thus, we present the model we have implemented in our
simulation code that solves the transport of electrical charges problem in NFR-TPV cells without
taking into account the low-injection approximation. The domains of validity of the different
approximations are then assessed. The following section is dedicated to new optimizations us-
ing the appropriate charge transport model of two parameters of the PV cell, the doping levels
and the thicknesses of the p and n-regions, in order to increase the performances of the NFR-
TPV systems. Eventually, we end this chapter by analyzing the performances of InSb cells with
a PIN architecture, that are PV cell candidates for the conversion of thermal energy emitted by
relatively low-temperature (≈ 800K) sources, and for which a full Drift-Diffusion modeling is

91

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés
Chap.V: High-injection effects on near-field radiation mediated thermophotovoltaics

required.

V.2 Numerical solutions of the full and approximate Drift-Diffusion

equations

As mentioned in Chapter II, in order to evaluate the performances of a NFR-TPV system, its elec-
trical behavior under illumination and applied voltage has to be modeled. We recall the basic
equations of the Drift-Diffusion (DD) model, in the 1D and steady state. First, Poisson’s equation
relates the electric field (or the electrostatic potential profile) to the spatial distribution of the
electrical charges inside the PV cell writes

d E(z) d 2 V (z) e
=− = − n(z) − p(z) + Na (z) − Nd (z) . (V.1)
dz dz 2 "
The transport equations state that the electrical current is due to the motion of the electrical
charges that are either swept by the electric field or diffusing from high to low concentration
regions

dn
Jn = en(z)µn E(z) + eDn (V.2)
dz
for the electrons, and

dp
J p = ep(z)µ p E(z) − eDp (V.3)
dz
for the holes. Eventually, the continuity equations relate the spatial variations of the electrical
currents with the net generation-recombination rates

1 d Jn (z)
− = G(z) − R(z), (V.4)
e dz
for the electrons, and

1 d J p (z)
= G(z) − R(z). (V.5)
e dz
for the holes. The number of equations can be lowered from 5 to 3 by combining the continuity
and transport equations, thus eliminating the electrical currents Jn and J p

d 2 n(z) d n(z) d E(z)

Dn + µn E(z) + µn n(z) = R(z) − G(z), (V.6)
dz 2 dz dz

d 2 p(z) d p(z) d E(z)

Dp − µ p E(z) − µp p(z) = R(z) − G(z). (V.7)
dz 2 dz dz
In Eq. V.6 and V.7, it is assumed that the mobilities µn and µ p and the diffusion coefficient Dn
and Dp do not vary spatially in each region. Mobilities and diffusion coefficients depend on the

92

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés
 V.2. Numerical solutions of the full and approximate Drift-Diffusion equations

temperature and the doping levels. As the PV cell is thin, its spatial temperature distribution is
almost uniform. Furthermore, we consider abrupt doping profiles, which means that the number
of donors and acceptors are constant in each region of the PV cell. To solve the charge transport
problem, Eqs. V.1, V.6 and V.7 have to be solved for the electric field E and the carrier densities
n and p. These three equations are non-linear, because the recombination terms are carrier-
density dependent (Eqs. II.16, II.17, II.19). Moreover, they are coupled by the fact that the
electric field depends on the carrier density and vice-versa. Two methods can be used to solve
the Drift-Diffusion equations. The simplest case is to assume a set of hypotheses, called under
the naming low-injection approximation, that simplifies the problem. However, the low-injection
approximation is questionable in the case of NFR-TPV systems. To solve the problem without
applying the low-injection approximation, an iterative method can be used. In the following, we
will present the two approaches, starting by the low-injection approximation.

V.2.1 The low-injection approximation and the separation of the minority and ma-
jority carriers (MCS model)

A schematic of the pn-junction under consideration is depicted in Fig. V.1.

Figure V.1: Schematic representation of the pn-junction under consideration, assuming the low-
injection approximation.

In the p-region, the concentration of holes in larger than the concentration of electrons, due to
the doping. In this case, the electrons are the minority carriers in this region, and the holes are
the majority carriers. In the n-region, the electrons are the majority carriers and the holes are the
minority carriers. The low-injection approximation consists in assuming that the minority carrier
densities n and p are always much lower than the doping levels Na and Nd in each region. It
means that these approximations are valid only under some level of illumination, or if the doping
level is high enough. The first impact of this approximation can be seen on Poisson’s equation,

93

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés
Chap.V: High-injection effects on near-field radiation mediated thermophotovoltaics

that becomes

d E(z) d 2 V (z) e
= = (Na (z) − Nd (z)) . (V.8)
dz dz 2 "
The dependence of the electric field on carrier densities is removed, meaning that it is now in-
dependent of the illumination and the applied voltage. The new Poisson’s equation can now be
solved for the electric field. As an abrupt doping profile in considered, Na and Nd are constant
in each region. As a consequence, the electric field is nil outside the depletion region, in the
quasi-neutral regions. The size of the depletion region can be calculated as [34]

√ Å ã
 2" 1 1
Ld p = l p + ln = V0 + (V.9)
e Na Nd

where V0 is the built-in voltage

 
kb T Na Nd
V0 = ln . (V.10)
e n2i

The thickness of the depletion region in the p and n-regions l p and l n can be expressed as

Nd Na
l p = Ld p , ln = Ld p . (V.11)
Na + Nd Na + Nd

When a forward bias Vf is imposed on the PV cell, the thickness of the depletion is calculated as
[34]

√ Åã
 2" 1 1
Ld p = l p + ln = (V0 − Vf ) + . (V.12)
e Na Nd

The next approximation is the full depletion approximation: if an electrical charge reaches the
depletion region, the electric field sweeps it toward the region where it is a majority carrier and
no recombination occurs. The electrical carrier generates a current. As a consequence, Eqs.
V.6 and V.7 need to be solved in the quasi-neutral regions, more specifically in the p-region for
electrons and in the n-region for holes. As E is nil outside the depletion region, Eqs V.6 and V.7
become

d 2 n(z)
Dn = R(z) − G(z) (V.13)
dz 2
for the p-region, and

d 2 p(z)
Dp = R(z) − G(z) (V.14)
dz 2
for the n-region. As shown in Chapter II, when the low-injection approximation is considered,

94

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés
 V.2. Numerical solutions of the full and approximate Drift-Diffusion equations

the recombination rate can reduce to

∆n
R= (V.15)
τn

for electrons in the p-region, and

∆p
R= (V.16)
τp

for holes in the n-region.

Equations V.13 and V.14 can then be written as

d 2 ∆n(z) ∆n
Dn − + G(z) = 0 (V.17)
dz 2 τn

for the p-region, and

d 2 ∆p(z) ∆p
Dp − + G(z) = 0 (V.18)
dz 2 τp

for the n-region. Eq. V.18 and V.19 are two second-order non-coupled and linear differential
equations, that can easily be solved numerically using the finite element method and a tridiagonal
matrix algorithm (TDMA) [116], or even analytically if an analytic expression of the generation
rate of electron-hole pair (G) can be provided (as in [28]). Because Eqs. V.18 and V.19 are
decoupled, it is assumed that the behavior of the minority and majority carriers are independent
in each region. Boundary conditions are required to solve these equations. At the boundaries of
the depletion region, the excess of carriers ∆n and ∆p is assumed to be nil due to the electric
field

p
∆n(z = zd p ) = ∆p(z = zdnp ) = 0, (V.19)

p
where zd p and zdnp are the spatial positions of the edges of the depletion region on the p-side and
n-side, respectively (Fig. V.1).
At the boundaries of the cell, the electrical current is balanced by the surface recombinations

d∆p(z = z p )
Dn = Sn ∆n, (V.20)
dz
in the n-region, and

d∆p(z = zn )
Dp = S p ∆p, (V.21)
dz
in the p-region. S p and Sn are the surface recombination velocities for holes and electrons, re-
spectively. zn is the position of the boundary of the cell in the n-region, and z p is the position of
the boundary of the cell in the p-region. Once the spatial distribution of the excess carrier concen-

95

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés
Chap.V: High-injection effects on near-field radiation mediated thermophotovoltaics

trations are found, the electrical currents can be derived. If electrical charges are photogenerated
in the depletion region, they are assumed to be collected. The corresponding electrical current
is:

∫ zdnp
Jd p = e G(z). (V.22)
p
zd p

As explained before, if charges are photogenerated in the quasi-neutral regions, they are assumed
to be collected if they reach the depletion region. The current of electron and holes can therefore
be written as proportional to the gradient of ∆p and ∆n at the edges of the depletion region

p
d∆n(zd p )
Jn = eDn , (V.23)
dz

d∆p(zdnp )
J p = −eDp . (V.24)
dz
The photogenerated current is then the sum of the three components

Jph = Jn + J p + Jd p . (V.25)

To evaluate the performances of NFR-TPV devices, the problem has to be solved for different
applied voltages Vf . When the cell is under dark conditions (i.e. not illuminated), a current
appear when the voltage increases, called dark current (Jdark ). To save computation time, only
the dark current is computed as a function of Vf . The dark current is obtained by solving the
transport equations (Eqs. V.17 and V.18) without illumination (G = 0). Once the J-V characteristic
under dark conditions is obtained, the total current is the calculated as

Jtot (Vf ) = Jph + Jdark (Vf ). (V.26)

Here, it is assumed that the dark current does not depend on the level of illumination. This is
possible by means of the superposition principle.

V.2.2 Questioning the low-injection approximation for NFR-TPV devices

In the preceding section, several approximations have been made to simplify the resolution of the
Drift-Diffusion equations:
◦ The density of minority carriers is much smaller than the doping everywhere,
◦ The electric field is not dependent on the level of illumination,
◦ The expressions of the recombination rates can be simplified using a spatially constant
lifetime,
◦ The recombination current depends only on the applied voltage, and not on the level of
illumination.

96

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés
 V.2. Numerical solutions of the full and approximate Drift-Diffusion equations

In the case of non-concentrated solar PV, these approximations can been made because of the level
of illumination is relatively low. Furthermore, in the case of silicon solar cells (the most commonly
used type of solar cell), because of the large bandgap of silicon, the intrinsic carrier concentration
is small. It induces smaller carrier concentrations at equilibrium, and therefore smaller carrier
concentrations under illumination and applied voltage. However in NFR-TPV systems, the semi-
conductors under consideration (III-V compounds) have smaller bandgap energies. In addition,
the levels of illumination are larger than in non-concentrated solar PV, as depicted in Fig. V.2(a).
In this figure, the radiative power emitted by a radiator at 2000 K and absorbed by a GaSb cell as
a function of the distance between the two bodies is presented. One can observe that the values

Figure V.2: (a): Radiative power absorbed by a 10.4 µm thick GaSb cell in the vicinity of a radiator
at 2000 K as a function of the distance separating them. From [7]. (b): Map of the radiative heat
flux as a function of the radiation energy and the depth inside p-region of a Gasb cell. The radiator
(2000 K) supporting surface polaritons is placed at 20 nm from the cell. From [8].

increase exponentially when the distance between the two bodies decreases, up to an absorbed
total radiative heat flux of ≈ 3.106 W.m−2 . In comparison, the value of the incident radiative
power from the sun at the sea is 103 W.m−2 for the standard AM1.5 spectrum [15]. It means that
the amount of photogenerated electrical charges inside a PV cell is much larger in NFR-TPV de-
vices than in non-concentrated solar PV devices. Even in concentrated solar photovoltaics where
the illumination can reach 46000 suns, the radiative power absorbed by the cell is lower than in
TPV systems in the extreme near-field. Moreover, it can be observed in Fig. V.2(b) that absorption
is very local, especially when using radiators supporting SPhPs. It is observed that most of the
radiative power is absorbed in the first dozens of nanometers due to the small penetration depth
of evanescent waves. Because of this, the validity of the low-injection approximation has to be
verified.

97

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés
Chap.V: High-injection effects on near-field radiation mediated thermophotovoltaics

V.2.3 Numerical solution of the full Drift-Diffusion equations (FDD)

If the low-injection approximation is not used, the Drift-Diffusion equations cannot be solved
directly, but rather with an iterative method. In this section, we will present Gummel’s iterative
procedure [117] used to solve the problem. It consists in taking an initial guess on the electric field
(or the electrostatic potential) and the carrier densities. Using the values of the carrier densities,
Poisson’s equation can be solved for the electrostatic potential, and the electrostatic potential is
then used to solve the transport equation for the carrier densities. The process is repeated until
convergence.

V.2.3.1 Variables, normalization and scaling

The first step is to redefine a set of variables to solve the problem. The unknowns are the electric
field E and the carrier densities n and p. However these variables are not convenient to solve the
full Drift-Diffusion problem. Therefore, we solve the Drift-Diffusion equations for the electrostatic
potential V , and the quasi-Fermi levels for the electrons and holes φn and φ p . The quasi-Fermi
levels describe the population of electrons and holes in the conduction and valence bands in non-
equilibrium (i.e. when a voltage is applied or under illumination). They represent the shift in
potential due to the excess of electrical carriers. They are related to the carrier concentrations
as
Å ãÅ ã
eV (z) −eφn (z)
n(z) = ni exp , , (V.27)
kb T kb T

Å ã 
−eV (z) eφ p (z)
p(z) = ni exp . (V.28)
kb T kb T

The red terms in Eqs. V.28 and V.29 correspond to the equilibrium carrier concentrations of the
electrons and holes inside the PV cell. To simplify the equations, normalized variables are intro-
kb T
duced. The electrostatic potential can be normalized by the thermal voltage V ∗ = e to define a
new variable

V
V̄ = , (V.29)
V∗
called normalized electrostatic potential. The quasi-Fermi levels can be written in the form of the
so-called Slotboom variables [118]

−eφn
Φn = exp , (V.30)
kb T

eφ p
Φ p = exp , (V.31)
kb T

98

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés
 V.2. Numerical solutions of the full and approximate Drift-Diffusion equations

where Φn and Φ p are the Slotboom variables for electrons and holes, respectively. Using these
notations, the carrier densities become

 
n(z) = ni exp V̄ Φn , (V.32)

 
p(z) = ni exp − V̄ Φ p . (V.33)
r
∗ "k b T
The space variable can be normalized by the intrinsic Debye length z = e2 ni

z
z̄ = . (V.34)
z∗

V.2.3.2 Solving the Poisson equation

Poisson’s equation can be rewritten using the variables introduced before

Å ã
V ∗ d 2 V̄ eni Na (z) − Nd (z)
= exp(V̄ (z))Φn (z) − exp(− V̄ (z))Φ p (z) + . (V.35)
z ∗2 dz̄ 2 " ni
V∗ eni
By noticing that the factor z ∗2
is equal to " , Poisson’s equation takes the form

Å ã
d 2 V̄ Na (z) − Nd (z)
= exp( V̄ (z))Φ n (z) − exp(− V̄ (z))Φ p (z) + . (V.36)
d 2 z̄ 2 ni

This differential equation is non-linear, meaning that an iterative method has to be used once
again. First, an initial guess on V̄ is chosen. Then, at the next iteration, a new potential is written
as V̄ i+1 = V̄ i + δ V̄ . Injecting this in Eq. II.37, we have

d 2 V̄ dδ V̄
+ =
dz̄ 2 dz̄ 2
Å ã (V.37)
¯ ¯ Na (z) − Nd (z)
exp(V̄ (z))exp(δV (z))Φn (z) − exp(− V̄ (z))exp(−δV (z))Φ p (z) + .
ni

To linearize this equation, we can use the first order Taylor expansion exp(±δ V̄ ) = 1 ± δ V̄

d 2 V̄ dδ V̄
+ =
d z̄
2 2 dz̄ 2
Å ã (V.38)
¯ ¯ Na (z) − Nd (z)
exp(V̄ (z))(1 + δ V (z))Φn (z) − exp(− V̄ (z))(1 − δ V (z))Φ p (z) + .
ni

The linearized Poisson’s equation can be solved for δ V̄ using a TDMA procedure. The potential at
iteration i + 1 can then be calculated as a function of that at step i. The process is then repeated
until δ V̄ reaches a value close to 0 at each spatial node of the system (i.e. lower than a prescribed
convergence criterion).

99

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2016LYSEI103/these.pdf

© [E. Blandre], [2016], INSA Lyon, tous droits réservés