Grover Thesis11 PDF
Grover Thesis11 PDF
Grover Thesis11 PDF
by
SACHIT GROVER
Doctor of Philosophy
2011
This thesis entitled:
_____________________________
_____________________________
Date ________________
The final copy of this thesis has been examined by the signatories, and we
find that both the content and the form meet acceptable presentation standards
Two types of ultra-fast diode are fabricated, characterized, and simulated for
iii
The RC time constant of MIM diodes is too large for efficient operation at near-
are fabricated using graphene and measured for response to infrared illumination.
the measured current-voltage characteristics are consistent with each other. I have
also derived a semiclassical theory, analogous to the one for MIM diodes, for
iv
Dedicated to my parents
&
my wife
Ginni
v
ACKNOWLEDGMENTS
The last five years of my life have been enriched in many ways through my
interaction with Prof. Garret Moddel. I wish to extend my sincere thanks to him for
Park, Charles Rogers, and Bart Van Zeghbroeck, for sharing their expertise on
various aspects of my thesis. I also thank Profs. E. F. Kuester, Kelvin Wagner, and
Leo Radzihovsky for helpful conversations. I sincerely thank Mike Estes for our
collaborations, and Mario Scurati for his questions related to rectenna design.
Thanks is due to the never tiring team at CNL, including Jan Van Zeghbroeck,
Tomoko Borsa, and Vincenzo LaSalvia, and to Bill Mitchell at UCSB for process
support. I also thank former Phiar Corporation employees for their help and
For making grad school an enjoyable experience and for being partners in
procrastination, I wish to thank current and former members of the Quantum Engr.
Lab. including Olga Dmitriyeva, Rahul Trivedi, David Doroski, James Zhu, Kendra
the same spirit, I also thank my friends outside the lab including Venkata Tamma,
I am eternally grateful to my parents, Kamlesh and Anand, for the strength and
and Shalini, and their families for constant love and support, with a high five to
Sumira, Arnav, Neal, and Kabir for enlivening my vacations. Finally, using the
clichd adage, behind every successful man there is a woman, I thank my wife
vi
TABLE OF CONTENTS
I. INTRODUCTION ........................................................................................................ 1
A. Rectennas ................................................................................................ 1
B. High-speed diodes ................................................................................... 4
C. Thesis outline .......................................................................................... 9
II. METAL-INSULATOR TUNNEL DIODES........................................................ 11
A. MIM diode simulations ......................................................................... 12
B. Single-insulator (MIM) diodes.............................................................. 19
C. Shortcomings of MIM diodes ................................................................ 25
III. DOUBLE-INSULATOR (MIIM) TUNNEL DIODES ............................................... 29
A. Simulation methodology ....................................................................... 29
B. Double-insulator configurations ........................................................... 31
C. Comparison of MIM and MIIM diodes ................................................. 36
D. Interface stability.................................................................................. 38
IV. RECTENNA CIRCUIT & EFFICIENCY ................................................................ 40
A. Circuit analysis ..................................................................................... 42
B. Power received by a rectenna ............................................................... 45
C. Impedance match and cutoff frequency ............................................... 49
D. Rectenna applications ........................................................................... 54
E. Rectenna design for solar energy harvesting ...................................... 59
F. Ultimate efficiency of rectenna solar cells ........................................... 61
V. SEMICLASSICAL THEORY OF RECTENNAS ....................................................... 64
A. Semiclassical theory ............................................................................. 66
B. Semiclassical rectifier-properties ......................................................... 73
C. Load-line analysis ................................................................................. 78
D. Correspondence between circuit semiclassical theory ........................ 84
VI. MIM TRAVELING-WAVE DETECTOR ................................................................ 88
A. Theory of operation ............................................................................... 88
B. TW concept and modeling..................................................................... 91
C. Performance calculation ....................................................................... 95
D. Comparison with IR detectors .............................................................. 99
vii
VII. FABRICATION AND CHARACTERIZATION OF GEOMETRIC DIODES ....... 102
A. Metal geometric diodes ....................................................................... 103
B. Graphene geometric diodes ................................................................ 106
C. Infrared characterization of antenna-coupled diodes........................ 112
VIII. QUANTUM SIMULATION OF GEOMETRIC DIODE ..................................... 117
A. NEGF formalism ................................................................................. 118
B. Tight-binding model (H) ..................................................................... 123
C. Contact self-energy () ....................................................................... 126
D. Self-consistent NEGF-Poisson solver ................................................. 127
E. A simulation example ......................................................................... 130
F. Simulated geometric diode I(V) characteristics ................................. 134
G. Comparison of simulated and experimental characteristics............. 140
IX. SEMICLASSICAL THEORY OF OPTICAL FREQUENCY RECTIFICATION IN
MESOSCOPIC DIODES ....................................................................................... 142
A. Mesoscopic junction under illumination ............................................ 143
B. Projecting illuminated characteristics from DC I(V)......................... 146
C. Discussion ........................................................................................... 150
X. CONCLUSIONS & FUTURE WORK ..................................................................... 152
A. Metal-insulator diodes ........................................................................ 152
B. Rectenna circuit and design ............................................................... 153
C. Semiclassical analysis of MIM diodes ................................................ 154
D. Geometric diodes ................................................................................. 155
E. Future work ........................................................................................ 155
REFERENCES ......................................................................................................... 158
APPENDIX - A METAL/INSULATOR/METAL FIELD EFFECT TRANSISTOR
.............................................................................................................................. 169
A. Introduction ........................................................................................ 169
B. Background ......................................................................................... 170
C. Device design and fabrication ............................................................ 171
D. Electrical characterization ................................................................. 179
E. Device modeling .................................................................................. 185
F. Conclusions and suggestions for future work .................................... 189
viii
LIST OF FIGURES
Figure II-2 Comparison of transfer-matrix, WKB, and quantum transmitting boundary methods.... 17
Figure II-3 Comparison of simulated and experimental I(V) characteristics for two MIM diodes ...... 18
Figure II-4 Current-voltage characteristics of two MIM diodes having different barrier heights....... 20
Figure II-5 Contribution to total current density by electrons at different energies ............................ 22
Figure II-6 Normalized current density vs. temperature for a range of barrier heights ...................... 24
Figure II-7 Power-law coefficient vs. barrier-height for temperature dependence ............................... 25
Figure II-8 Responsivity and resistance vs. barrier asymmetry for single-insulator diodes ............... 27
Figure III-1 Energy-band profiles for the resonant and step MIIM diodes ........................................... 32
Figure III-2 Transmission probability and current density for resonant and step MIIM diodes. ....... 33
Figure III-3 Current density, resistance and nonlinearity for resonant and step MIIM diodes .......... 35
Figure III-4 Resistance vs. responsivity at zero bias for single- and double-insulator diodes ............. 37
Figure IV-4 Effect of varying the diode size on the antenna to diode coupling efficiency. ................... 51
Figure IV-5 Separating the effect of impedance match from cutoff frequency ...................................... 52
Figure IV-8 Fraction of solar energy vs. minimum wavelength that can be harvested efficiently. ..... 61
Figure V-1 Conduction band profile of an MIM diode modulated by an AC voltage ............................ 67
Figure V-2 Conduction band profile of an MIM diode with photon assisted transport ........................ 68
ix
Figure V-3 Obtaining semiclassical resistance from the inverse of the slope of secant ....................... 71
Figure V-4 Semiclassical resistance and responsivity vs. photon energy .............................................. 72
Figure V-5 Obtaining Iillum vs. voltage curve starting with a step Idark.................................................. 74
Figure V-6 Calculated I(V) characteristics for an illuminated MIIM diode .......................................... 75
Figure V-7 Calculated |I(V)| characteristics on a log-scale with varying signal strength. ................. 76
Figure V-9 Iillum vs. voltage for a piecewise linear I(V). .......................................................................... 80
Figure V-10 Illuminated characteristics for piecewise linear, and exponential dark I(V) curve. ........ 83
Figure VI-2 Finite element analysis of a Ni-NiO-Ni traveling-wave diode for 100 THz ...................... 94
Figure VI-3 Calculated characteristic impedance of the traveling-wave diode vs. wavelength. .......... 96
Figure VI-5 Responsivity comparison of the lumped-element and the traveling-wave detectors........ 98
Figure VI-6 Comparison of MIM, semiconductor, and thermal detectors ........................................... 100
Figure VII-1 Antenna-coupled metal geometric diode with a 20 nm neck. ......................................... 103
Figure VII-3 SEM image of the antenna and the four-point probe contacts ....................................... 105
Figure VII-6 Four point probe configuration for measuring the diode I(V) ......................................... 110
Figure VII-7 Current, resistance, and responsivity for a graphene geometric diode.......................... 111
Figure VII-8 Setup for measuring response of antenna-coupled geometric diodes to illumination... 113
Figure VII-9 Short circuit current vs. angle between field and antenna polarization........................ 114
Figure VIII-1 The device region represented by a Hamiltonian H and potential U. ......................... 119
Figure VIII-2 Determining the tight-binding Hamiltonian (H) for armchair graphene..................... 124
x
Figure VIII-3 Programming a geometric-diode-shaped graphene........................................................ 125
Figure VIII-5 Representation of semi-infinite quasi-1D reservoirs connected to a device. ................ 126
Figure VIII-11 Transmission vs. energy obtained from the self-consistent solution. ......................... 132
Figure VIII-13 Current vs. energy calculated from T(E) and Fermi distribution. .............................. 134
Figure VIII-14 Current vs. energy for opposite voltages across the diode........................................... 135
Figure VIII-15 Asymmetric I(V) characteristics of the geometric diode obtained from simulation... 136
Figure VIII-16 Comparison of current vs. energy at positive- and negative-bias for majority electron
Figure VIII-17 Forward-to-reverse current asymmetry vs. diode voltage with changing Eoffset......... 138
Figure VIII-18 Symmetric I(V) curve for a symmetric sheet of graphene. .......................................... 139
Figure VIII-19 Symmetric I(V) curve for a symmetric conductor with confinement. ......................... 139
Figure VIII-20 Comparison of I(V) curves obtained from Drude model, self-consistent NEGF +
Figure IX-1 A two terminal device with an arbitrary shape, subjected to an AC potential V(r,t). .... 143
xi
CHAPTER I
INTRODUCTION
A. Rectennas
Extensive research is being conducted on new materials and concepts for the
next generation of energy conversion and sensing devices (Luryi, 2010). In order to
make solar energy as cost effective as coal, the requirements for a third generation
solar cell are low-cost and high power conversion efficiency (Green, 2001). Similar
cost and efficiency requirements are imposed on infrared detector technologies for
applications like automotive night vision. The rectenna is a device that has the
potential to deliver the desired performance at low-cost for both photovoltaics and
at any frequency.
by the diode. The radiation induces an AC signal on the antenna arms, which gets
channeled into a region, called the antenna feedpoint, where the diode is connected.
cycles of the oscillating current induced on the antenna. Depending on whether the
1
Figurre I-1 Schem
matic of an antenna-cou
a upled diode rectifier, allso
known as a recteenna.
Sign
nificant research wa
as conducte
ed in the 1960s and
d 70s towa
ards the usse of
frequen
ncies, recten
nnas with power-conv
version effficiencies g
greater tha
an 90% (Brown,
1976) have
h been demonstra
ated and are
a used in
n a variety
y of energy
y transmisssion
(Shinoh
hara, 1998) and harve
esting (Hag
gerty, 2004
4), (Singh, 2
2006) applications.
investig
gated (Fum
meaux, 1998
8). Infrared
d rectenna
a-detectors have several advanttages
require cooling an
nd there is no
n lower liimit on thee frequency
y of the sign
nal that ca
an be
detected
d. Moreove
er, rectenna
as can dete
ect the env
velope of a carrier sig
gnal modullated
at an extremely
e high frequency. For infrared
d carrier w
wavelength
hs, this alllows
rectennas to have modulatio
on bandwid
dths severa
al orders oof magnitude higher than
Pho
otovoltaic rectification
n is anothe
er applicatiion where rrectennas can potenttially
gained significant
s attention.. Here agaiin, the abssence of a b
bandgap im
mplies thatt the
efficienccy limitatio
ons of a sem
miconducto
or solar celll are not a
applicable.
2
Despite the decades of work done in this field, infrared and visible applications
of the rectenna are still in the research phase. On the antenna side, historically,
planar metal antennas have been used in infrared detectors (Fumeaux, 1998) and
work is underway for their large scale integration into flexible substrates for
resistive losses in the antenna-metal limit the antenna efficiency to below 50%
(Gonzlez, 2005). A proposed alternative is dielectric antennas that collect and
channel radiation by means of a dielectric rod (Sarehraz, 2005). Such antennas can
The choice of a suitable diode for a rectenna is based on its operating frequency.
The transit time of charges in semiconductor p-n junction diodes limits their
using GaAs Schottky diodes have been designed (Yoo, 1992). Schottky diodes are
also used at terahertz and far-infrared frequencies (Brown, 2004), (Kazemi, 2007).
optical frequencies. These include the MIM diode and a new diode, referred to as
geometric diode. In conjunction with my work on the diodes, I lay emphasis on the
or a solar cell. In the next section, I provide a background on MIM diodes, which
have been extensively used in optical rectennas. I also explain the operation of
3
B. High-speed diodes
1. MIM diode
apart by an extremely thin (few nanometers) insulator. In the MIM diode, the
metals have a higher work function than the electron-affinity of the insulator
E
Vacuum level
Electron
affinity
Workfunction ()
() Conduction band
e- Electron tunneling
the thickness and height of the barrier (Stratton, 1962), which changes with the
voltage across the diode. This gives rise to the nonlinear dependence of the tunnel
4
Electron tunneling in MIM junctions occurs on a femtosecond timescale (Nagae,
1972). This inherently fast charge transport across the tunnel barrier allows MIM
conduction mechanism, the thickness of the insulating layer should not be more
diodes, and achieve small junction areas without requiring fine lithography. Point-
contact diodes use a simple fabrication technique that can test several metal wires
for the same oxidized metal sheet at a high throughput (Periasamy, 2010).
However, such diodes are unreliable due to the mechanically unstable nature of the
unintended native oxide layer or impurities can significantly alter the diode
in experiments on the detection and mixing of infrared radiation (Riccius, 1984) and
made more reliably. Moreover, the diodes can be made using thin-film materials.
The insulator can be a grown oxide obtained by oxidizing a metal film to the desired
stack of metals and insulators provides the freedom to choose the barrier materials
independent of the metals. These techniques allow the MIM tunnel barrier to be
interfaces.
Despite the progress in the fabrication methods, rectennas based on MIM diode
lack the efficiency required for successful application in detectors or solar cells. A
major factor for the inefficiency is the large RC time constant of the rectenna circuit
5
that uses MIM diodes (Sanchez, 1978). A proposed solution for reducing the RC
value is to reduce the junction capacitance by making smaller diodes. However, this
2. Geometric diode
junction can show rectification due to interaction of carriers with the conductor
effect has been observed in small conductors having geometric asymmetry due to
through the device (Song, 1998). Songs device uses a four terminal configuration,
similar to a bridge rectifier, in which the AC signal is applied across two opposite
terminals and the DC voltage is tapped from an orthogonal pair of terminals. The
geometric diode reported here is a two terminal device and does not require ballistic
transport.
6
explained later, the relevant size requirement for an asymmetric-hourglass-shaped
geometric diode shown in Figure I-3 is that the neck-size should be smaller than lc.
Blocked
Channeled
Neck < lc
Shoulder
based the Drude model (Ashcroft, 1976, p.1). Free charge carriers in a conductor
move randomly at a velocity equal to the Fermi velocity. Assuming the conductor to
have perfect edges from which the charges can reflect specularly, on striking an
edge, the direction of motion for the charge is changed while maintaining the speed.
In the region to the right of the neck charges moving leftwards are deflected in the
opposite direction due to the vertical edge, while charges in the left region moving
rightwards collide with the slanting edge, and funnel through to the right of the
neck.
the effect of the geometry. Therefore, only on length-scale smaller than the inelastic
collision length (lin), a charge near the neck senses the geometric asymmetry
that occurs due to the local redistribution of the charge. However, on applying a
7
voltage, the magnitude of current varies depending on the direction of bias due to
from the edges, and the motion of charges through the neck without losing
directionality, can be expressed in terms of the wave nature of the charge particles
(Thouless, 1980). The quantum equivalent of the specular reflections is that the
edges define the shape of an electronic wavepacket and influence its diffusion. The
distance over which the wavepacket is able to diffuse elastically before undergoing
an inelastic collision is controlled by lin and the elastic diffusion length le. This
distance is the electron phase-coherence length (lc). Under the condition that the
charges have more wave-character than corpuscular lin >> le, the lc is obtained as
(Nimtz, 1988)
lc = lelin 3
Eq. I-1
the conductor must be smaller than lc. As explained in chapter VI, it is difficult to
meet this requirement using metal thin-films due to a small lin and thereby small lc.
Asymmetric characteristics of geometric diodes have been demonstrated using
their applicability for rectenna-based optical detectors and solar cells. For high-
unlike the parallel plate structure of the MIM diode, the geometric diode being
planar does not have a large capacitance. Second, the absence of a tunnel junction
8
results in a lower resistance than MIM diodes. These factors facilitate a lower RC
time constant for the rectenna circuit and a better impedance matching between the
C. Thesis outline
In chapter II, I develop a basic understanding for the properties of MIM diodes
MIM diodes are characterized by their resistance and the nonlinearity in their
efficiency is carried out. These include the RC time constant for the rectenna, the
impedance match between the antenna and the diode, and the area of spatially
coherent radiation that gets collected by one rectenna element. I estimate the
performance of rectenna-based detectors and solar cells using MIM and geometric
solar cell is developed and correspondence between the semiclassical and classical
theories is established.
9
In chapter VI, I analyze a technique for improving the infrared bandwidth and
rectifiers made from graphene. The simulations are based on the non-equilibrium
In chapter IX, I carry out a derivation that extends the semiclassical theory
10
CHAPTER II
1969), as infrared detectors (Grover, 2010), and in the emerging field of terahertz
electronics (Estes, 2005). MIM diodes are also used for detection and mixing of
The operation of the MIM diode is based on the quantum tunneling of electrons
between two metal electrodes that are spaced apart by several nanometers of
characteristics that depend on the shape of the barrier. To estimate this nonlinear
I(V), several analyses of MIM diodes have been carried out. To simplify analysis,
most of these calculate the tunnel probability using the WKB approximation
simplify the Fermi distribution of electrons in the metals (Chapline, 2007), (Guo,
11
low-barrier diodes as compared to a more rigorous solution. Here, I use the transfer-
matrix method (TMM) to model the MIM diode and explain the barrier height- and
1. Methodology
I assume a charge-free oxide region to determine the shape of the tunnel barrier
tunnel probability, which along with the Fermi distribution of electrons provides
the tunnel current. Consider the barrier shown in Figure II-1 with an arbitrary
potential profile.
12
Figurre II-1 Enerrgy-band dia agram for aan asymmettric (L R)
tunneel barrier. Here,
H i is th
he electron affinity of tthe insulatoor,
is the
t metal work
w functiion, and VD is the volltage applieed
acrosss the diode e. The Fermmi level of the left meetal electrod de
(EFL) is held fixeed while tha
at of the rig
ght electrodde varies with
applieed voltage across the diode. Thee rectangullar barrier is
modiffied by the e image-forrce barrierr lowering to give th he
effective profile (dashed).
(
An electron
e wiith total en
nergy E, ha
as an x-direected comp
ponent of eenergy Ex a
and a
transmiission prob
bability T(E
T x). Assu
uming an isotropic distributioon of elecctron
velocitie
es in the metal
m electrodes, the formula foor the tunnel curren
nt from thee left
(cathode
e) to the rig
ght (anode) electrode
e can be wrritten as (Simmons, 1963)
44mL e
h 3 0
J LR (VD ) = T(E x )dE x f L (E)
E){1 - f R (E + eVD )}dE
Ex
Eq.. II-1
electrod
des are give
en by
13
1
f L (E) =
E - E FL
1+ exp
kT
Eq. II-2
1
f R (E + eVD ) =
E - (E FL - eVD )
1+ exp
kT
Eq. II-3
In Eq. II-1 the inner integral is over all possible total energies E with incident
energy Ex for which there are filled states on the left and empty states on the right.
The outer integral then multiplies this total number of electrons with the
transmission probability T(Ex) and sums the product over all Ex. The net tunnel
current is the difference between the currents from the left to the right electrode
(JLR) and from the right to the left electrode (JRL), where JRL can be written in a
form similar to that of Eq. II-1. Assuming effective masses in each metal region, mL
= mR = m0, where m0 is the electron rest mass, the net current is given by
4m0 e
h 3 0
J(VD ) = J L R - J R L = T(E x )dE x {f L (E) - f R (E + eVD )}dE
Ex
Eq. II-4
wavefunction and its first derivative is applied at each interface. Unlike the basic
2x2 matrices need to be multiplied, the approach I use combines all the continuity
equations into a (2N+2) by (2N+2) near-diagonal square matrix (Probst, 2002). This
14
Applying the conditions that the amplitude of the incoming wave is unity (
A 0+ = 1 ) and that there is no reflected component in the N+1th region ( A N +1 = 0 ),
k N +1 | AN+ +1 |2
T(E X ) =
k0 | A0+ |2
Eq. II-5
For calculating T(Ex), I use an adaptive step size for Ex based on the slope of
T(Ex). This helps preserve accuracy when required without having a very fine grid
The effective mass of the electron in the insulator (me) is assumed to be equal to
the rest mass (m0). This assumption is made in the absence of a more accurate
estimate. For crystalline semiconductors, the me can be obtained from the band
structure (Kittel, 1996, p.209). However, for the grown or deposited amorphous
The shape of the potential barrier is determined by the work function of the
metals, the electron affinity of the insulators, and the applied voltage. In addition,
an electron in the vicinity of a metal experiences an image potential that causes
e2 1 1
Vimage (x) = +
16i 0 x x0 x
Eq. II-6
15
moving near a metal plane but its application to tunneling, which is a quantum
mechanical phenomenon, has been questioned (Hartstein, 1978). For the quantum
treatment of image potential, a number of theories have been proposed (Puri, 1983),
(unji, 1991). Due to the lack of consensus in what has been proposed in these
models, I chose to use the classical result of Eq. II-6. From the results reported by
unji it can be seen that the barrier lowering obtained from the quantum
mechanical image potential is smaller than that from the classical image potential.
For tunnel barriers with a high dielectric constant (i), Vimage is small, which means
that the difference between classical and quantum barrier lowering is minor. Hence
the choice of using the classical result for the image force is reasonable.
conduction through defects, surface states and charge build-up at the interfaces
that can affect the current. Scattering of the electrons in the insulator also needs to
method (TMM), with those obtained from the WKB approximation, which is dated
but has been used recently (Chapline, 2007), and the more current quantum
EFL is the Fermi level referenced to the bottom of the conduction band in the left
metal. The electron wavefunction for Ex < max, where max is the highest potential
on the modified barrier as shown in Figure II-1, decays with increasing x inside the
barrier. The T(Ex) rises sharply with increasing energy, as seen in the plot of Figure
II-2(a). When the electron energy (Ex) rises above the barrier (max), the
16
wavefun
nction beco
omes oscillatory and
d the transsmission p
probability stays closse to
unity. For Ex > max, therre are resonances in
n T(Ex) du
ue to interrference off the
wavefun
nction insid
de the insu
ulator.
The
e TMM calculation of T(Ex) is in cclose agreeement wiith the W
WKB
approximation fo
or Ex < EFL (10 eV
V). Howev
ver, at hiigher enerrgy the W
WKB
overestiimates the
e transmisssion proba
ability and gives T(E
Ex) = 1 for Ex > maxx. As
shown in
i Figure II-2(b), thiss results in
n a significa
ant deviatiion of I(V) characteriistics
on usin
ng the WK
KB method. On the other
o hand
d, in both the figurees, there iis no
differen
nce between the resu
ults obtain
ned from tthe TMM and the Q
QTBM, botth of
which give
g solutions to the Schrdinger
S r equation..
17
3. Com
mparison with expe
erimental characte
eristics
Sim
mulated and
d experime
ental chara
acteristics of two asy
ymmetric M
MIM diodess are
compare
ed in Figure II-3. Th
he diodes are
a made ffrom sputteered insula
ator and m
metal
layers and
a the dim
mensions shown
s are the targetted thickneess of the iinsulators. The
The
e simulated
d and the experiment
e tal curves match welll without the use off any
fitting parameters
p s besides the
t choice of
o effectivee mass equ
ual to the rrest mass. The
are eith
her book va
alues or exttracted from
m measureements (Ph
hiar, 2007)..
Metal Workfun
nction Insulatorr Electron affinity Dielectric
(eV)) (eeV) constant
Nb 4.33
3 Nb2O5 44.23 25
NbN 4.7 Ta2O5 3.83 20
18
B. Single-insulator (MIM) diodes
Using the simulation methodology described above, I now analyze the properties
affects the tunnel current is facilitated by comparing two diodes with different
current that arises due to the Fermi distribution is also developed. I extend this
analysis to show how the variation of current with temperature decreases for large
barrier-heights.
As given by Eq. II-4, the Fermi distribution and the transmission probability of
electrons tunneling across the barrier together determine the tunnel current. For
two diodes that differ only in their barrier heights, the Fermi distribution is
tunnel current on the barrier-height, consider a low-barrier (0.5 eV) and a high-
barrier (1 eV) diode. Referring to Figure II-1, max is closer to EFL in the low-barrier
case. Therefore, as seen in Figure II-2(a), the sharp peak in T(Ex) near Ex = max,
adds significantly to the tunnel current only in the low-barrier diode. In the high-
barrier diode, this rise in T(Ex) is insignificant, as the Fermi distribution results in
than for the low-barrier. This leads to a smaller tunnel current in the high-barrier
diode.
To show that the contribution of electrons near max is dominant only in the
low-barrier case, I compare the tunnel currents calculated using the T(Ex) obtained
from the plane-wave solution and the WKB approximation. This comparison
19
inaccura
ate estima
ate of T(Ex) near the peak
p of thee barrier. T
The sensittivity decreeases
with inccreasing ba
arrier heig
ght as the contributio
c on of electrrons near tthe peak off the
Howeve
er, for the high-barrie
h er diode (1
1 eV) the W
WKB and p
plane-wavee results arre in
agreeme
ent.
20
The above results show that in the high-barrier case, the electrons near max
have a smaller contribution to the total current. They also show the limited validity
tunnel current, the Fermi distribution also determines the temperature dependence
of the tunnel current. To explain the variation in tunnel current with temperature, I
analyze the 0.5 eV symmetric tunnel barrier of Figure II-4, biased at 0.3 V. The
tunnel current at any Ex is the product of the transmission probability and the
Figure II-5, I plot the transmission probability T(Ex), the Fermi distribution fL(Ex)
and the argument of the outer integral in Eq. II-4, J(Ex), calculated using TMM for
77 and 300 K.
21
Figurre II-5 Contrribution to total
t curren
nt density by y electrons at
different energie es J(Ex) (A A/cm2/eV) ffor a symm metric diod de.
Transsmission prrobability, T(E T x), is ob btained from m the plan ne-
wave solution off the Schrd dinger equa ation. Fermii distributioon
on thhe left and d right metal electrod des, fL(Ex) and fR(Ex) ,
respectively, is needed
n to de
etermine J(E(Ex). The simmulated diod de
has a barrier heiight of 0.5 eV e and a th hickness of 2 nm, and is
biasedd at 0.3 V. Since fL at a 77 K drrops faster with energ gy
compa ared to fL att 300 K, J(E Ex) follows thhe same treend. A seconnd
peak is seen in J(EJ x) at 300 0 K as T(Ex ) rises sharrply near th he
top off the barrierr, briefly compensating g the effect of decreasin
ng
fL. Th
he increase ini area und der J(Ex), froom 77 to 3000 K, leads to
the te
emperature dependence e of tunnel ccurrent.
The
e J(Ex) is th
he currentt density as a functioon of the x
x-directed eelectron en
nergy
and its integral giives the tottal currentt density. T
The sharpeer drop of tthe fL(Ex) a
at 77
K as com
mpared to 300 K is ev
vident. The
e fR(Ex) is a left-shifteed version of fL(Ex) on
n the
using th
he correspo
onding fL an
nd fR at both 77 K an
nd 300 K.
The
e J(Ex) curv
ve at 77 K indicates that a sig
gnificant coontribution
n to the tu
unnel
Howeve
er, at room temperatu
ure the Ferrmi functioon has a la
arger sprea
ad with smaller
slope, and so a sh
harp rise in
n T(Ex) nea
ar the peak
k of the ba
arrier (10 < Ex < 10.4
4 eV)
adds to
t the tota
al current. As the temperatu
t ure rises, tthe spread
d in the Fermi
22
distribution causes a larger contribution from the high (above EFL) energy electrons.
Electrons at higher energies have a greater probability of tunneling across the
energy electrons and their greater probability of tunneling result in a larger tunnel
current.
As the barrier-height of the diode is increased, the peak of the barrier (max)
shifts away from EFL. This reduces the Fermi distribution near max, which
decreases the contribution of the second peak in J(Ex). Hence the temperature
barrier diodes. To demonstrate this trend, I calculate the tunnel current at various
from 0.2 to 2 eV. The insulators are 2 nm thick and the temperature is varied from
normalized by its value at 50 K for each barrier height and is plotted in Figure II-6.
The variation with temperature is larger for smaller barriers. As shown in Figure
II-5, this happens because the tail of the Fermi distribution is significant near the
23
Figurre II-6 Normmalized currrent densitty vs. temperature for a
range n a symmeetric MIM tunnel diod
e of barrierr heights in de
biasedd at 0.3 V. The thicknness of the insulator iis 2 nm. Thhe
tempe erature deppendence iss larger forr smaller bbarriers. Th
he
changge in currrent with barrier-heiight increa ases as th he
tempe erature rise
es.
The
e temperatu
ure depend
dence of cu
urrent can be reduced
d to a power law rela
ation
of the ty
ype
y = ax b + c
Eq.. II-7
The
e data of Fiigure II-6 is fitted to
o Eq. II-7 a
and compa
ared with the analytically
obtained
d temperatture depen
ndence of th
he form
J(T 0) c1kT 1
= = 1+ (c1kT)2 +
J(T = 0) sin(
c1kT) 6
Eq.. II-8
actual power-law
p temperature depen
ndence of current att low barrrier heightts is
24
Figurre II-7 Poweer-law coeffficient b vs. barrier-heiight from th
he
curve fit for temp
perature dependence off tunnel currrent. For th he
2 nm m diode un nder considderation, on nly for barrrier heighhts
greateer than 0.8 eV does thhe temperatture depend dence becomme
quadrratic as pred
dicted by th
he analytica al formula in
n given in E
Eq.
II-8.
The
e coefficien
nt c in Eq. II-7 is 1 within
w a 5%
% margin, in
ndicating tthe accuracy of
tunnel currents.
c Only
O for barrrier heigh
hts greater than 0.8 eV does thee behavior sstart
to exhib
bit the quadratic temperature dependence
d e given in E
Eq. II-8. Th
his observa
ation
is depen
ndent on se
everal variiables inclu
uding the ttemperaturre range in
n considera
ation
and the
e width of the barrie
er. Nonethe
eless, it signifies tha
at the anallytical form
mula
does nott provide th
he correct temperatu
t ure dependeence for low
w-barrier d
diodes.
C. Shortcom
mings of MIM
M diode
es
Elia
asson (Elia
asson, 2001
1) has exte nalyzed thee possible variations of a
ensively an
single-in
nsulator MIM
M diode. The I(V)) characterristics of tthe diode depend on
n the
shape of the tunne
el barrier, which is determined
d d by the meetals and iinsulators used
25
be optimized are the differential resistance and responsivity of the diode. The
antenna, and is achieved by keeping the barrier heights low (Eliasson, 2001). A high
1978). Here I analyze these characteristics at zero bias, which reduces the
complexity of comparing the resistance and responsivity of several diodes. At zero
bias, the responsivity is determined by the degree of asymmetry in the shape of the
tunnel barrier, which causes the asymmetry in the I(V) curve. In Figure II-8(a)&(b),
I plot the responsivity and resistance vs. the difference in barrier height on the left
(L)and the right (R). Experimentally, this can be achieved by varying the metal
on the left while keeping the insulator and the metal on the right fixed.
26
Figurre II-8 (a) Responsiviity and (b)) resistancee vs. barriier
asymm metry for siingle-insula
ator diodes. The diode tthickness an
nd
the le
eft barrier height
h (L) are varied
d while the right barriier
heigh
ht (R) is keept fixed at 0.1 eV. Thhe responsivvity increases
with increase
i in asymmetry y but saturaates for hig
gh L-R. FFor
the same asymm metry, the responsivitty is largerr for thick ker
diodes. Increasinng asymmetry and inccreasing thiickness, both
lead to
t larger ressistance.
As explained in chapte
er V, an asymmetrric I(V) iss necessary
y for self--bias
generation and effficient rectification. The zero-b
bias respon
nsivity is a
an indicatoor of
this asy
ymmetry. As
A seen in Figure
F II-8
8(a), for a fiixed asymm
metry, the responsiviity is
increasiing thickn
ness, so do
oes the re
esistance a
as shown in Figuree II-8(b). IIn a
between
n the anten
nna and the
e diode beccomes worsse.
27
The above characteristics are not representative of trends at a non-zero bias.
The non-zero bias responsivity of the diodes described above may be sufficient for
the operation of a biased detector. Another possibility is that the asymmetry in the
28
CHAPTER III
this chapter, I analyze two mechanisms that can improve the performance of multi-
insulator (MIIM) diodes based on these mechanisms are simulated and their
which may significantly alter the diode behavior. In another MIIM configuration
understanding of these effects and use them to design MIIM diodes with improved
A. Simulation methodology
In chapter II, I gave the framework for simulating MIM diodes using the
insulators play an important role in determining the voltage drop across each
insulator layer. To determine the energy-band profile at a certain bias (VD), I apply
the condition for continuity of the electric displacement vector at each insulator
xj j
V j = (VD Vbi )
xj j
Eq. III-1
where, xj and j represent the thickness and dielectric constant, respectively, of the
jth layer, and eVbi ( = L R ) is the built-in potential.
e2 1 1
Vimage(x) = x
+ L
16 0 (x')dx'
0 x (x')dx'
Eq. III-2
K
where, L = x j with K being the number of insulator layers. The integrals in the
1
denominator represent the effective distance of an electron from the left or the right
accounts for the resonant tunneling but assumes an empty well. However the
resonant level has a certain probability of being filled and is likely to be occupied if
its energy is close to or less than the higher metal-Fermi level. This correction is not
30
B. Double-insulator configurations
To obtain a high responsivity and low resistance diode, one can design an MIIM
increasing voltage) for electrons tunneling from the higher Fermi level (Matsumoto,
1996). Both these mechanisms can occur in the same diode and the overall
asymmetry of the I(V) curve is regulated by the one that dominates. I examine these
parameters for the oxide were listed in chapter II. The work function of tungsten is
4.55 eV (Camp, 1965). This choice of materials and dimensions is not optimized for
between the resonant tunneling dominant in MIIM1 and the step change dominant
in MIIM2.
For the two diodes, the conduction band profiles under positive and negative
bias are shown in Figure III-1. A quantum well is formed in both the MIIM diodes
under positive bias (a) & (b). However, only in the MIIM1, the quantum well is wide
enough to have a resonant energy level. On the other hand, under negative bias (c)
& (d), the step barrier-profile leads to an abrupt change in the tunneling distance
for the electrons near the Fermi level on the right metal-electrode. The Fermi level
31
Figurre III-1 Ene ergy-band profiles
p for the resona ant and steep
MIIMM diodes. Forward
F and d reverse b bias profilees are show wn
respectively in (a
a) and (c) forr the resona
ant, and in ((b) and (d) ffor
the sttep diode. The
T dotted lines
l show tthe profiless with barriier
lowerring. The diode
d param meters are given in T Table I. Th he
thicknness of the Nb2O5 layerr is the only y differencee between th he
two diodes.
In Figure
F III-2
2, I plot the
e transmission proba
ability T(Ex) and the ccurrent den
nsity
magnitu
ude of the bias voltag
ges, the negative biass curves arre offset along the en
nergy
axis from those att positive bias
b by app
proximately
y 0.4 eV. T
The fact th
hat the offsset is
mirrored along th
he verticall, the diffe
erence in Fermi lev
vels betweeen the barrier
profiles of opposite
e bias is 0.4 eV and hence
h the ooffset. For llow electroon energiess, the
transmiission prob
bability T(E
Ex) for the step
s (MIIM
M2) diode, rrepresenteed by curvees (b)
is becau
use it is eassier to tunn
nel across a thin barrrier as com
mpared to a thick one.
32
Figurre III-2 Ellectron traansmission probabilityy T(Ex) an nd
currennt density J(Ex) vs. energy
e for the resonaant and steep
MIIMM diodes off Figure IIII-1. A sh arp resona ance peak is
observved in the resonant diiode under forward bia
as due to th
he
formaation of a qu
uantum welll.
The
e resonant diode under positive
e bias (a) h
has a sharp
p rise in T
T(Ex) exceeeding
(b) at th
he resonan
nce peak. Except
E for the
t resona
ance peak, the T(Ex) n
near the toop of
insulato
or, the tra
ansmission
n probability exhibitss oscillatoory behavioor for all four
tunnelin
ng and is sinusoidal
s through
t th
he remainin
ng region oof transmisssion abovee the
33
tunnel barrier causing the oscillations in T(Ex). These oscillations modify the
probability of tunneling through the higher barrier to give the net T(Ex).
The above trends in T(Ex) influence the current density J(Ex). The area under
the resonant diode J(Ex) at positive bias (a) is greater than at negative bias (c).
Therefore the current in the resonant diode is greater at positive bias as shown in
the I(V) curve of Figure III-3 (a). For the step diode, the area under the J(Ex) curve
at negative bias (d) is greater than under positive bias (b). This asymmetry is also
seen in the I(V) curve shown in Figure III-3(a). Comparing the resonant and the
step diode J(Ex) curves, the narrow resonance peak in (a) is large enough to give a
current greater than that in the step diode under positive bias (b) but not enough to
exceed the current in the step diode under negative bias (d).
34
Figurre III-3 (a) Current
C den
nsity vs. volttage for the MIIM diod des
shown n in Figure III-1, an nd a compa arable asym mmetric-MIIM
diode. The step MIIM diode has hig gher curren nt magnitud de
underr negative bias due to o the directt tunneling g of electron
ns
acrosss the high-barrier. Th he resonannt MIIM diiode has th he
oppossite asymm metry in itss I(V) cha aracteristic, due to th he
formaation of re esonant quantum welll under p positive biaas.
Compparing these e with the asymmetric
a c-MIM diodee we see th hat
both the
t MIIM diodes
d have a smaller rresistance ((b) and largger
nonlinnearity (c) in
n their prefferred directtion of cond
duction.
In Figure
F III-3(a), I also
o compare the MIIM
M diodes w
with an asy
ymmetric-M
MIM
diode th
hat has the barrier heights
h corrrespondin
ng to the W
W-Nb2O5 in
nterface on
n the
35
left and the Ta2O5-W interface on the right and an insulator thickness of 2 nm. The
asymmetric-MIM diode is essentially the MIIM2 diode without the abrupt step in
the conduction band profile. This is confirmed by their similar current densities
under positive bias. However, under negative bias, the step change in tunnel
distance in MIIM2 causes a sharp increase in tunnel current. This difference is also
evident in the resistance and responsivity curves in Figure III-3(b) and (c) where,
under negative bias, the sharp increase in current for MIIM2 leads to a lower
significantly higher at zero bias but becomes comparable to the thinner diodes near
VD = 0.4 V. The large change in resistance also accounts for the higher magnitude of
with higher responsivity and lower differential resistance than an equivalent MIM
The above example shows that just changing the thickness of an insulator in an
MIIM diode made with the same pair of materials, can lead to different asymmetry
and nonlinearity. It does not suggest which of the mechanisms for achieving larger
nonlinearity is preferable. I have analyzed several MIIM diodes designed for
also be applied to barriers with more than two insulators (Korotkov, 1999).
The comparison of thick and thin double-insulator diodes shows that the bias
36
the currrent for thiis polarity can become larger th
han that un
nder negative bias. In
n the
absence
e of a reson
nant level, the step change
c in tunneling distance u
under nega
ative
bias cau
uses a larg
ger current than unde
er positive bias. Com
mpared to ssingle-insullator
a largerr responsiv
vity and a smaller
s resistance.
In Figure
F III-4
4, I comparre the resisstance vs. rresponsivitty at zero b
bias for sev
veral
resistan
nce values, the respon
nsivity of double-insu
d ulator diod
des is largeer than that of
single-in
nsulator diiodes. Therrefore doub
ble-insulattor devices are able tto achieve both
responsivity.
37
D. Interface stability
control over layer thicknesses. However, every combination of metals and insulators
ensure that the intended barriers are obtained in an experimental device. For the
MIIM diodes discussed in Section III-B, I have carried out a Gibbs free energy
analysis (Silberberg, 2007, p.666) for reaction between all interfacial pairs of
materials using the FACTsage web software (Bale, 2011). I analyzed each of the
The stability analysis for the Ta2O5-Nb interface that occurs in diodes 5 and 6 of
Figure III-4 is reproduced here. Using the FACTsage web software (Equilib-web),
the equation obtained for the reaction between Nb and Ta2O5 at 300 K and 1 atm is
Nb also forms several other sub-oxides that were not considered. To confirm
that the above reaction occurs, a Gibbs free energy calculation can also be made
where sys denotes the system represented by Eq. III-3 and the denotes change in
formation (H) and the entropy (S) data for the materials is obtained from the
CRC handbook. The calculation for Gsys is given in the table below.
38
Material H S T*S G multiplier
(kJ/mol) (kJ/mol/K) (kJ/mol) (kJ/mol)
Nb 0 0.036 10.8 -10.8 -5
Ta2O5 -2046 0.143 42.9 -2088.9 -1
NbO -405.8 0.048 14.4 -420.2 5
Ta 0 0.042 12.6 -12.6 2
On multiplying G by the multiplier and adding the values for all materials,
Gsys = 16.7 kJ/mol. For a spontaneous reaction to occur, the Gsys should be
indicates that a reaction does occur. One reason for this discrepancy can be the
individual reactants and products. Even if a spontaneous reaction were to occur, the
small Gsys implies that it would be slow. This is understandable from the fact that
39
CHAPTER IV
the MIM diode is intrinsically fast and allows rectification at optical frequencies. In
this chapter, I estimate the performance of a high-frequency rectenna using circuit
analysis applicable to both MIM and geometric diodes. In the next chapter, I model
Other than the signal strength, the semiclassical analysis shows that for large
photons or the frequency of the incoming wave. The results for the semiclassical
rectification theory (Torrey, 1964). The choice of the relevant model depends on the
diode I(V) curve and the magnitude of the AC signal applied across the diode
(Eliasson, 2001, p.10). For an AC voltage smaller than the voltage scale on which
the diode nonlinearity is evident, the square-law analysis is applicable. For larger
AC voltages the linear model is used, where the diode switches between high- and
harvesting, a small AC voltage is applied across the diode. This is due to the low
intensity of the incoming radiation and the small receiving area of the antenna.
40
Therefore, the analysis presented in this chapter is based on the square-law
rectification theory.
The derivation of the square-law rectification theory leading to the formula for
(Eliasson, 2001). Here I develop a deeper insight into the derivation and investigate
the factors that contribute to the rectenna efficiency. The requirements for
rectennas to operate as efficient detectors and photovoltaic (PV) rectifiers are
discussed and calculations for efficiency based on MIM and geometric diodes are
presented. Although these diodes are used as examples, the concepts given are more
generally applicable. I also provide design guidelines for rectenna solar cells and
An outline of the circuit analyses presented in this chapter and their key
41
A. Circuit analysis
diode is given by
V {t} = VD + V cos ( t )
Eq. IV-1
and the corresponding current, to second order, is given by the Taylor expansion
Eq. IV-2
V2 I {VD }
I diode = I {V {t}} = I {VD } +
4
2
V
= I {VD } + {VD } = I {VD } I DC
{VD }
2 RD {VD }
Eq. IV-3
in which the first component corresponds to the DC resistance of the diode Rdiode =
VD/I{VD} while the second component is the DC current generated due to
rectification. Without a DC load across the diode, the two currents are equal and
opposite and the net DC power in the diode is zero. The time averaged total power
V2
Pdiode = V {t} I {V {t}} = VD I {VD } VD I DC
{VD } +
2 RD {VD }
Eq. IV-4
where the first two terms correspond to the DC current while the third term gives
42
The efficiency () of a rectenna is determined by the combination of several
= asc j
Eq. IV-5
the efficiency of propagating the collected energy to the junction of the antenna and
the diode; c is the coupling efficiency between the antenna and the diode; and j is
the efficiency of rectifying the power received in the diode. The efficiency of the
was discussed in Chapter II. This sets the overall units of to be A/W implying the
AC power delivered to the diode resistance to the power of the antenna voltage
source. This ratio can be calculated from the analysis of the small-signal circuit of
equivalent and the diode by the parallel combination of a capacitor and a voltage-
dependent resistor.
43
Antenna Diode
RA
VA RD CD
~
Assuming that the input power is delivered to the diode as AC power, the power
R A RD
4
c =
PAC , RD
=
( R A + RD ) 2
PA 2
R A RD
1+ CD
( R A + RD )
Eq. IV-6
where, PA = VA2 (8RA ) . In the above equation, the numerator gives the impedance
and RD. The denominator in Eq. IV-6 determines the cutoff frequency of the
combination of the antenna and the diode resistance and the diode capacitance.
Above the cutoff frequency, the capacitive impedance of the diode is smaller than
the parallel resistance leading to inefficient coupling of power from the antenna to
44
The responsivity form of the overall efficiency, given by = ci, is a measure for
the performance of a detector and indicates the current or voltage produced per
efficiency (load) which is given by the ratio of the DC power delivered to the load
The IDC,load is calculated from the current division of IDC between Rload and
Rdiode, giving the expression for efficiency
load i2 PAc2
Eq. IV-8
From Eq. IV-8, the power-conversion efficiency depends on four factors, namely,
the diode responsivity, the strength of the AC signal that depends on the power
received by the rectenna, the impedance match between the antenna and the diode,
and the RC time constant of the circuit. Another component that decreases the
efficiency of the rectenna is the series resistance of the leads from the antenna to
the strength of the voltage source equivalent to the solar spectrum (100 mW/cm2)
45
radiation is proportional to the wavelength squared. Using a nominal RA = 100 ,
an order-of-magnitude calculation gives VA 1 mV. The low magnitude of the
voltage source agrees with the choice of using a small-signal circuit analysis.
approach that accounts for photon energy of the radiation. In the case of a PV
IV-8. Also, for a detector, even though the responsivity does not change with area,
the noise performance or the detectivity improves on increasing the receiving area
(Grover, 2010). Possible methods of increasing the area include using concentrators
or feeding the signal from an array of antennas into a diode. The second
However, the area from which radiation is fed into one diode cannot be made
determines the maximum area that can be coupled to a diode. Sources of interest for
rectennas are the sun for photovoltaics, recycled heat for thermo-photovoltaics,
humans or animals for infrared (IR) detectors in automotive, to name a few. All
temperature and the radiation emitted from different parts of the source is
use the example of the sun and calculate the distance on earth over which the solar
radiation is coherent.
46
The degree of coherence for a quasi-monochromatic source is given by the van
Cittert-Zernike theorem (Born, 1999, p.572) that calculates the correlation in field
extended source as shown in Figure IV-1. I apply this theorem to a black body
spectrum, which can be treated as quasi-monochromatic (Marchenko, 2007, p.321).
U(r2)
Rsun U(r1)
Rsun
2
1
r
Eq. IV-9
where is the wavelength of the source and r is the distance from it. The radius of
the sun is approximately 7x105 km and at = 500 nm, the far zone of the sun is at a
distance greater than 3x1021 km. The distance between the sun and the earth is
only 1.5x108 km (ls-e), implying that the earth is in the near-zone of the sun.
However, numerical calculation of the degree of coherence without using the van-
47
Cittert Zernike
Z th
heorem shows that th
he theorem is applicab
ble even in
n the near field
(Agarwa
al, 2004). Coherence
e in radiattion emerg
ges as closse as a few
w wavelen
ngths
away fro
om the sun
n.
Neg
glecting the
e variation
n in the brightness off the sun, tthe degree of coheren
nce is
given by
y
where, k = 2/ an
nd J1 is Be
essel functtion of firstt kind. Thee dependen
nce of on
n is
The
e angle ove
er which the
t radiatiion from tthe sun ha
as a degree of coherrence
1
= 2sin
2 1
kRsun
Eq. IV
V-11
48
From this the diameter of the circular area at earth over which sunlight is
0.16ls e
dc =
Rsun
Eq. IV-12
This gives dc = 19 m and a spatially coherent area of 283 m2. With the
increased area, the strength of the antenna voltage source discussed earlier is
and located 50 m away, I apply Eq. IV-12 to obtain an approximate value for the
dc = 90 m is obtained.
Efficient coupling of power from the antenna to the diode requires impedance
matching between them. Moreover, having a small RC time constant for the circuit
implies that the product of the antenna resistance (RA) in parallel with the diode
resistance (RD) and the diode capacitance (CD) must be smaller than the time period
(2/) of radiation incident on the rectenna. This ensures that the signal from the
antenna drops across the diode resistor (RD) and is not shorted out by CD. This
The parameters that can be varied to achieve these conditions are the diode
area, the antenna resistance, and the composition of the diode. A less resistive diode
will give a higher coupling. Therefore, for this analysis I choose the Ni/NiO(1.5
nm)/Ni MIM diode, which has an extremely low resistance at zero bias and was
49
used in several high-frequency rectennas (Fumeaux, 1998), (Wilke, 1994), (Hobbs,
2007). The coupling analysis for any other practical diode is likely to be worse than
the present case. Here I am disregarding the fact that low resistance and/or
semiclassical resistance accounts for the energy of the optical photons. This concept
I choose a nominal antenna impedance of 377 and vary the diode area. In
Figure IV-4, the coupling efficiency c is plotted vs. the diode edge length for a
The semiclassical resistance is lower than the classical resistance and gives a
higher c. The peak in both the curves occurs at the same edge length, and is an
outcome of the balance between the needs for impedance matching and low cutoff
frequency. In either case, the maximum efficiency is much smaller than unity.
50
Figurre IV-4 Effect of varyinng the diodee size on th
he antenna to
diode coupling effficiency. Th
he peak in tthe efficienccy occurs du ue
e tradeoff be
to the etween impedance mattch and cutooff frequenccy.
The resistance
r of
o the Ni-N NiO(1.5 nm))-Ni diode is calculateed
from its
i simulate ed I(V) curvee classically
y and semiclassically (ffor
a phooton energyy Eph=1.4 eV V, or air=0 .88 m) as discussed in
Chaptter V. A Ni-NiO barriier height oof 0.2 eV iss used in th he
simullation (Hobbbs, 2007).
To understand
u d the coup
pling efficiiency betteer, one can
n separatee the effeccts of
impedan
nce matchiing given by
b the num
merator (ideeally RD/RA=1) and cu
utoff frequ
uency
given by
b the den
nominator (ideally (RA||RD)C
CD=0) in E
Eq. IV-6. U
Unity coup
pling
51
Figurre IV-5 Ante enna-to-diod de coupling
g efficiency as a functioon
of dio
ode edge lenngth, separa ating the eff
ffect of impeedance matcch
from cutoff frequ uency for two antenna a impedancce values: ((a)
RA=37 77 , and d (b) RA=1 10 k. RP denotes the paralllel
combiination of RA and RD. TheT curves labeled RPCD show th he
coupliing efficien
ncy when only the ccutoff frequ uency is th he
limitiing factor, and those labeled RD /RA show the couplin ng
efficie
ency when only the impedance match is the limitin ng
factorr. The overa
all efficiency
y remains sm
maller than n either curv
ve.
The
e tradeoff between
b im
mpedance match
m to th
he antenna
a, for which
h a small RD is
desired,, and a high cutoff fre
equency, fo
or which a small CD is desired, iis fundameental
for para
allel plate devices. Ev
ven increasing the an
ntenna im
mpedance, if feasible, does
not help
p. In such a case a higher
h RD can be accoommodated
d, allowing the diode area
to be sm
maller, and
d resulting in a desira
able smalleer CD. Unffortunately
y the higheer RA
improve
ement in the
t couplin
ng efficienccy. In otheer words, to improv
ve the coup
pling
52
curve shifts to the right and RD/RA curve shifts to the left. The condition under
which the constraints simultaneously lead to a high coupling efficiency is obtained
by combining
RD 2
( RA || RD ) CD << 1 & = 1 RDCD <<
RA
Eq. IV-13
For the model Ni-NiO-Ni diode discussed above, this condition is not satisfied
for near-IR-light frequencies ( = 0.88 m), where 2/=9.4x10-16 s is much smaller
independent of the diode area and is determined solely by the composition of the
MIM diode. As already noted, the Ni/NiO/Ni diode is an extremely low resistance
diode and NiO has a small relative dielectric constant (r) of 17 at 30 THz. Even if
one could substitute the oxide with a material having comparable resistance and
lower capacitance (best case of r=1), the RDCD would still be off by an order of
result from a breakdown-level current density of 107 A/cm2 at, say, 0.1 V, giving a
resistance of 108 -cm2. A near-ideal capacitance would result from a vacuum
The resulting RDCD would be ~10-15 s, again too large for efficient coupling at visible
wavelengths.
On the other hand, the planar structure of the geometric diode results in a very
elements: the capacitance between the arrow-shaped conductor on one side of the
neck and the square area on the other side and the quantum capacitance of the neck
region. The total capacitance is calculated to be 0.06 fF (Zhu, 2011). Since the
53
resistance of graphene can be adjusted by doping, the impedance of the diode can be
tailored to match the antenna impedance, which is about 200 ohms. For these
values the time constant for the geometric diode rectenna is RDCD=7x10-15 s. These
calculations are based on an initial fabricated device and not for an optimized
structure or the best graphene possible. Despite that, the RC time constant for the
geometric diode is a factor of 10 lower than that of the Ni-NiO-Ni MIM diode, which
implies that the geometric diode under consideration is not RC limited for infrared
imposed by Eq. IV-13 is easier to meet. The RDCD can also be artificially reduced by
compensating the capacitance of the MIM diode with an inductive element, but this
D. Rectenna applications
In this section, I use diode characteristics obtained from DC I(V) curves, and
project the performance capability of rectenna-based IR detectors and solar cells.
1. Infrared detector
for use in automobiles. Adequate performance and low cost has led bolometers (Yon,
2003) to be popularly used in these systems. More recently, the suspended thermo-
diode technology (Reinhart, 2009) has shown promise in providing a low-cost FIR
54
rectennas with the currently prevalent options, and examine the noise equivalent
automotive night vision is to have a 10 m detector with an NETD of less than 300
mK (ADOSE, 2008).
I noise BW
NETD =
dP dT
Eq. IV-14
limit of Plancks law cannot be used to calculate dP/dT (Brown, 2004) as that
and in the spectral range of 8 to 14 m, dP/dT = 3.185 nW/K. The effective area over
which the antenna receives radiation is chosen to match the area of a bolometer
element.
50% (as = 0.5) (Gonzlez, 2005), the antenna to diode coupling efficiency (c), and
the current responsivity of the diode (i). I calculate the responsivity of the detectors
17 is used for NiO. The capacitance of the geometric diode is estimate from a strip-
55
line and a quantum self-capacitance (Zhu, 2011). The resistance and responsivity
The responsivity of the geometric diode is higher due to a lower capacitance and
rectennas based on the geometric and MIM diode with the suspended thermo-diode
and bolometer.
Suspended
Rectenna Rectenna Bolometer
thermo-diode
Geometric MIM (Yon, 2003)
(Reinhart, 2009)
NETD 34 mK 137 K < 300 mK 56 mK
Receiving
35x35 m2 35x35 m2 225 m (pitch) 35x35 m2
area
Detector
0.2 A/W 5x10-5 A/W 150 V/W
responsivity
Noise
4 pA/Hz 4 pA/Hz
current
The above analysis for the rectennas is for zero-bias across the diode. This
removes the diode shot noise and the noise due to the biasing power supply, thereby
56
minimizing noise current. In situations where the gain in responsivity on operating
traveling-wave structure (Estes, 2006), (Grover, 2010). This design circumvents the
restrictions imposed on the MIM diode by the coupling efficiency. Akin to a
transmission line where the geometry determines the impedance, the distributed
RC enhances the coupling between the antenna and the traveling-wave structure.
However, losses in the metallic regions of the waveguide limit its efficiency as the
frequency approaches that of visible light. A detailed analysis for the traveling-wave
2. Solar cell
The efficiency of a rectenna solar cell given by Eq. IV-7 assumes a bias across
the diode. Another approach to get a better estimate for the efficiency is to consider
a self-bias voltage developed by the rectifier. The diode parameters that are used in
the efficiency calculation depend on this DC voltage (VDC). As with the efficiency,
the DC voltage depends on the power incident on the rectenna (Pin).
requirements of the rectenna and analyze the efficiency of a rectenna solar cell from
the diode perspective. The Norton equivalent circuit for the solar cell at DC is
shown in Figure IV-6. Assuming that the rectenna responsivity in Eq. IV-5 depends
only on the diode responsivity j = , the current source due to rectification, for a
57
V2
I DC = {VD }
2RD {VD }
Eq. IV-15
V2
VD = ( Rdiode || Rload ) {VD }
2RD {VD }
Eq. IV-16
The operating point for the diode is determined by the load resistance (Rload).
For maximum transfer of DC power to the load, Rdiode>>Rload, giving
V2
VD = Rload {VD }
2 RD {VD }
Eq. IV-17
to obtain the DC voltage for a given diode I(V) characteristics and incident power.
Iload
+
IDC +
Rdiode VD Vload
Rload
-
-
For the set of constraints outlined above, the power conversion efficiency for the
58
V2
VD {VD }
I loadVload 2 RD {VD } VD {VD }
load = = =
Pin
(
1 + VD {VD }) V 2
(1 + VD {VD } )
2 RD {VD }
Eq. IV-18
The above equation for efficiency is not independent of incident power, as there
value of |VD| which is smaller than unity as discussed from a quantum efficiency
perspective in the next chapter. Therefore, the maximum efficiency for a small-
The results presented here are obtained from a circuit analysis that uses
efficiency is an electron per photon, which imples that max=e/. Therefore, for at
high frequencies, a more accurate model for the rectenna solar cell is given by the
semiclassically derived diode parameters and I(V) curve under illumination. This is
The above calculation for the solar cell efficiency assumed rectification of the
total power available in the solar spectrum. An actual rectenna would be limited by
the bandwidth received by the antenna and the central frequency it is designed to
operate at. The spectral irradiance of the sun received on the earth can be written
59
The
e intensity distributio
on normaliized to thee maximum
m at 633 nm
m is plotteed in
wavelen
ngth of 633
3 nm allowss efficient energy-har
e rvesting.
As calculated
c in Section
n IV.C, th
he wavelen
ngth at wh
hich a recttenna operrates
efficienttly is deterrmined by
y the RC time consta
ant of the diode. In Figure IV
V-8, I
than 4 m,
only 1%
% of the tottal solar en
nergy is ha
arvested.
60
Figurre IV-8 Fracction of sola
ar energy ass a function
n of minimu
um
wavellength that can be harv vested efficiiently.
From
m Figure IV-8, it is evidentt that 60% of the solar eneergy existts at
wavelen
ngths less than 0.8 m.
For an
n infrared solar cell, operating above 0.8 m,
40% of the
t total so
olar energy
y is availab
ble for harv
vesting.
F. Ultimate
e efficienc
cy of recte
enna solarr cells
The
e maximum
m efficiency
y with whicch the enerrgy radiateed by a bla
ack body soource
can be converted
d to electtrical enerrgy is giv
ven by th
he Landsb
berg efficiiency
(Landsb
berg, 1979)). In this thermodyn
t namic mod el for efficciency the entropy off the
photonss generate
ed at a hiigh tempe
erature is discarded
d at a low
w tempera
ature
depende
ence for the
e efficiency
y as shown
n in Figure IV-9.
61
Figurre IV-9 Therrmodynamiccally-limited d maximum m efficiency of
a solaar cell as a function of tempera ature of thee black bod
dy
sourcee. The solarr cell is assu
umed to be at 300 K. A
Also shown is
the peeak wavelen ngth of the black bodyy spectrum as a functioon
of tem
mperature.
The
e Landsberrg efficienccy is based
d on a theermodynam
mic model for black b
body
to recten
nnas (Cork
kish, 2002). In the inttermediatee absorber model (de Vos, 1992)), the
radiatio
on is assum
med to be converted
d into heatt by the an
ntenna an
nd then Ca
arnot
at low frequencies
f s where the
e noise spe
ectrum of a resistor iis equal to the black b
body
spectrum
m (Dicke, 1946).
1
low and
d depends on
o the tem
mperature of
o the heatt source. T
This fact iss overlookeed by
62
researchers who propose to harvest the energy radiated out by the earth at night
(Midrio, 2010). Energy can be harvested by a solar cell only at a temperature lower
than that of the source. Again, the efficiency depends on the temperature ratio of
63
CHAPTER V
In the circuit analysis presented in the previous chapter, the resistance and
responsivity of the diode were used to calculate the efficiency of the rectenna.
Classically, differential (or small signal) resistance is the inverse of the first
derivative of current with respect to voltage and the responsivity is proportional to
the ratio of the second to the first derivative. These calculations are based on the
DC I(V) characteristics of the diode. However, only at a relatively low frequency can
energy of the incident photons (Eph = ) exceeds the voltage scale times e of the
is required (Tien, 1963). At high frequencies, the quanta of energy of the incident
(Dayem, 1962).
The qualification of high frequency is relative to the voltage scale on which the
Dayem, the voltage scale on which the diode nonlinearity is apparent corresponds to
the the superconducting bandgap of the metal films ~ 0.1 meV. Photons in this
energy range fall under microwave frequencies. For optical frequencies, the relevant
energy scale is greater than 100 meV, which is also the voltage scale for the
the photon with the electron is termed as photon-assisted transport, which can be
64
analyzed using several approaches (Platero, 2004). Here, I follow the derivation
given by Tien and Gordon (Tien, 1963) for a superconducting junction and
generalized for tunnel devices by Tucker (Tucker, 1978). I examine the properties of
between the antenna and the diode. Also, the bandwidth limitation due to the
antenna, for a constant incident power the amplitude of the AC signal may no
longer be smaller than the voltage scale over which the diode nonlinearity becomes
that a 1st order approximation can be applied to the results derived. A summary of
65
Correspondence and comparison of circuit model and semiclassical
rectification theory
Both the first-order and higher-order semiclassical models are obtained from
A. Semiclassical theory
diode is
Vdiode = VD + V cos ( t )
Eq. V-1
Fermi level on one side of the tunnel junction while holding the other side at a fixed
potential, as shown in Figure V-1. Effectively, the low frequency AC signal results
in an excursion along the DC I(V) curve around a bias point given by VD.
66
E
electron tunneling
VD
V
x
dependent term in the Hamiltonian H for the contact (Tien, 1963), written as
H = H 0 + eV cos ( t )
Eq. V-2
( x, y, z, t ) = f ( x, y, z ) eiEt /
Eq. V-3
The harmonic perturbation in Eq. V-2 leads to an additional phase term whose
( x , y , z , t ) = f ( x , y , z ) e iEt / exp (i / ) dt ' eV cos ( t ' )
t
Eq. V-4
Integrating over time and using the Jacobi-Anger expansion, the wavefunction
can be written as
67
+
eV
( x, y, z, t ) = f ( x, y, z ) J n ei( E +n )t /
n =
Eq. V-5
where Jn is the Be
essel functiion of order n. The m
modified wa
avefunction
n indicates that
an electtron in the
e metal, previously located
l at energy E,, can now be found at a
multitude of energ
gies separa
ated by the
e photon en
nergy ( E ph = ) as sh
hown in Fiigure
V-2. Th
he amplitud
de of the electron
e be
eing found at energy
y E+nEph iis given by
y the
Bessel function
f off order n, where
w n corrresponds to the num
mber of phootons absoorbed
or emittted by th
he electro
on in a multi-photo
m on process. The tim
me depen
ndent
wavefun
nction is normalized
n since the infinite su
um of the ssquare of B
Bessel term
ms is
wavefun
nction and therefore to
t the squa
are of the B
Bessel funcction.
A heuristic
h ex
xplanation for the efffect of thee wavefun
nction mod
dulation on
n the
tunnel current
c is based on the
t fact tha
at all such single-electron statees will und
dergo
voltage,, there is an
a addition
nal voltage nEph/e tha
at is applied across th
he diode wiith a
68
weighting factor Jn2() where =eV/Eph. The DC current under illumination is then
given by
I illum (VD ,V ) = J n2 ( )I dark VD + n
e
n =
Eq. V-6
frequency.
Apart from the DC component of the tunnel current given by Eq. V-6, there is a
dependent formulation for the illuminated tunnel current (Tucker, 1979), the first
harmonic is given by
I = J n ( ) J n+1 ( ) + J n1 ( ) I dark VD + n
e
n =
Eq. V-7
From Eq. V-6 and Eq. V-7 one can calculate the resistance and responsivity
V SC I illum
RDSC = ; i =
I 1
V I
2
Eq. V-8
and Iillum is the incremental DC current due to the illumination. The Iillum =
Iillum(V)-Iillum(0) and the approximation used for Bessel functions are J0(x) 1-x2/4,
69
From here on up to the end of section V.B, a small-signal AC voltage is assumed
such that Bessel function terms only up to first order in n are needed. From Eq. V-7,
to first order in n(=-1, 0, 1), RDSC is given by (Tucker, 1979)
SC 2 E ph e classical 1
RD =
I dark (VD + E ph e) I dark (VD E ph e) I
Eq. V-9
The semiclassical responsivity is similarly found out from the first order
approximation of Eq. V-6 and Eq. V-7, and is given by (Tucker, 1979)
Eq. V-10
Again, Eq. V-10 in the limit of small photon energies leads to the classical
formula for responsivity given by 1/2 the ratio of second derivative of current to the
quantities given by Eq. V-9 and Eq. V-10. As shown in Figure V-3, the semiclassical
resistance is the reciprocal of the slope of the secant between the currents at VD
/e, instead of the usual tangent at VD for the classical case.
70
Figurre V-3 The semiclassiccal resistan nce is obtain
ned from th
he
inversse of the slope
s of seccant betwe en Idark(VD /e). AAn
experrimentally measured
m Idark for a Nb//Nb2O5(3 nm
m)/Ta2O5(1.7
75
nm)/N
NbN double insulator diode
d is used
d.
The
e semiclasssical responsivity refflects the cchange in the slope of the seccant,
rather than
t the cu
urvature at
a the bias point for the classiccal case. In
n Figure V
V-4, I
plot the
e semiclassical resista
ance and re
esponsivity
y at zero bias vs. the photon en
nergy
resistan
nce of the diode
d decreases as the
e slope of th
he secant iis larger th
han the slop
pe of
the tang
gent.
The
e AC resista
ance given by Eq. V-9
9 is the dioode impeda
ance seen b
by the anteenna.
The dro
op in the AC
A resistan
nce at high
h frequenciies improv
ves the imp
pedance m
match
between
n the ante
enna and the
t diode. As shown
n in Chaptter IV, thee semiclasssical
diode re
esistance le
eads to a hiigher coupling efficie ncy.
The
e change in
n the slope
e of the se
ecant decreeases as itt becomes more verttical,
responsivity apprroaches th
he limit off e/, wh
hich is th
he maximu
um achiev
vable
responsivity corre
esponding to
t a quantu
um efficien
ncy of 1. T
Thus, even a diode wiith a
The
e Tien-Gord
don formu
ulation use
ed in abovee derivatioon assumees that thee AC
excitatio
on is appllied on one
e of the co
ontacts. Th
his leads tto the prooblem of ga
auge
invarian
nce whereby the ressponse of the diode is not sttrictly dependent on
n the
differen
nce in volta
age applied
d across the
e junction (Pedersen,, 1998). A modificatioon to
the Tien
n-Gordon approach
a that
t essenttially leadss to the sa
ame resultss was prop
posed
72
and that the applied field does not influence the charge distribution in the junction
(Pedersen, 1998). For tunnel diodes, these approximations are valid as there is a
B. Semiclassical rectifier-properties
frequency or DC current generated in the diode is given by the Iillum vs. voltage
relation of Eq. V-6. In Figure V-5, I show the procedure for calculating the
illuminated I(V) for a diode with step characteristics (Eliasson, 2001, p.142),
(Eliasson, 2005). In the next section, I will argue that the step diode is not realistic
assume that the strength of the AC signal (V) is constant. This implies that for
different bias points along the illuminated I(V) curve the power delivered by the
source varies. If V is applied using a function generator, this would indeed be the
case, but for a rectenna illuminated with a constant intensity signal, the V varies
73
I
Idark(VD)
J02Idark(V)
J12Idark(VD+Eph/e) J-12Idark(VD-Eph/e)
Eph/e Eph/e V
Iillum(VD) Eph/e
Eph/e V
In the case of a detector, the DC voltage is fixed and the additional current
generated due to the AC signal is given by Iillum(V) Idark(V). The Iillum for the ideal
case gives an open-circuit voltage of Eph/e and a short circuit current ISC that
depends on the strength of the AC signal V via the argument of the Bessel
function. Eliasson (Eliasson, 2001, p.143) shows that in the ideal diode, the DC
explained in the next section, the maximum conversion efficiency of a realistic diode
74
For a rectenna
a solar celll, the Iillum vs. voltagee relation determines the quad
drant
of the I(V) curve
e where itt provides power. U
Unlike con
nventional semicondu
uctor
junction
n solar cells, which operate in
n the fourtth quadran
nt, rectenn
nas operatte as
solar ce
ells in the second. Ussing the Iillum
i (V) cha
aracteristiccs, the operating poin
nt of
the sola
ar cell can be calcula
ated from a load-linee analysis as discusssed in the next
section.
In Figure
F V-6,, I calculatte Iillum usiing the exp
perimentallly measured MIIM d
diode
Idark sho
own in Fig
gure V-3. This
T diode has a hig
gh forward-to-reversee current rratio,
which is
i required
d for obtain
ning a sig
gnificant sh
hort-circuitt current or open-cirrcuit
voltage.
The
e effect of the illumin
nation is more noticea
able on a log scale. In
n Figure V
V-7, I
correspo
onding to the
t Idark. As
A V incre
eases, correesponding to an increeasing num
mber
75
of photo
ons incide
ent on a diode,
d the zero-crosssing of thee illuminatted I(V) sshifts
leftward
ds. This sh
hows that as the pow
wer inciden
nt on the d
diode increeases, a hiigher
operatin
ng voltage and thus a greater efficiency
y can be acchieved by
y the recteenna.
This co
onclusion is
i consiste
ent with the
t requireement for a large ccollection area
describe
ed in Chapter IV.
In Chapter
C IV
V, the strrength of the AC s ource (V) was calcculated ass the
equivale
ent of the overall sollar spectru
um. This a
approach ca
annot be a
applied herre as
the freq
quency dep
pendence of
o the sourrce also neeeds to bee accounted
d for. Elia
asson
(Eliasso
on, 2001, p.132)
p has calculated
d an electrric field sttrength (and voltagee) by
only ap
pplicable to
o a static field. For an AC fieeld, energy
y is not sttored acrooss a
capacito
or. Furtherr problems with Eliasssons meth
hod are the arbitrary
y choice of a 10
76
nm window around each discrete step in photon energy, and counting the number of
unnatural and arbitrarily set it to 1 for all frequencies. This causes a significant
to find the number of photons corresponding to the field. This is reflected in the
description of the semiclassical model given by Eq. V-2, where the electromagnetic
field is included classically (Vcos(t)) but its effect on the electrons is modeled
quantum mechanically (H). In the case of a detector, where the input power is
dissipated only as AC power in the diode, the voltage can be found out directly from
V = 2 P RDSC
Eq. V-11
The second change is required in selecting the interval over which the spectrum
should be integrated to find the power around a particular frequency. For this, I
propose that the window of integration be such that the frequencies within it
frequency remains coherent should be greater than the average time over which at
least one electron flows across the barrier. This time is proportional to the inverse of
Donges considers the coherence of the overall black body spectrum, which
77
should be greater than 80 A. If only a part of the spectrum is considered, the
coherence time will be larger. For a minimum of one electron to flow during this
the appropriate bandwidth around a central frequency, which satisfies the condition
C. Load-line analysis
The step Idark used in Figure V-5 is a misleading representation of a diode. Near
VD=0+ there is essentially zero resistance to the flow of current. As the voltage
constant. Such an increase in resistance may actually be due to the series resistance
of the diode. However, this trend for resistance is opposite of that for a real diode in
which the resistance is higher at zero voltage and decreases with increasing voltage.
A more realistic model for the ideal diode is one with no current for V<0 and a
linear I(V) for V>0. I modify this I(V) as having a higher resistance (RN) for V<0 and
lower resistance (RP) for V>0. The forward bias resistance is kept fixed in the
ensuing calculations at 100 . I also define a current asymmetry ratio A = RN/RP.
78
I
I=V/RP
I=V/RN
rectify only around VD=0 V. However, at high frequencies, the semiclassical analysis
implies that the MIM I(V) curve is sampled in steps of Eph/e even if the amplitude of
the AC signal is small. Therefore, there is non-zero responsivity even for VD0 V.
shown in Figure V-9, with a fixed V. The Iillum corresponds to the current that can
be extracted out of a diode and the operating point depends on the load resistance.
79
I
Idark(VD)
I
J02Idark(V)
J12Idark(VD+Eph/e) J-12Idark(VD-Eph/e)
Eph/e Eph/e V
Iillum(VD) Eph/e
Eph/e V
Figure V-9 Iillum vs. voltage for a piecewise linear I(V) with a
high forward-to-reverse current ratio.
order to keep the input power constant. Under this condition, I now derive the
modified load-line curve and find out the operating point corresponding to the
maximum obtainable power from a solar cell based on this diode. The rectified
power given by Prect=VD*Iillum can be expressed in terms of Idark using Eq. V-6 as
2 2 2
Prect = VD 1
I dark (VD ) +
4 4
( )
I dark (VD + E ph e ) + I dark (VD E ph e )
Eq. V-12
80
where Iillum can be separated into Idark and the remaining terms that depend on the
strength of the signal making up IDC. Therefore IDC gets divided into the load
current and the dark current in the diode, as shown in Figure IV-6. The total power
received by the rectenna is equal to the sum of AC power dissipated in the diode
and the rectified DC power, which is equal to VD*IDC. When the DC resistance of
the diode at the bias point is much larger than the load resistance, the DC power is
most efficiently delivered to the load. For power generation, the diode has to operate
in the second quadrant implying that Eph/e<VD<0. The total power equal to the
1 1 1
Pin = PAC + PDC = V I + V D I DC = V I + V D iSC V I
2 2 2
V2
= SC
2 RD
(
1 + V D iSC )
Eq. V-13
Under the assumption that little or no power is lost to the diode dark current at
reverse bias, the efficiency of the rectenna solar cell at a bias point is given by
VD iSC
load =
(1 + V )
D i
SC
Eq. V-14
According to the Iillum in Figure V-9, the largest absolute operating voltage of
the rectenna for power generation in the second quadrant is /e, while the
maximum responsivity of the diode from Eq. V-10 is e/. Therefore the maximum
value of load is 1/2 implying that under the first-order semiclassical regime, the
maximum efficiency of operation for the rectenna is 50%. The remaining 50% or
81
I now apply the above analysis to a piecewise linear and an exponential dark
I(V) curve. The forward bias resistance of the piecewise linear I(V) is 100 and the
forward to reverse current ratio is 108. The reverse saturation current of the
exponential characteristics is 10-10 A and the coefficient in the exponential is 100 V-
1. An input power of 10-8 W is assumed, at a photon energy of 1.5 eV. The power
corresponds to the solar intensity of 100 W/cm2 received over an area of 10 m2.
The basic assumption in the above analysis is that the first order theory is
the illuminated I(V), the variation of , and the load efficiency as a function of the
DC bias voltage.
82
-9
x 10
6
5
Iillum (A)
4
3
2
Piecewise linear
1 Exponential
0
-1.5 -1 -0.5 0
0
10
: V/Vph
-5
10
-10
10
-1.5 -1 -0.5 0
1
0.8
Efficiency: Load
0.6
0.4
0.2
0
-1.5 -1 -0.5 0
VD (V)
equal to the input power of 10-8 W. The assumption of <<1 is satisfied for majority
of the illuminated characteristics and is violated only near VD=VOC. The effect of a
varying V is evident in the illuminated I(V) curve leading to the trapezoidal shape
that results in a peak efficiency close to 50%. This is to be contrasted with Figure
83
V-9, where the triangular illuminated I(V), under the assumption of constant V,
As V increases at large negative VD, the efficiency rises. This suggests that
enhancement of the field by confining a fixed power in a smaller diode region leads
resistance, responsivity, illuminated I(V) and power need to be derived with higher-
order Bessel terms included in Eq. V-6 and Eq. V-7. Another consideration at large
order analysis. These factors may modify the illuminated characteristics and
In the limit of photon energy smaller than the voltage scale of nonlinearity in
the diode, the semiclassical formulas for resistance and responsivity reduce to their
between the classical and semiclassical theories for rectification. Here, I analyze
this correspondence and also explain why a small-signal classical rectifier and even
the first-order semiclassical one are only able to operate at 50% efficiency.
84
Both the classical small-signal and the semiclassical first-order models, under
dissipated as AC power in the diode. At any < 1/2, the AC power dissipation is
correspondingly increased.
In the classical rectifier, the AC signal from the antenna feeds into an AC diode
resistance RD, causing dissipation in the diode. Under the first-order semiclassical
influenced by the large photon energy, suggesting that the dissipation occurs due to
photon-assisted tunneling.
I have emphasized the fact that the diode is always dissipating in order to point
out the difference between the classical small-signal and large-signal (linear)
rectifier. In the large-signal circuit model, the diode turns ON only for a short
85
I
VD
V (t)
V
time
In this case, the best case self-bias DC voltage that can develop across the diode
is VD=-V. In doing so, the rectenna essentially becomes a level shifter (Sedra, 2004,
p.194). If the level shift is such that the diode never turns on, there is no DC current
generated. For supplying power to a load, the diode must turn on for a part of the
AC voltage cycle. This happens for V(t)+VD > 0 allowing the rectenna to draw
charge at a very low resistance and therefore low loss. In the remaining cycle, the
diode is turn off, and the charge that was pulled up is supplied to the load. The AC-
to-DC conversion efficiency now depends on the power that is dissipated in the
86
diode when it turns ON. Depending on the diode asymmetry, conversion efficiency
described above and the higher-order semiclassical model considered in the previous
consistent with the circuit model in the limit of small photon energies. This is
crucial for understanding whether efficiencies higher than 50% are achievable in
87
CHAPTER VI
from literature, I calculate its normalized detectivity for comparison with several IR
detectors. I also identify ways for improving the detectivity of the TW device.
A. Theory of operation
signal in the diode. The difference between rectennas using a lumped-element diode
and the TW diode is twofold. First, they differ in the manner in which the signal is
transferred from the antenna to the diode. Second, as the names suggest, the
traveling-wave rectifies a wave while the lumped-element has the same signal
matching and RC time constant for a lumped-element model. Here I describe the
concept of a TW device and explain how it can achieve the above mentioned
requirements.
The MIM traveling-wave detector was proposed as a method for improving the
efficiency of antenna coupled MIM rectifiers operating in the infrared (Estes, 2006).
A 3D view of this device is shown in Figure VI-1(a). It consists of an antenna
connected to two metals M1 and M2, with a thin insulator between them. The
direction. The plasmon develops a voltage between the two metal electrodes causing
asymmetry in the I(V) curve about the bias point, as characterized by the
frequency components of the rectified signal do not travel over a large distance.
89
Figurre VI-1 (a) An isome etric view oof the anttenna-coupleed
travelling-wave detector.
d Thhe antenna a arms con nverge into a
paralllel-plate wa
aveguide wiith a thin (2 2 nm) insullator betweeen
the metals
m M1 and
a M2. On n the other end, thesee metals forrm
the leads
l to thhe contact pads (nott shown) oor the loa ad.
Parammeters requ uired for caalculating tthe perform mance of th he
travelling-wave include
i the
e characterristic imped dance of th he
waveg guide (ZC); the
t plasmon n decay lenngth (-1); an
nd the tunnnel
diode resistance (RD) and responsivitty (i). (b) Small sign nal
circuiit represenntation of the detecttor. Imped dance of th he
travelling-wave diode
d can be readily ma atched to thee antenna (ZZA
= ZC*).
* (c) A 3D D view of th he travelin g-wave MIM M diode. Thhe
structture has a length L and a wid dth w. To ccalculate th he
responsivity, I divide
d the entire
e lengtth of the trraveling-wavve
into N equal secttions of widtth z.
The
e technique
e of rectifyiing a surfa
ace-plasmon
n wave exttends the b
bandwidth
h and
waveguide as sho
own in the equivalen
nt circuit iin Figure V
VI-1(b). Th
his leads tto an
improve
ed power transfer frrom the antenna
a too the diod
de. Also, tthe distrib
buted
next secction, I ex
xplain the techniques required
d to modell the perfoormance off the
travelin
ng-wave dettector.
90
A variation of the traveling-wave has been implemented (Hobbs, 2007) by
model the responsivity of the distributed rectifier, the MIM waveguide is divided
into N parts of width z as shown in Figure VI-1(c). The antenna excites a surface-
plasmon wave in the MIM waveguide. Assuming that a power of 1 W is incident on
the waveguide cross-section (x-y), this wave develops a certain voltage distribution
(V1W(z)) between the two metal plates. The 1 W of power is chosen for normalization.
smaller. Resistive losses in the metal cause the plasmon-wave amplitude to decay
V1W ( z ) = V0 e z
Eq. VI-1
where is the decay constant or the inverse of the decay length of the plasmon, and
V0 is the voltage at z = 0. Here I have ignored the decay of the plasmon wave due to
rectification. As calculated later, the rectified energy is a small fraction of the total.
To simplify the derivation, I assume a constant voltage distribution along the x-axis.
The net AC power across the diode is given by the following sum
N
V1W 2 ( z )
PAC =
1 2 R D / A ( w z )
Eq. VI-2
where RD/A is the resistance per unit area of the diode, and w is the width along the
91
idc = i PAC
Eq. VI-3
Combining Eq. VI-2 and Eq. VI-3, and applying the limit of N (z0) gives
iV02 w 1 e 2 L
idc =
2 RD / A 2
Eq. VI-4
where L is the length of the waveguide. Since the incident power was normalized to
1 W, the DC current in Eq. VI-4 also gives the responsivity. Choosing the length of
the waveguide to be larger than the decay length of the surface-plasmon L > 1 , the
responsivity is approximated as
iV02 w
TW =
4 RD / A
Eq. VI-5
Eq. VI-5 provides an estimate for the efficiency of the traveling-wave detector
and helps to identify ways for improving the device, which are discussed later.
I now describe how to obtain the parameters required for computing the
responsivity using Eq. VI-5. The semiclassical diode-properties (i and RD/A) have
already been discussed in chapter V. Next, I describe the procedure for calculating
the V0, , and the characteristic impedance (ZC) of the TW device. These quantities
are then used to estimate the responsivity and noise performance of the TW
detector.
To find the parameters for the traveling-wave section, I first calculate the
is assumed that the waveguide is infinite in the x-direction, ensuring that only TM
modes exist. These modes can have an even- or an odd-symmetric field distribution
92
variation on the antenna arms, which matches with the odd TM mode of the MIM
the antenna impedance, the impedance matching and the mode matching allow
by using a matrix method (Davis, 2009) to obtain the dispersion relation for the TM
mode. With the estimate for and fixing the width of the waveguide to 100 nm I
solve for the corresponding hybrid mode (TE + TM) using a commercial finite
from which the plasmon wavelength and the distance over which it propagates can
be calculated.
The solver also provides the field distribution in the 2D cross-section of the
waveguide. In Figure VI-2, I plot these results for a Ni-NiO-Ni (MIM) TW structure
simulated at 100 THz. As seen in Figure VI-2(a), the wave is confined mainly to the
93
Figurre VI-2 Finiite element analysis off a Ni-NiO--Ni travelin ng-
wave diode for 100 THz. (a) Cross-ssectional diistribution of
powerr flowing in the z-directtion. The bllack rectanggles mark thhe
metall regions of the MIM diode.
d The ppower is connfined main nly
to thee 2-nm thicck insulatorr. (b) The yy- and z-directed electrric
fields as a function of the vertical
v possition for x=
=0. (c) The x-
directted magnetiic field as a function oof the horizoontal positioon
for y==0. The EY(y) and the HX(x) aree used to ccalculate th he
charaacteristic im
mpedance off the traveliing-wave, w which is 45
for th
his analysis.
From
m the field
d distribution, the ch
haracteristtic impedan
nce (Huang, 2009) off the
TW stru
ucture is ob
btained as
M2 M2
V E.dy E dy Y
ZC = = M1
= M1
.dr
I H
, y = 0
H X dx
Eq. VI-6
Figure VI-2(b). Th
he currentt is calcula
ated using
g Ampress law by iintegrating
g the
94
Figure VI-2(c), the magnetic field falls rapidly outside the waveguide. Thus at an
infinite distance, it is negligible and the current can be approximated by an integral
along x from - to +. For the validity of the above formula, the metal needs to be a
perfect electric conductor or have its magnitude of dielectric constant much larger
than the insulator |metal|>>insulator. At infrared wavelengths, the latter is the case
Next, I calculate the V0, which is proportional to the voltage given by the
normalization factor given by the integration of the z-directed power density in the
E dy y
V0 = M1
area
PZ dA
Eq. VI-7
C. Performance calculation
match between the antenna and the diode, and overcome the RC bandwidth
MIM material combinations that provide a small resistance (RD/A) for the tunnel
responsivity for the TW detector. The MIM diodes compared are Ni/NiO/Ni (0.2 eV)
95
(Hobbs, 2007), Nb b (0.1 eV) (Bain, 198
b/Nb2O5/Nb 85), and Ta/Ta2O5/Ta
a (0.4 eV). The
bracketed quantitty denotes the barriier height of each d
diode. Sym
mmetric tun
nnel-
barrierss have been
n chosen to
o keep the analysis
a simple.
1. Cha
aracteristiic impeda
ance and rresponsiv
vity
The
e characterristic impedance (ZC) of the tr aveling-wa
ave diode iis plotted as a
function
n of wavelength (1.5 to 10 m
m) in Figurre VI-3. Th
he ZC is w
well within
n the
range of typical antenna
a im
mpedances. For a prrecise matcch with the antenna, the
impedan
nce of the TW structure can be
e fine tuned
d by varyin
ng its widtth along-x. The
broadba
and detectiion. The sm
mall variatiion in impeedance witth frequenccy is attrib
buted
to varia
ation in th
he experim
mental valu
ues for thee metal diielectric coonstants (metal)
compara
able as the
eir metal are
e similar.
96
Assu
uming no losses in the anten
nna, the reesponsivity
y of the T
TW detectoor is
calculatted at a DC
C bias of 0.1
0 V. The bias
b ensurres an operrating poin
nt at which
h the
diode re
esponsivity
y (i) is non
n-zero. I asssume thatt a perfect impedancee match ca
an be
made be
etween the
e traveling-wave diod
de and the antenna. T
The detector responssivity
calculatted from Eq
q. VI-5 is shown
s in Figure
F VI-4
4. The resp
ponsivity iss higher forr the
for Nb and
a Ni, and
d its absen
nce in Ta, can
c be expllained from
m the trend
ds in resisttance
(RD/A) and
a respon
nsivity (i). In the sem
miclassicall model, th
hese quantities vary with
frequen
ncy. Since the
t Ta-Ta2O5-Ta is a higher t unnel-barrrier than tthe Nb and Ni
MIM diiodes, it ha
as a differrent variattion of RD//A and i d
due to its larger turrn-on
voltage.
The
e maxima in the Nb
b curve at a wavellength of 3 m givees a quan
ntum
efficienccy of 3.6%.
97
2. Com
mparison with the lumped-el
l lement de
etector
I compare the
e responsiv
vity of the TW detecttor with a lumped-eleement deteector
that hass an antenna impeda
ance of 100 and a d
diode area oof 100x100
0 nm2. Both
h the
shown in
i Figure VI-5.
V For th
he lumped-element d
device, the calculation
n is based on
= ci where
w c acccounts for the
t RC tim
me constantt and the im
mpedance mismatch.. The
perform
mance of th
he lumped--element detector
d degrades witth increasee in frequeency.
As expllained in Chapter
C IV
V, only a minor imp
provementt in the lu
umped-elem
ment
device can
c be achieved by ch
hoosing a sm
maller area
a for the diiode.
Thu
us, the resp
ponsivity of
o the TW detector iss significan
ntly improoved and m
much
and the
e responsiv
vity allow the
t detectio
on bandwiidth to exteend over several miccrons
98
in the MWIR and the LWIR ranges. The spectral response of the TW detector can be
In order to compare the TW detector with other detector technologies used for
defined as
1 TW
D* = ( Ad f ) 2
In
Eq. VI-8
where Ad is the detector area over which the radiation is received, f is the
bandwidth of the readout circuitry, and In is the noise current in the diode.
Figure VI-6, I compare the calculated D* of the traveling-wave detector with some
explained later in this section. If implemented, the improved noise performance will
place the traveling-wave detector at par with thermal detectors (Rogalski, 2003).
99
R
References
[1] (Rogalski, 2000)
[2] (Gunapala a, 2008)
[3] (Rogalski, 2000)
[4] (Jayaweerra, 2008)
[5] (Phiar, 2002)
[6] (Tiwari, 20
009)
Figurre VI-6 Calcculated dete
ectivity com parison of tthe travelin
ng-
wave (TW) dete ector with semiconducctor, therm mal and MIIM
lumpe ed-element detectors. The initiall calculation n of the TW
devicee response is based on o the Nb--Nb2O5-Nb device. With
improovements, th he performance of the TW detecttor can be a at-
par with
w thermall detectors.
With
h its proje
ected perfo
ormance ca
apability oon par with
h thermal detectors,, the
travelin
ng-wave de
etector can be used in
n a range of applications inclu
uding activ
ve IR
imaging
g. A signifiicant advan
ntage of th
he TW deteector over the semicoonductor b
based
and dep
position of amorphous
a s thin-filmss, the detecctor can bee easily ma
ade into a ffocal-
plane-arrray on top
p of existin
ng CMOS circuits. T
The antenn
na preservees the phase of
the inco
oming signa
al (Middleb
brook, 2008
8), thereforre the TW detector ca
an also be used
in IR phased-arr
p rays. Unlik
ke therma
al detectorrs (Rogalsk
ki, 2000, p
p.68), the TW
bandwid
dths in the
e terahertzz. The larg
ge bandwid
dth is ben
neficial in ccommunica
ation
100
1. Scope for improvement
Three criteria for improving the performance of the traveling-wave detector can
be identified from Eq. VI-8. Since the detector is not limited by background
radiation noise, i.e., it is not a background limited infrared photodetector (BLIP)
(Rogalski, 2000, p.16), the receiving area can be increased without affecting the
noise performance. Also, the responsivity of the TW diode, given in Eq. VI-5, can be
improved. This can be done by improving the design of the traveling-wave structure
and/or by using a diode with low resistance and high responsivity. These diode
as discussed in Chapter III. Finally, reducing the current noise In can also improve
the D*. This can be done by eliminating the shot noise, which is proportional to the
DC current of the diode by operating at zero bias. For a rectenna to work at zero
101
CHAPTER VII
The concept of a geometric diode was introduced in Chapter I and the efficiency
small critical dimensions on the scale of the mean-free path length for charge
carriers are required for geometric rectification. Successful fabrication at such
conductors.
electron-beam (e-beam) lithography. Here I give the fabrication details for the e-
beam process carried out at the University of California, Santa Barbara. Antenna-
coupled geometric diodes made from both metal and graphene thin-films were
fabricated.
102
A. Metal ge
eometric diodes
d
previous attempt at
a making aluminum hu, 2010), it was reallized
m geometricc diodes (Zh
that the
e extremely m is difficult to measu
y low resistance of a metal film ure. Thereffore I
chose to
o make the gold geom
metric diode
es with a m
metal-thick
kness of 25 nm along with
a 5 nm chromium
m or titaniu
um adhesiion layer. The diode was desig
gned to ha
ave a
neck sizze of 20 nm
m as shown
n in Figure
e VII-1, and
d was conn
nected at the feedpoint of
a bow-ttie antenn
na. The antenna
a was
w ned to recceive 10.6 m radia
design ation
(Gonzllez, 2005) in
n an edge-ffed configu
uration (Weeiss, 2003).
An oxidized
o silicon waferr with 750 nm oxide w
was used a
as the subsstrate on w
which
Cr/Au (5/45
( nm) metal-elect
m trodes werre placed u
using phottolithography and lifft-off.
Alignme
ent marks for e-beam
m lithograp
phy were a
also pattern
ned during
g this step. The
103
next ste
ep was to pattern th
he shape off the geom
metric diod
de and the antenna. This
which allows
a high
h resolution
n lithograph
hy. The lit hography w
was done b
by Bill Mitcchell
determiine the ap
ppropriate exposure-dose for Z
ZEP on th
he substra
ate used. A
After
patterniing the an
ntenna reg
gion at a dose of 4
400 C/cm
m2, the opttimal dosee for
the diod
de area getts partially
y exposed due
d to the p
proximity. Therefore,, for patterrning
the diod
de, a smalller dose iss required to compen
nsate for th
his uninten
nded expossure.
The resu
ulting pattterns are sh
hown in Fiigure VII-2
2 and Figurre VII-3.
104
Figurre VII-3 SE
EM image of
o the anteenna and th
he four-point
probe
e contacts for
f to the geometric diode show
wn in Figuure
VII-2.
Afte
er depositin
ng the Cr/A
Au (5/25 nm)
n using ee-beam eva
aporation, the ZEP rresist
The
e DC I(V) cu
urves for the metal geometric
g d
diodes weree observed to be linea
ar. A
possible
e cause forr the linea
ar characte
eristics is the seriess resistancce between
n the
device and
a the prrobes, as seen
s in Fig
gure VII-3
3. The deviices were also tested
d for
response to infra
ared illumination ussing the ssetup desccribed in section C.. No
ation was observed.
rectifica o
The
e failure off metal ba
ased geom
metric diodees can be attributed
d to the g
grain
boundarries in the
e thin-film
m that limiit the mea
an-free-patth. As seen in the S
SEM
image in
n Figure VII-4,
V the grain
g boundaries occu
ur on a sca
ale smallerr than the bulk
mean-frree-path an
nd compara
able to the
e thicknesss of the film
m and the neck-size. The
(Durkan
n, 2007, p.1
107) ratherr than at th
he edges off the condu
uctor.
105
Figurre VII-4 Gra
ains in the evaporated
e C
Cr/Au thin-ffilm.
B. Graphen
ne geomettric diodes
is a two
o-dimension
nal crystallline sheet of carbon atoms with a large m
mean-free-path
chose to
o make botth the diod
de and the
e antenna u
using grap
phene. Later efforts h
have
modified
d the process to hav
ve a metal antenna cconnected to a graph
hene geom
metric
diode.
The
e first ste
ep in the fabricatio
on of grap
phene geoometric diiodes was the
exfoliatiion of grap
phene flak
kes onto an
n oxidized silicon wa
afer with a 90 nm tthick
oxide. The
T 90 (or 300 nm) SiO
S 2 enhan
nces the coontrast of the graphene, makin
ng it
easier to locate an
nd to estim
mate its thiickness usiing an optiical microsscope (Rodd
daro,
2007). Graphene
G films with
h 1-10 layers were u
used. As tthe grapheene flakes had
106
random locations and orientations, they were separated by cleaving the wafer and
processed individually. While cleaving the wafer, precautions were taken to prevent
silicon dust from falling on the graphene. This was achieved by shielding the
graphene using a glass slide that also serves as a line to scribe along.
Cr/Au (5/45 nm) contacts for four-point probe measurement were placed using
lift-off. The contacts also helped to keep the graphene clamped to the substrate
during the subsequent processing steps. Negative resist NR9-1000PY was used for
patterning the contacts. The resist was spun at 3000 rpm for 40 s and prebaked for
1 min at 150 C. Alignment between the contact patterns on the mask and the
graphene on the wafer was guided with scratch marks that were made while
locating the graphene. The grayscale images displayed on the screen of the MJB4
mask-aligner made it easier to locate and align the graphene. Once aligned, the
NR9 resist required a 3 s exposure on the MJB4. The resist was post-baked at 100
C for 1 min followed by a 7 s dip (agitated) in the resist developer RD6. The usual
plasma. As graphene is sensitive to oxygen plasma this step was not performed. To
a noticeable change on the shape of the patterns. This was followed by a 20 s dip in
DI water and N2 blow-dry. The metals for the contacts were evaporated using a
thermal evaporator. The lift-off was done in acetone (without ultrasonic) followed by
beam resist (hardens where exposed) so that the unprotected areas of graphene
which has a resolution of about 50 nm, was used with HMDS as an adhesion layer.
A 300 nm thick resist layer was obtained by coating the wafer at 3000 rpm for 30 s.
The exposure dose for e-beam lithography done at UCSB was between 1000-1400
107
C/cm2. The resist was developed for 45 secs in AZ 300MIF developer followed by a
60 s rinse in DI water. The exposed graphene was etched using oxygen plasma (100
the resist was stripped by soaking the wafer in Shipley 1165 at 80 C for 15 min.
devices. After imaging the patterns generated on the resist with an SEM, stripping
the resist became difficult. It was concluded that the SEM beam was depositing a
background CO, CO2, or hydrocarbons) that prevented the stripper from accessing
the resist (Mitchell, 2010). Furthermore, ultrasonic could not be used to speed-up
the lift-off or the resist removal as the contacts and graphene were prone to
developing cracks.
Figure VII-5 (a). The diode is connected to a graphene antenna as shown in Figure
VII-5 (b).
108
(a) (b)
The
e antenna-ccoupled graphene de
evices weree not tested for DC II(V) due too the
series resistance of
o the lead
ds and also
o because tthe graphene devicess were pron
ne to
ut under the
burn-ou t param
meter analyzer (HP4
4145B). Graphene d
diodes witthout
antenna
as fabricate
ed using a similar prrocess havee subsequeently been tested usiing a
were pe
erformed with
w a fourr-point pro
obe setup tto avoid noonlinearity
y at the m
metal-
graphen
ne contact (Perello,
( 20
010), (Ran,, 2009).
109
Figurre VII-6 Fou
ur point pro
obe configurration for m
measuring th
he
diode I(V) withou
ut including
g nonlinearitty at contaccts.
In th
he above fiigure, the Si
S substratte is conneccted to a ga
ate bias that modifiess the
Fermi level
l relatiive to the bandstruccture of grraphene. A
At zero gatte voltage,, the
graphen
ne is p-type
e due to itss interactio
on with th
he SiO2 sub
bstrate (Ka
ang, 2008). The
majority
y carriers in graphe
ene can be made n -type by a
applying a positive gate
voltage greater than the volttage corressponding too the Diracc point (Noovoselov, 20
005),
The
e DC chara
acteristics of
o a graph
hene geomeetric diode at gate vooltages of 2
20 V
carrierss for VG = 40 V. At VG = 20 V,
V the currrent is higher at poositive tha
an at
negative
e VDS indiicating tha
at the prefferred direection of current flow
w is along
g the
funnel, which is also the preferred direction for p-typee carrier ttransport. The
resistan
nce is high
her for the
e opposite voltage d
direction th
hat blocks the flow. The
asymme
etry is also
o seen in the
t zero-biias responssivity. Thee opposite characteriistics
are observed for VG = 40 V as the majority ccarriers arre n-type. The prefeerred
direction
n for charg
ge flow is still along the
t funnel h
however n-type carriiers requiree the
opposite
e bias polarrity to flow
w in that direction.
110
Figurre VII-7 (a) Current, (b)) resistancee, and (c) ressponsivity v
vs.
voltag
ge for a graaphene geom metric diod e made usiing exfoliateed
graphhene.
direction
n of bias th
hat blocks the flow of holes. Th
he asymmeetry in the I(V) curve also
leads to
o a non-ze
ero respon
nsivity at zero-bias,
z thereby a
allowing geeometric-diiode-
based re
ectennas to
o operate without
w an external b
bias.
111
C. Infrared characterization of antenna-coupled diodes
from a CO2 laser (=10.6 m). Performing this measurement at far-infrared instead
of visible or near-infrared is preferred as the resistive losses, both in metals (Fox,
A Synrad CO2 laser (Synrad, 2010), model number 48-1 SWJ, is used in the
frequency of 5 kHz is applied to keep the gas ionized without lasing. The laser needs
to be warmed-up in this state for more than 15 min to obtain a stable output. The
high-time of the pulse is increased to enable lasing. The output of the laser is pulsed
according to the applied signal. Therefore, the output power of the laser is
nonlinearity of the diode, VOC and ISC are expected to be extremely low. This
necessitates the use of a lock-in detection technique with the setup shown in Figure
VII-8.
112
Figurre VII-8 Setup for meassuring resp ponse of anttenna-coupleed
geome etric diodess to 10.6 m
m illuminatiion. The wa
afer is rotateed
in th
he horizontal plane to o change tthe angle between th he
polariizations of the
t radiation n and the anntenna.
Infrrared radia
ation from the CO2 laser is spllit into twoo beams, oone of whicch is
measure
ement. A heat sink is used to block th
he IR laserr prior to and afterr the
measure
ement. The wafer is rotated to change th
he angle (
) between polarizatioon of
the rad
diation and
d the ante
enna. The radiation emitted b
by the laseer is vertically
polarize
ed.
The
e infrared beam is chopped at a frequen
ncy (f). Th
he rectified
d signal in
n the
device is
i also mo
odulated att the same
e frequenccy. From tthe choppeer controlleer, a
frequen
ncy referencce is fed in
nto the refe
erence portt of the lock
k-in ampliffier which then
measure
es the amp
plitude of the
t signal from
f the deevice at th
hat frequen
ncy. The locck-in
input ca
an be set to
o detect either curren
nt or voltag
ge.
113
The
e variation
n of ISC wiith for a graphen
ne geometrric diode cconnected to a
graphen
ne antenna
a is shown in Figure VII-9.
V The maximum
m response ooccurs at = 0
when th
he antenna
a is aligne
ed parallel to the eleectric-field. An orderr of magniitude
as the ratio of th
he current generated
d to the ap
pproximate power in
ncident on
n the
antenna
a. The lase
er beam ha
as a diame
eter of 4 m
mm at the source and
d expands at a
laser be
eam as unifform over an
a 8 mm diameter.
d F
For a total power of 1 W, choppeed at
area of 10 m2 the
e power recceived is 50
0 nW. At =
=0, this leeads to a reesponsivity
y of 3
pA/50 nW
n ~ 6x10
0-5 A/W. Th
he low efficciency can
n be attribu
uted to a n
number off loss
mechan
nisms inclu
uding anten
nna efficien
ncy, impeda
ance mism
match, and poor quanttum-
114
As the complete device is made out of graphene, the sheet resistance is constant
throughout the antenna and the diode. The rectified current generated in the
geometric diode flows through the antenna and the leads before being detected,
which weakens the current signal measured by the lock-in detector. This is avoided
by measuring the open-circuit voltage. In Figure VII-10, I plot the VOC vs. for
varying power of the laser.
0.5
0.6 W
0.8 W
0.4
Voltage: V OC ( V)
0.9 W
1W
0.3
0.2
0.1
-75 -30 0 30 75
Angle: (degree)
For =0 and 1 W power on the laser, the diode + series resistance is given by
the ratio VOC/ISC (=RD+S). The measurements shown in Figure VII-9 and Figure
VII-10 are on different devices, therefore the calculation of resistance gives at best
an order of magnitude estimate. The RD+S = 0.5 V/3 pA = 167 k also includes the
geometric diode indicates that the signal occurs due to the presence of the antenna
and thereby a device a connected to the antenna. Ideally, the polarization sensitive
115
response should be smallest at 90. As seen in Figure VII-9, this may be true for
the negative angles but is not the case for positive angles. Also, in Figure VII-10,
this is not the case for all power levels. As the response is extremely small, it is
conceivable that this is due to measurement error. It could also be due to variation
in the position of the device with respect to the IR beam in going from one rotation
to the other. A possible way to eliminate the latter is to mount the device on a TO-8
can, making sure that the position of the rectenna coincides with the rotation axis of
the can.
The response observed is not bolometric (Richards, 1994) as both VOC and ISC
were measured in the absence of any external bias. Thermo-electric effect arising at
the metal-graphene interface due to unequal heating of different regions can lead to
116
CHAPTER VIII
using the Drude model and also using a wavefunction approach. Characteristics of
geometric diodes simulated using the Drude model (Zhu, 2011) are consistent with
the expected direction of asymmetry in the I(V) curve. However, these simulations
are based on a classical model and account for the shape of the conductor through
the specular reflection of charges from the boundaries. Moreover, a constant field
distribution across the patterned thin-film is assumed, without accounting for the
effect of charge or shape on the potential. Material properties only enter through
the mean-free-path length and Fermi velocity. Finally, it is not possible to model
graphene waveguide (Li, 2009). However, this approach assumed the wavefunction
to be coherent across the device. Modeling the geometric diode accurately requires
the interaction of the charge-carriers with quanta of energy (=) of the AC signal.
the Hamiltonian that describes the material and the structure of the geometric
117
diode, the NEGF approach can be used to model geometry-dependent transport,
the NEGF method. Development of the simulator is an ongoing project and in its
current version, it can model ballistic transport through graphene while calculating
the charge and potential self-consistently. The simulator can be further improved to
include scattering and, as explained in Chapter IX, it can be used in conjunction
transport.
Though all the results in this chapter are specific to the 2D hexagonal lattice of
obtained by converting the physical problem defined by the NEGF formalism into a
set of matrix equations that are solved to obtain the I(V) curve for the geometric
diode.
A. NEGF formalism
manner similar to the use of Fourier series for solving partial differential equations
the system under consideration is open, i.e. it interacts with the surroundings. For
the geometric diode, such an interaction occurs due to the flow of current to and
from the contacts and can be simulated using the NEGF approach (Datta, 2002).
Consider a central device region with contacts at two ends that act as charge
118
properties, shape, charge, and field in the device region are modeled by the
Hamiltonian (H) and the potential (U). In the absence of contacts, the solution to (E
Contacts
Device
1 2
[H+U]
1 2
The effect of the contacts on the central device region is three-fold (Datta, 2005,
p.183). First, the contacts determine the potential boundary-conditions at the two
ends of the device. Second, they act as charge reservoirs from which carriers are
connection between the device and the reservoirs. Third, on connecting a set of
states from the contact to the device. The last two effects are a consequence of the
time*energy uncertainty relation between the lifetime of an interacting state and its
Even though the reservoirs are significantly larger in size compared to the
the escape rate, that has the same size as H (Datta, 2000). A detailed explanation of
119
Scattering mechanisms can be modeled using NEGF through the appropriate
and gets affected by the charge distribution in the device region, S has to be
determined self-consistently with the charge. In the results presented in the later
sections, I have not accounted for scattering. However, it can be added to the basic
The formulation of the simulator along with the explanation of the relevant
quantities is given below. If the reader finds it difficult to follow the matrix-version
For an isolated system, the eigenenergies and the corresponding set of wave
EI ( H + U ) { } = 0
Eq. VIII-1
i1 i2
EI ( H + U ) + + { } = 0
2 2
Eq. VIII-2
By adding , the wavefunction in the device gets modified such that it decays
i
=
2
Eq. VIII-3
120
More generally, can have a real part as well, which represents the effect of the
contact on the device Hamiltonian. The outflow or the decay of electrons from the
EI ( H + U ) 1 2 { } = {S }
Eq. VIII-4
in = {S }{S }+ = [ ] f
Eq. VIII-5
occupation of energy-levels in the reservoir, and the coupling between the device
{ } = [G r ]{S}
Eq. VIII-6
zero for t<0 (retarded). As seen in Eq. VIII-5, the physical quantities in the system
(e.g. the strength of the source) are given by the square of the wavefunction-like
121
itself has no physical significance. However, two essential quantities that can be
derived from it are the electron density and the density of states.
+ +
G n = { }{ } = G r {S }{S } G r + = G r inG r +
Eq. VIII-8
in = 1in + in2 = 1 f1 + 2 f 2
Eq. VIII-9
The density of states is given by the diagonal elements of the spectral function
(A) defined as
A = G r 2G r + + G r 1G r +
or
A = i [G r G r + ]
Eq. VIII-10
The expression for A is derived from the sum of Gn and a function similar to it
with f replaced by (1-f) in Eq. VIII-9, such that both filled and empty states are
counted.
Finally, the current in the device from both the contacts is calculated for a
particular energy as
I ( E ) = I1 + I 2 =
e
h
(
Trace 1in A 1G n + in2 A 2G n )
Eq. VIII-11
The inA term represents the current flowing from the reservoir to the device
and depends on the strength of the source and the density of states in the device.
122
The Gn term represents the current flowing out of the device, which is a function of
the escape-rate and the electron density in the device. The total current is the
(Datta, 2000)
e
I (E) =
h
(
T ( E )( f1 f 2 ) ; T ( E ) = Trace 2G 1G + )
Eq. VIII-12
on the angle of the edges with respect to the lattice, two edge profiles can be
obtained. Based on the type of edge, the graphene is referred to as zigzag and
Hamiltonian (H) of the device region is given by a matrix whose diagonal elements
give the Bloch energy (2p) at each atomic location and the off-diagonal elements (1)
give the interaction energy with the first nearest-neighboring atoms. Interactions
with second- and third-nearest neighboring atoms can also be taken into account.
eV, 1 = -2.7 eV. In Section VIII-F, I will introduce the effect of a gate voltage on the
Fermi level in graphene through a term Eoffset, which is accounted for as 2p = Eoffset.
The H-matrix is of the size NxN where N is the number of atoms in the device.
matrix.
The
e on-site en on atom i s containeed within the -matrices
nergy for each carbo
along with
w the nea
arest-neigh
hbor intera
action betw
ween the atooms enclossed in the ssame
rectanglle. The intteraction between two -matricces is proviided in thee -matrix. The
TBM Hamiltonian
n consists of
o the -ma
atrices alon
ng the diagonal and the -matrices
The
e simulatorr is progra
ammed to accept an
n input forr the neck
k and shou
ulder
obtain the
t asymm
metric hourg
glass geom
metry shown
n in Figuree VIII-3. Th
he width of the
uncut-re
egions on the
t left and
d the right can be varried.
124
60
For the graphene geometric diode in Figure VIII-3, the assignment of and
1 1 2 2 3 3 4 4 5 5 6 6 7
125
C. Contact self-energy ()
feasible (Blanter, 2007). However, such an interface may not have linear I(V)
characteristics (Ran, 2009). In order to ensure that the contacts do not contribute to
the nonlinearity or asymmetry in the simulated I(V) curve, I choose to model the
contacts as semi-infinite graphene leads stretched out to the left and the right
Greens function (gR) (Datta, 2005, p.190). The gR is a subset of the overall Greens
function in the reservoir and accounts for the atoms in the reservoir closest to the
device region. This simplification is based on the periodic lattice of the contact and
accounts for the atoms in the first period. Due to the limited computational
capability, I restrict the width of the contact to be equal to the shoulder-size of the
diode. If 1 and 7 in Figure VIII-4 extend indefinitely to the left and the right
as shown in Figure VIII-5. Condensing the vertical direction into matrix form allows
To obtain gR, the following equation needs to be solved for g iteratively (Datta
(a), 2009)
1
g n +1 = ( E + i ) I g n +
Eq. VIII-13
126
The self-energy matrix is calculated from gR as
= gR +
Eq. VIII-14
For a description of the iterative procedures for solving Eq. VIII-13, see the
work of Sancho and Rubio (Sancho, 1984) and references therein. To avoid an
proposed (Umerski, 1997). However, both the iterative and the closed-form methods
are unstable due to the ill-conditioned -matrices for graphene. The difficulty arises
A two-step semi-analytical approach for finding the self-energy, even with ill-
conditioned matrices, is given by Rungger and Sanvito (Rungger, 2008). In the first
step, the singular (non-invertible) matrix is rewritten in a form that separates the
2011). Next, the singularities are removed by reducing the size of the matrix. The
In the previous two sections, I have outlined the procedure for solving the
quantum physics of the geometric diode. Another factor that needs to be accounted
for in the overall simulation is the potential distribution across the device. The
conditions. Since the potential affects the solution of the NEGF equations, which in
turn modifies the Poissons equation via the charge distribution in the device, the
127
I use the partial differential equation (PDE) solver in MATLAB (MATLAB,
2010) to solve the Poissons equation on a 2D region defined by the boundary of the
reservoirs, while the top and the bottom boundaries are floating with Neumann
boundary conditions.
Neumann
Dirichlet (V=-VD /2)
Neumann
without any charge (0=0). This is used as a starting value in the NEGF equations
1
dE G n ( E )
2 Vol .
=
Eq. VIII-15
where Vol. is the volume associated with each atomic location. This is approximated
as a2*tG where a (=2.46 ) is the lattice constant and tG (=3.4 for monolayer) is the
128
Using the charge 1 calculated from Eq. VIII-15 as a starting point for the
Poissons equation, an updated potential U1 is obtained. Convergence of the self-
consistent procedure implies that the difference between the updated (U1) and the
U1 U 0
x y
change = < RelTol
U1 + U 0
x y
Eq. VIII-16
Figure VIII-7 is repeated with a new estimate for the potential (Unew) used for
Ui
Solve Poisson
(account for DC Calculate charge
voltage and charge) distribution
predicting Unew based on several previous Us. The prediction technique that I have
used is called Anderson mixing (Eyert, 1996) and is based on minimizing the
residual vector. The residual vector is the difference in potential used in the NEGF
129
equations and the potential obtained from solving the Poisson equation. Anderson
mixing calculates weights for the linear combination of previous Us such that the
E. A simulation example
I will now compute and explain the various quantities that are obtained from
the simulation conducted on the geometry shown in Figure VIII-3. Applying a bias
of VD = 0.28 V across the graphene results in the potential distribution shown in
Figure VIII-8. This is the starting point of the iterative loop in Figure VIII-7,
Potential (V)
0.1
0.05
0
-0.05
-0.1
4
4
-9
2 3
x 10 2 -9
1 x 10
y (m) 0 0
x (m)
130
On the termination of the self-consistent iterative procedure described in the
previous section (with a RelTol of 10-4), the potential distribution takes the form
shown in Figure VIII-9.
Potential (V)
0.1
0.05
0
-0.05
-0.1
4
4
-9
2 3
x 10 2 -9
1 x 10
y (m) 0 0
x (m)
Almost all the voltage drop in the potential distribution of Figure VIII-9 occurs
equally at the two contacts. This is due to the absence of scattering in the device
region (Datta, 2000). The voltage drop also results in a higher charge density near
the contacts with the electrons accumulated at the positive- and holes at the
131
Figurre VIII-1
10 Self-co
onsistent charge distributioon
corressponding to the potentiial shown in
n Figure VIIII-9.
I calculate the
t curren
nt using Eq. VIII--12, which
h requiress finding the
transmiission as a function off energy ass shown in Figure VIIII-11.
1.5
T(E)
1
Transmission:
i i
0.5
T
0
-1 -0.5 0 0.5 1
Energy (eV
V)
Figurre VIII-11 Transmission
T n vs. energy
y obtained ffrom the self-
consisstent solutio
on.
132
At negative energies, the transmission is due to holes and at positive energies it
is due to electrons. The confinement due to the neck introduces a bandgap (0.37 eV
in this case) that prevents the transmission at energies close to zero. This bandgap
is also observed in the plot of density of states vs. energy shown in Figure VIII-12.
70
60
)
-1
Density of states (eV
50
40
30
20
10
0
-1 -0.5 0 0.5 1
Energy (eV)
Figure VIII-12 Density of states vs. energy. The bandgap and
step change with increasing energy are evident.
The density of states has a step change with energy due to the abrupt addition
of conduction channels with increasing energy. The rounding of the steps occurs due
to the interaction with the contacts that leads to broadening of the channels in the
device.
In Figure VIII-13, I plot the energy dependent current calculated from the
transmission T(E) and Fermi distribution according to Eq. VIII-12. Even though the
T(E) increases at higher |E|, the difference in Fermi distributions on the left- and
bandgap.
133
-10
x 10
1.5
I D (A/eV)
1
0.5
0
-1 -0.5 0 0.5 1
Energy (eV)
Figure VIII-13 Current vs. energy calculated from T(E) and
Fermi distribution.
The integral of the above curve gives the total current at the DC bias of 0.28 V.
Electrons and holes move in opposite directions under a drift field. However, the
geometric effect funnels both polarities of charges in the same direction. For an
equal contribution from electrons and holes to the total current, the geometric effect
gets canceled and reversing the voltage does not lead to a change in the magnitude
of current. Hence, there is no asymmetry in the I(V) curve and thereby no geometric
As seen in Figure VIII-13, electrons and holes have unequal contributions to the
current. This unequal, net p-type conduction is attributed to the sharper rise in
density of states for holes as compared to electrons near the bandgap as seen in
Figure VIII-12. Even though it is not visible in Figure VIII-11, this difference also
exists for T(E). The type of carriers that carry the larger current is determined by
the direction in which the asymmetry favors transport in conjunction with the
134
applied bias, and also by the relative location of the Fermi-level with respect to the
The effect of asymmetry and bias on the contribution of n- and p-type carriers to
the current is shown in Figure VIII-14, where I plot the current vs. energy for a
positive and a negative biased diode.
-10
x 10
3
VD=0.28 V
2
VD=-0.28 V
1
I D (A/eV)
-1
-2
-3
-1 -0.5 0 0.5 1
Energy (eV)
Figure VIII-14 Current vs. energy for opposite voltages across
the diode.
As seen before, for the positive bias, p-type carriers have a larger contribution
as they flow in the direction favored by the geometry. However, for the same reason,
at negative bias n-type carriers dominate. Overall, the current under negative bias
135
-9
x 10
2
I D (A) 0
-1
-2
-0.3 -0.2 -0.1 0 0.1 0.2 0.3
VD (V)
strongly n- or p-type by changing the location of the Fermi level w.r.t. the
voltage. In the simulation, this effect is modeled by changing the Bloch energy (2p)
at each graphene atom by an amount Eoffset as was introduced in section VII.B. This
change is equivalent to changing the position of the Fermi level with respect to the
bandstructure.
To demonstrate the effect of Eoffset, in Figure VIII-16, I plot the current vs.
energy curves for two bias voltages with different offset energies chosen to have
dominant hole (Eoffset, = 0.05 eV) or electron (Eoffset = -0.05 eV) transport for the same
136
Figurre VIII-16 Comparison
C n of currentt vs. energy y at positiv
ve-
and negative-bia
n as for majorrity electron
n (Eoffset = --0.05 eV) an
nd
hole (E
( offset = 0.05
5 eV) carrierrs.
Whe
en holes arre the majo
ority carrie
ers, the currrent for a positive vooltage is la
arger
than at a negative
e voltage im
mplying th
hat a greateer current flows alon
ng the direcction
of the funnel
f in the geome
etrically asymmetricc conductorr. The opp
posite currrent-
asymme
etry is obse
erved when
n electrons are the ma
ajority carrriers as a llarger electtron-
Cha
ange in the
e preferred direction of
o current flow whilee transition
ning from a
an n-
to p-typ
pe graphene is shown
n in Figure VIII-17. F
For Eoffset > 0.02 eV, tthe graphen
ne is
p-type and
a a high
her curren
nt flows in
n the reverrse bias. F
For Eoffset < 0.02 eV,, the
graphen
ne is n-typ
pe and a higher currrent flows at positivee bias. Thee saturatioon in
asymme
etry for neg
gative offse
et values su
uggests tha
at an optim
mal Eoffset m
may be foun
nd.
137
1.8 Eoffset (eV)
-0.06
1.6
-0.04
Asymmetry: A
1.4
-0.02
1.2 0
1 0.02
0.8 0.04
0.06
0.6
0 0.1 0.2 0.3
VD (V)
Figure VIII-17 Forward-to-reverse current asymmetry
(A=|ID(+VD)/ID (-VD)|) vs. diode voltage with changing Eoffset.
The asymmetry in the I(V) curve occurs due to unequal transmission at different
voltages. The effect of the potential on transmission increases with the magnitude
higher voltages.
As a check, the asymmetric shape used for the above simulations was flipped
along the vertical and simulated. The I(V) curve obtained was the same as the
original one mirrored at the origin, indicating no geometry related simulation error.
To further ensure that the asymmetry in I(V) occurs due to the geometry, I
simulated two symmetric structures. The first one is a sheet of armchair graphene
with 5 nm edge lengths. The I(V) curve for this conductor is shown in Figure
VIII-18. The magnitude of current flowing under forward and reverse bias is equal.
138
-7
x 10
6
4
2
I D (A)
0
-2
-4
-6
The second shape is a symmetric hourglass shaped conductor that emulates the
-10
x 10
2
1.5
1
0.5
I D (A)
0
-0.5
-1
-1.5
-2
-0.3 -0.2 -0.1 0 0.1 0.2 0.3
VD (V)
139
G.. Compariison of sim
mulated and
a experiimental c
characteriistics
The
e simulatio
ons shown above are
e on a leng
gth scale ttwo orderss of magniitude
smaller than the experimen
ntal devicess. The sma
aller dimen
nsions lead
d to conducction
sub-ban
nds that arre farther apart in energy tha
an in an a
actual dev
vice. Thereefore,
diode asymmetry
a is consisttent for th
he quantum
m simulattion, Drud
de model ((Zhu,
2011), and
a an experimental diode. To plot
p them ttogether, th
he curves are scaled by a
factor given
g in th
he legend. The smalller asymm he experimental devicce is
metry in th
partly due
d to seriies resistance. Greatter asymm
metry is observed in measured I(V)
curves at
a a larger voltage sca
ale.
The
e NEGF sim
mulation in
n Figure VIII-20 corrresponds too a diode w
with a neck
k size
of 1 nm and a shou
ulder width of 7 nm with
w an an
ngle of 60. The Drud
de simulatioon is
for a de
evice with a neck size
e of 50 nm
m, shoulderr size of 20
000 nm, an
ngle of 45, and
140
mean-free path length of 200 nm. The measured I(V) curve is for the device reported
in Figure VII-7.
141
CHAPTER IX
MESOSCOPIC DIODES
From the semiclassical analysis for the optical response of an MIM diode,
(Tien, 1963), but applicable even to a non-tunneling based device, is required for the
Platero and Aguado (Platero, 2004) have reviewed several techniques that can
than Tien and Gordons approach, all the analyses culminate in a form requiring
The simplicity of the Tien and Gordon formulation comes with the drawback
that the theory is not gauge-invariant. Another limitation of this method is that it
does not account for charge and current conservation. To ensure conservation, an
AC transport theory such as the one that solves the non-equilibrium Greens
Here I derive a formula analogous to the Tien and Gordon approach but also
142
analytical relation between the DC and illuminated characteristics. Even though
The starting point for the derivation is the NEGF theory for an illuminated
junction given by Datta and Anantram (Datta, 1992). In this section I reproduce
IX-1. It is assumed that charge transport from one contact to the other occurs
Here, the contacts refer to charge reservoirs, much larger than the device region,
held at a fixed potential. Electrons gain or lose energy through scattering in the
contacts. The interaction of charge carriers with photons occurs in the device region
through a time-varying potential V(r,t). This interaction, even though inelastic
(changes the energy of charge particles), does not cause phase incoherence.
Contact 1: 1 Contact 2: 2
Junction: V(r,t)
143
The energy-domain version of Schrdingers equation for the device in the
2 2 i R
E + VS ( r ) + G0 ( r , E ; r ', E ' ) = ( r r ' ) ( E E ' )
2m 2 ( r , E )
H 0 ( r , E ) G0R ( r , E ; r ', E ' ) = ( r r ' ) ( E E ' )
Eq. IX-1
where G0R is the retarded Greens function that represents the impulse response of
Schrdingers equation. The subscript 0 refers to the Greens function for the un-
illuminated case. The wavefunction at any energy E can be obtained from G0R. Vs is
the static potential in the device. The is the scattering (phase-breaking) time in
the contacts.
Eq. IX-2
Eq. IX-3
The avg is the time over which the current or the transmission is averaged to
find the DC component. The above equations are analogous to Eq. VIII-12 with 1/
representing the broadening () due to the contacts and |GR|2 representing GG+.
The un-illuminated case is obtained by replacing GR by G0R in Eq. IX-3.
144
Eq. IX-2 is different from the usual form for the transport equation that takes
into account the exclusion principle by counting the filled states on one contact and
the empty states on the other. The applicability of this equation for phase-coherent
perturbation term added to the RHS is the strength of the broadening due to
interaction with the field (Datta, 2005, p.259). This effect is included via a self-
energy similar to the one used for the contacts (Kienle, 2010). Under a first order
Born approximation, the solution to the equation of motion given by the modified
Eq. IX-5
Eq. IX-3 by the above expression. The GR can be computed numerically using the
technique described in Chapter VII. However in the next section, I simplify the
145
B. Projecting illuminated characteristics from DC I(V)
I propose two simplifications to the expression for GR given in Eq. IX-5. The first
is a uniform strength of interaction with the field over the device area (V(r, ) =
gauge (Eliasson, 2001, p.127). The dipole approximation requires that the
wavelength of the EM field be much larger than the size of the device. This
condition is easily satisfied for an MIM diode, and for small geometric diodes. A
further complication in geometric diodes is the field non-uniformity due to the shape
of the conductor. For this, a field strength averaged over the geometry would serve
as an initial correction.
do not differ significantly for two energy levels spaced apart by . The GR for the
two energies are similar if the photon energy is small compared to the energy of
surface, the relevant energy for comparison is the Fermi energy (<<Ef) measured
with respect to the band edge. This assumption is similar to the nearly elastic
scattering case considered by Datta (Datta, 2005, p.274). The ramifications of this
Eq. IX-6
where the Greens function in the integral is simplified under the assumption E E
as
146
G0R ( r , r '; E ) = G0R ( r , r ''; E ) G0R ( r '', r '; E ' )
Eq. IX-7
The spatial integral on the RHS of Eq. IX-6 leads to the volume of the devices
region (vol.) as the r dependence in the integrand has been removed. The Greens
Eq. IX-8
t21 ( E , E ' )
2
G ( r , r '; E ) ( E E ' ) + vol.V ( )G0R ( r , r '; E ) ( E E ' )
R
0
2
= G0R ( r , r '; E ) 2 ( E E ' )
R R
cc
+ G0 ( r , r '; E ) ( E E ' ) vol.V ( )G0 ( r , r '; E ) ( E E ' ) + cc
2
+ vol.V ( )G ( r , r '; E ) ( E E ' )
R
0
Eq. IX-9
where cc denotes complex conjugate. The term inside the square bracket has a
third term, the square of the summation over delta functions is equal to the
summation of the squares as all cross terms with different are always zero due to
t21 ( E , E ' )
G0R ( r , r '; E ) 2 ( E E ' ) + vol . V ( ) G0R ( r , r '; E ) ( E E ' )
2 2 2
147
Eq. IX-10
energy correction to the Greens function for the un-illuminated case. This leads to
e 2
I illum =
T dE dE ' dr dr '
r2
r1
G R ( r, r '; E ) 2 ( E E ' ) 2
G0R ( r ', r ; E ' ) ( E ' E )
f1 ( E ' ) f2 ( E ) +
0
4 2 ( r , E ) ( r ', E ' ) 4 2 ( r , E ) ( r ', E ' )
G R ( r, r '; E ) 2 ( E E ' )
0 f ( E ' )
4 2 ( r , E ) ( r ', E ' )
1
2 2
Vol V ( ) R 2
G0 ( r ', r ; E ' ) ( E ' E )
f ( E )
4 2 ( r , E ) ( r ', E ' )
2
Eq. IX-11
I now perform the integral with respect to E in Eq. IX-11 using the sifting
property of the delta function. Under the condition <<Ef, I also approximate that
does not vary significantly from E to E+. The validity of this is in-line with the
approximation made for GR in that the interaction of an electron with the reservoirs
is similar at two closely spaced energies. With this simplification, the illuminated
current is given by
148
e 2
I illum =
T dE dr dr '
r2
r1
2 2
G0R ( r , r '; E ) G0R ( r ', r ; E )
2 f1 ( E ) 2 f2 ( E ) +
4 ( r , E ) ( r ', E ) 4 ( r , E ) ( r ', E )
R 2
G ( r , r '; E )
2 0 f ( E )
4 ( r , E ) ( r ', E )
1
2 2
Vol V ( ) 2
G0R ( r ', r ; E )
2 f2 ( E )
4 ( r , E ) ( r ', E )
Eq. IX-12
e 2
I illum =
T dE dr dr '
r2
r1
2
G0R ( r , r '; E )
2
4 ( r , E ) ( r ', E )
{ f1 ( E ) f 2 ( E )} +
R 2
2 G ( r , r '; E )
Vol V ( ) { }
2 0
f ( E ) f ( E )
4 2 ( r , E ) ( r ', E )
1 2
Eq. IX-13
The second term in the above equation that represents the additional current
applied between the two contacts (VD). Assuming that contact 2 is the ground, Eq.
149
e 2
I illum =
T dE dr dr '
r2
r1
2
G0R ( r , r '; E )
2
4 ( r , E ) ( r ', E )
{ f ( E eV D ) f ( E )} +
2
2 G0R ( r , r '; E )
Vol V ( )
2
{
4 2 ( r , E ) ( r ', E )
f ( E ( eVD + ) ) f ( E )}
Eq. IX-14
I illum (VD ) = I dark (VD ) + Vol 2 V ( ) I dark VD +
2
e
Eq. IX-15
C. Discussion
a dark I(V) curve, similar to the first order version of Eq. V-6 depicted graphically
using a piece-wise linear dark I(V) curve in Figure V-9. Although the equation for
diode in Figure V-2, Eq. IX-15 indicates that the interaction of the electrons in the
device region is equivalent to the modulation of electron energies in the contact.
This simplified picture for the interaction emerges because the electronic transport
additional photon energy () acquired by the electrons in the device region. The
interaction with the photon from the device region to the contacts.
150
The assumption of <<Ef is the most significant consideration of the
derivation. Earlier, I stated this condition without analyzing its physical
be a strong function of energy as seen in Figure II-2(a). However, in the case of MIM
diodes the photon-electron interaction does occur only in the contacts due to the
absence of electrons in the insulator. This is the basis for the semiclassical theory
narrow energy ranges, the assumption of constant transport behavior has greater
validity. Ultimately, the effect of the added photon energy is relative to the existing
electron-energy. If this energy is large, such that the transmission is highly likely,
the change in transmission with additional photon energy will be small. A measure
of interest for the existing electron energy is the Fermi energy and hence the
condition <<Ef.
A final point of discussion concerns the material for the thin-film used in the
device region. Eq. IX-15 is applicable under the assumption that the photon energy
(Eph) is small compared to the Fermi energy (Ef) . This depends on the value of the
Ef. For metals, Ef is on the order of a few electron-volts so that the result holds even
151
CHAPTER X
optical rectennas with MIM and geometric diodes as potential rectifiers. Efficient
optical rectennas require diodes that are inherently fast, highly nonlinear and/or
asymmetric and have the product of resistance and capacitance (RC) smaller than
Below, I have grouped my conclusions into four sections based upon metal-
MIM diodes, and geometric diodes. This is followed by suggestions for future work.
A. Metal-insulator diodes
diodes can have improved current-voltage I(V) characteristics satisfying both these
requirements. Two mechanisms for improved MIIM diodes, resonant tunneling and
also via book values for thermodynamic quantities used in a Gibbs free energy
analysis.
with low resistance and capacitance are required for the RC time constant of the
rectenna to be smaller than the reciprocal of the operating frequency. This ensures
adequate bandwidth and efficient coupling between the antenna and the diode.
operation. Infrared detectors with reduced efficiency due to the high RC of MIM
diodes may still be practical. However, the impact of RC on the power conversion
I have considered two approaches for improving the efficiency of rectennas. The
more radical approach is to use the planar geometric diode, which unlike MIM
diodes does not have a parallel-plate structure and thereby has a smaller RC time
constant. The less radical approach uses MIM diodes in a traveling-wave structure
to obtain a distributed rectifier that removes the RC limitation on bandwidth.
responsivity and reduced resistance per unit area. A multi-insulator diode can be
at shorter wavelengths with scope for further improvement. Both the impedance
153
and the responsivity of the traveling-wave detector allow the detection bandwidth to
extend over several microns in the MWIR and the LWIR ranges. Its efficiency is
frequencies using a semiclassical approach that accounts for the photon energy of
radiation samples the DC I(V) curve at discrete voltage steps separated by the
photon energy (divided by the electronic charge). As a result, the diode resistance
and responsivity differ from their classical values. At optical frequencies, a diode
efficiency.
The power efficiency in an optical rectifier with a non-ideal diode improves with
a stronger AC signal. For solar radiation, the maximum illumination area that can
be tapped by one rectenna element corresponds to a coherence area of 283 m2. This
limits the strength of the AC voltage at the diode. The AC voltage can be increased
by making the diode smaller, effectively, increasing the field strength for the same
input power. With the AC signal signal strength larger than the photon energy
divided by electron charge, the first-order assumption breaks down and higher
the first-order semiclassical models is 50%. On the other hand, while the small-
signal circuit model gives 50% efficient rectennas, the large-signal classical circuit
154
D. Geometric diodes
obtaining low capacitance and low resistance diodes for optical rectennas. As
theorized, asymmetric constrictions on the length-scale of the mean free path length
diodes fabricated using graphene yielded characteristics that are consistent with
Hamiltonian and solve for the charge and potential self-consistently. Although
these simulations are performed device geometries that are much smaller than
A generalized semiclassical analysis was carried out for predicting the optical
compared to the Fermi energy measured from the band edge, the response turns out
to be similar to that for MIM diodes.
E. Future work
The double insulator tunnel diodes that I have simulated are only
constants may lead to diodes with higher nonlinearity and lower resistance. An
155
tunnel diode with a low-barrier low-dielectric-constant insulator. Doing so may
the barrier height and dielectric constant. Therefore, implementing this approach
insulators.
The tunnel diode simulations were carried out using a transfer-matrix method
simulating the MIIM diode is to consider the occupation of the resonant well. A
better estimate for the effective mass, measured or extracted from curve-fitting, is
also desirable.
signal circuit analysis as well as the semiclassical analysis for the MIM diode. In
applications such as infrared detectors and solar cells, the radiation is expected to
be broadband. For the rectenna, not only does this imply that the antenna should be
broadband but also requires the diode to rectify each frequency optimally. If and
how this is possible requires consideration both from the perspective of circuit
at efficiencies higher than 50%. This may require including the higher-order terms
For the geometric diode, so far, only sub-10 nm geometries have been simulated
binding model that is used to define the device Hamiltonian. A simpler approach,
e.g., an effective mass model, may allow larger geometries to be simulated quantum
including interaction with photons and phonons can be added onto the basic
before and after the absorption of an energy quantum equal to the photon energy. I
have provided qualitative implications of this assumption. It would be worthwhile
to quantify the error incurred and also to analyze the applicability of the derived
157
REFERENCES
Abdel-Rahman, M.R., Gonzalez, F.J., Zummo, G., Middleton, C.F. & Boreman, G.D.,
2004. Antenna-coupled MOM diodes for dual-band detection in MMW and LWIR.,
Proc. SPIE, 5410, 238.
ADOSE, 2008. Reliable Application Specific Detector of Road Users with Vehicle On-
board Sensors. European Commision.
Agarwal, G.S., Gbur, G. & Wolf, E., 2004. Coherence properties of sunlight. Optics
Letters, 29(5), pp.459-61.
Ashcroft, N.W. & Mermin, N.D., 1976. Solid state physics. Orlando: Harcourt
College Publishers.
Bailey, R.L., 1972. A proposed new concept for a solar-energy converter. Journal of
Engineering for Power, (73).
Bain, R.J.P. & Donaldson, G.B., 1985. Sputtered all-niobium Josephson tunnel
junctions with barrier oxide protected by gold. J. of Phys. C: Solid State Phys.,
18(12), pp.2539-48.
Berger, C., Song, Z., Li, X., Wu, X., Brown, N. et al., 2006. Electronic Confinement
and Coherence in Patterned Epitaxial Graphene. Science, 312(5777), pp.1191-96.
Berland, B., 2003. Photovoltaic Technologies Beyond the Horizon: Optical Rectenna
Solar Cell. Final report. NREL Report No. SR-520-33263.
Born, M. & Wolf, E., 1999. Principles of Optics. 7th ed. Cambridge University Press.
Brown, W.C., 1976. Optimization of the Efficiency and Other Properties of the
Rectenna Element., 1976.
Brown, W.C., 1984. The History of Power Transmission by Radio Waves. Microwave
Theory and Techniques, IEEE Transactions on, 32(9), pp.1230-42.
Brown, E.R., 2004. A system-level analysis of Schottky diodes for incoherent THz
imaging arrays. Solid-State Electronics, 48, pp.2051-53.
158
Bttner, H. & Gerlach, E., 1968. Microwave assisted tunneling between
superconductors. Phys. Lett., 27A(4), pp.226-27.
Camp, M. & Lecchini, S.M.A., 1965. The work function of polycrystalline tungsten
foil. Proc. of the Physical Society, 85(4), p.815.
Chapline, M.G. & Wang, S.X., 2007. Analytical formula for the tunneling current
versus voltage for multilayer barrier structures. Journal of Applied Physics, 101(8),
p.083706.
Corkish, R., Green, M.A. & Puzzer, T., 2002. Solar energy collection by antennas.
Solar Energy, 73(6), pp.395-401.
Dagenais, M., Choi, K., Yesilkoy, F., Chryssis, A.N. & Peckerar, M.C., 2010. Solar
Spectrum Rectification Using Nano-Antennas and Tunneling Diodes. Proc. of SPIE,
7605, pp.76050E-1.
Datta (a), S., 2009. ECE 659 Lecture 19: Self Energy.
http://nanohub.org/resources/6430.
Datta, S., 2000. Nanoscale device modeling: the Green's function method.
Superlattices and Microstructures, 28(4), pp.253-78.
Dayem, A.H. & Martin, R.J., 1962. Quantum Interaction of Microwave Radiation
with Tunneling Between Superconductors. Phys. Rev. Lett., 8(6), pp.246-48.
159
de Vos, A., 1992. Endoreversible thermodynamics of solar energy conversion. Oxford:
Oxford Science.
Dicke, R.H., Beringer, R., Kyhl, R.L. & Vane, A.B., 1946. Atmospheric Absorption
Measurements with a Microwave Radiometer. Phys. Rev. , 70, pp.340-48.
Donges, A., 1998. The coherence length of black-body radiation. Eur. J. Phys. , 19,
pp.245-49.
Eliasson, B. & Moddel, G., 2005. Detectors at the Intersection of Photons and
Electromagnetic Fields or, Where Einstein Meets Maxwell. In Spring Meeting of the
Materials Research Society. San Francisco, March 28-April 1, 2005.
Estes, M.J. & Moddel, G., 2005. Terahertz interconnect system and applications. US
Patent 6,967,347.
Estes, M.J. & Moddel, G., 2006. Surface plasmon devices. U S Patent No. 7,010,183.
Evenson, K.M., Wells, J.S., Matarrese, L.M. & Elwell, L.B., 1970. Absolute
Frequency Measurements of the 28- and 78- m cw Water Vapor LASER Lines.
Applied Physics Letters, 16(4), pp.159-62.
Ferry, D.K., Goodnick, S.M. & Bird, J., 2009. Transport in Nanostructures. 2nd ed.
Cambridge: Cambridge University Press.
Fox, M., 2001. Optical Properties of Solids. 1st ed. Oxford University Press.
Fumeaux, C., Herrmann, W., Kneubhl, F.K. & Rothuizen, H., 1998. Nanometer
thin-film Ni-NiO-Ni diodes for detection and mixing of 30 THz radiation. Infrared
Physics & Technology, 39(3), pp.123-83.
160
Gonzlez, F.J. & Boreman, G.D., 2005. Comparison of dipole, bowtie, spiral and log-
periodic IR antennas. Infrared Physics & Technology, 46(5), pp.418-28.
Grover, S., Dmitriyeva, O., Estes, M.J. & Moddel, G., 2010. Traveling-Wave
Metal/Insulator/Metal Diodes for Improved Infrared Bandwidth and Efficiency of
Antenna-Coupled Rectifiers. Nanotechnology, IEEE Transactions on, 9(6), pp.716-
22.
Gunapala, S., Bandara, S., Hill, C., Ting, D., Liu, J. et al., 2008. Quantum Well and
Quantum Dot Based Detector Arrays for Infrared Imaging. Mater. Res. Soc. Symp.
Proc., 1076.
Guo, Y., Gu, B.-L., Yu, J.-Z., Zeng, Z. & Kawazoe, Y., 1998. Resonant tunneling in
step-barrier structures under an applied electric field. Journal of Applied Physics,
84(2), pp.918-24.
Hagerty, J.A., Helmbrecht, F.B., McCalpin, W.H., Zane, R. & Popovic, Z.B., 2004.
Recycling ambient microwave energy with broad-band rectenna arrays. Microwave
Theory and Techniques, IEEE Transactions on, 52(3), pp.1014-24.
Hartstein, A. & Weinberg, Z.A., 1978. On the nature of the image force in quantum
mechanics with application to photon assisted tunnelling and photoemission.
Journal of Physics C: Solid State Physics, 11(11), p.L469.
Haug, H.J.W. & Jauho, A., 2008. Quantum Kinetics in Transport and Optics of
Semiconductors. 2nd ed. Berlin: Springer-Verlag.
Hegyi, B., Csurgay, A. & Porod, W., 2007. Investigation of the nonlinearity
properties of the DC I-V characteristics of metal-insulator-metal (MIM) tunnel
diodes with double-layer insulators. Journal of Computational Electronics, 6,
pp.159-62. 10.1007/s10825-006-0083-9.
Hobbs, P.C., Laibowitz, R.B., Libsch, F.R., LaBianca, N.C. & Chiniwalla, P.P., 2007.
Efficient waveguide-integrated tunnel junction detectors at 1.6 m. Opt. Express,
15(25), pp.16376-89.
Huang, J.-S., Feichtner, T., Biagioni, P. & Hecht, B., 2009. Impedance Matching
and Emission Properties of Nanoantennas in an Optical Nanocircuit. Nano Letters,
9(5), p.18971902.
Hbers, H.-W., Schwaab, G.W. & Rser, H.P., 1994. Video detection and mixing
performance of GaAs Schottky-barrier diodes at 30 THz and comparison with metal-
insulator-metal diodes. Journal of Applied Physics, 75(8), pp.4243-48.
161
Jablan, M., Buljan, H. & Soljai, M., 2009. Plasmonics in graphene at infrared
frequencies. Phys. Rev. B., 80, p.245435.
Jayaweera, P.V., Matsik, S.G., Perera, A.G., Liu, H.C., Buchanan, M. et al., 2008.
Uncooled infrared detectors for 35 mu m and beyond. Applied Physics Letters,
93(2), p.021105.
Jonsson, B. & Eng, S.T., 1990. Solving the Schrodinger Equatoin in Arbitrary
Quantum-Well Potential Profiles Using the Transfer Matrix Method. IEEE J. of
Quantum Elec., 26(11), pp.2025-35.
Kale, B.M., 1985. Electron tunneling devices in optics. Optical Engineering, 24(2),
pp.267-74.
Kang, Y., Kang, J. & Chang, K.J., 2008. Electronic structure of graphene and
doping effect on SiO2. Phys. Rev. B, 78, p.115404.
Kazemi, H., Shinohara, K., Nagy, G., Ha, W., Lail, B. et al., 2007. First THz and IR
characterization of nanometer-scaled antenna-coupled InGaAs/InP Schottky-diode
detectors for room temperature infrared imaging. Infrared Technology and
Applications XXXIII, 6542(1), p.65421.
Kittel, C., 1996. Introduction to Solid State Physics. 7th ed. John Wiley & Sons.
Kotter, D.K., Novack, S.D., Slafer, W.D. & Pinhero, P., 2008. Solar Nantenna
Electromagnetic Collectors. ASME Conference Proceedings, 2008(43208), pp.409-15.
Landsberg, P.T. & Tonge, G., 1979. Thermodynamics of the conversion of diluted
radiation. J. Phys. A: Math. Gen., 12(4), pp.551-62.
Lent, C.S. & Kirkner, D.J., 1990. The quantum transmitting boundary method.
Journal of Applied Physics, 67(10), pp.6353-59.
Liu, J. & Giordano, N., 1990. Nonlinear response of a mesoscopic system. Physica B,
165&166, pp.279-80.
162
Li, H., Wang, L., Lan, Z. & Zheng, Y., 2009. Generalized transfer matrix theory of
electronic transport through a graphene waveguide. Phys. Rev. B, 79(15), p.155429.
Luryi, S., Xu, J. & Zaslavsky, A., 2010. Future Trends in Microelectronics: From
Nanophotonics to Sensors to Energy. John Wiley \& Sons.
Marchenko, O., Kazantsev, S. & Windholz, L., 2007. Demonstrational Optics Part 2,
Coherent and Statistical Optics. Springer.
Marks, A.M., 1984. Device for conversion of light power to electric power. US Patent
No. 4445050.
Matsumoto, Y., Hanajiri, T., Toyabe, T. & Sugano, T., 1996. Single Electron Device
with Asymmetric Tunnel Barriers. Jpn. J. Appl. Phys., 35, pp.1126-31.
Middlebrook, C.T., Krenz, P.M., Lail, B.A. & Boreman, G.D., 2008. Infrared phased-
array antenna. Microwave and Optical Technology Letters, 50(3), pp.719-23.
Midrio, M., Boscolo, S., Locatelli, A., Modotto, D., Angelis, C.D. et al., 2010. Flared
Monopole Antennas for 10 m Energy Harvesting. In Proc. 40th European
Microwave Conf. Paris, France, 2010.
Moddel, G., 2001. Fractional bandwidth normalization for optical spectra with
application to the solar blackbody spectrum. Applied Optics, 40(3), pp.413-16.
Moddel, G. & Eliasson, B., 2004. High speed electron tunneling device and
applications. U.S. Patent No. 6,756,649.
Novoselov, K.S., Geim, A.K., Morozov, S.V., Jiang, D., Katsnelson, M.I. et al., 2005.
Two-dimensional gas of massless Dirac fermions in graphene. Nature, 438, pp.197-
200.
163
Palik, E.D., 1985. Handbook of Optical Constants of Solids. Orlando: Academic
Press, Inc.
Palik, E.D., 1991. Handbook of Optical Constants of Solids II. San Diego: Academic
Press, Inc.
Pedersen, M.H. & Bttiker, M., 1998. Scattering theory of photon-assisted electron
transport. Phys. Rev. B, 58(19), pp.12993-3006.
Perello, D.J., Lim, S.C., Chae, S.J..L.I., Kim, M.J., Lee, Y.H. et al., 2010. Current
anisotropy of carbon nanotube diodes: Voltage and work function dependence. Appl.
Phys. Lett., 96, p.263107.
Periasamy, P., Bergeson, J.D., Parilla, P.A., Ginley, D.S. & O'Hayre, R.P., 2010.
Metal-insulator-metal point-contact diodes as a rectifier for rectenna., 2010.
Phiar, 2002. Infrared Photon Field Detectors. University of Colorado at Boulder and
Phiar Corporation.
Probst, O.M., 2002. Tunneling through arbitrary potential barriers and the
apparent barrier height. American Journal of Physics, 70(11), pp.1110-16.
Puri, A. & Schaich, W.L., 1983. Comparison of image-potential theories. Phys. Rev.
B, 28(4), pp.1781-84.
Raether, H., 1988. Surface Plasmons, vol. 111 of Springer-Verlag Tracts in Modern
Physics. New York: Springer-Verlag.
Ran, Q., Gao, M., Guan, X., Wang, Y. & Yu, Z., 2009. First-principles investigation
on bonding formation and electronic structure of metal-graphene contacts. Appl.
Phys. Lett., 94(10), p.103511.
Reich, S., Maultzsch, J., Thomsen, C. & Ordejn, P., 2002. Tight-binding description
of graphene. Phys. Rev. B, 66, p.035412.
Reinhart, K.F. & al., e., 2009. Low-cost Approach for Far-Infrared Sensor Arrays for
Hot-Spot Detection in Automotive Night Vision Systems. In Advanced Microsystems
for Automotive Applications. Springer Berlin Heidelberg.
164
Riccius, H.D. & Siemsen, K.J., 1984. Point-contact diodes. Applied Physics A:
Materials Science & Processing, 35, pp.67-74. 10.1007/BF00620632.
Richards, P.L., 1994. Bolometers for infrared and millimeter waves. J. Appl. Phys.,
76(1), pp.1-24.
Richards, P.L., 1997. Bolometers for infrared and millimeter waves. J. Appl. Phys.,
76(1), pp.1-24.
Rockwell, S., Lim, D., Bosco, B.A., Baker, J.H., Eliasson, B. et al., 2007.
Characterization and Modeling of Metal/Double-Insulator/Metal Diodes for
Millimeter Wave Wireless Receiver Applications. In Radio Frequency Integrated
Circuits (RFIC) Symposium, IEEE., 2007.
Roddaro, S., Pingue, P., Piazza, V., Pellegrini, V. & Beltram, F., 2007. The Optical
Visibility of Graphene: Interference Colors of Ultrathin Graphite on SiO2. Nano
Lett., 7(9), p.27072710.
Rogalski, A., 2000. Infrared Detectors. Gordon and Breach Science Publishers.
Rogalski, A., 2003. Infrared detectors: status and trends. Progress in Quantum
Electronics, 27(2-3), pp.59-210.
Rungger, I. & Sanvito, S., 2008. Algorithm for the construction of self-energies for
electronic transport calculations based on singularity elimination and singular
value decomposition. Phys. Rev. B, 78, p.035407.
Sanchez, A., C. F. Davis, J., Liu, K.C. & Javan, A., 1978. The MOM tunneling diode:
Theoretical estimate of its performance at microwave and infrared frequencies.
Journal of Applied Physics, 49(10), pp.5270-77.
Sancho, M.P.L., Sancho, J.M.L. & Rubio, J., 1984. Quick iterative scheme for the
calculation of transfer matrices: application to Mo(100). J. Phys. F: Met. Phys., 14,
pp.1205-15.
Sarehraz, M., 2005. Novel rectenna for collection of infrared and visible radiation.
PhD Thesis. University of South Florida.
Sarehraz, M., Buckle, K., Weller, T., Stefanakos, E., Bhansali, S. et al., 2005.
Rectenna developments for solar energy collection. In 31st IEEE Photovoltaic
Specialists Conference., 2005.
Schift, H., 2008. Nanoimprint lithography: An old story in modern times? A review.
Journal of Vac. Sc. & Tech. B: Microelectronics and Nanometer Structures, 26(2),
pp.458-80.
165
Sedra, A.S. & Smith, K.C., 1997. Microelectronic circuits. 4th ed. New york: Oxford
University Press.
Sedra, A.S. & Smith, K.C., 2004. Microelectronic Circuits. 5th ed. New York: Oxford
University Press.
Shinohara, N. & Matsumoto, H., 1998. Experimental study of large rectenna array
for microwave energy transmission. Microwave Theory and Techniques, IEEE
Transactions on, 46(3), pp.261-68.
Simmons, J.G., 1963. Generalized Formula for the Electric Tunnel Effect between
Similar Electrodes Separated by a Thin Insulating Film. Journal of Applied
Physics, 34(6), pp.1793-803.
Singh, P., Kaneria, S., Anugonda, V.S., Chen, H.M., Wang, X.Q. et al., 2006.
Prototype silicon micropower supply for sensors. Sensors Journal, IEEE, 6(1),
pp.211-22.
Solymar, L. & Walsh, D., 2010. Electrical Properties of Materials. 8th ed. Oxford
University Press.
Song, A.M., Lorke, A., Kriele, A., Kotthaus, J.P., Wegscheider, W. et al., 1998.
Nonlinear Electron Transport in an Asymmetric Microjunction: A Ballistic Rectifier.
Phys. Rev. Lett., 80(17), pp.3831-34.
Spieler, H., 2005. Semicondutor Detector Systems. New York: Oxford University
Press Inc.
unji, M. & Marui, L., 1991. Dynamical effects in electron tunneling: Self-
consistent semiclassical image potentials. Phys. Rev. B, 44(16), pp.9092-95.
Synrad, 2010. Series 48, L version lasers: Operator's manual. [Online] Available at:
http://www.synrad.com/Manuals/48series_manual.pdf.
Sze, S.M. & Ng, K.K., 2006. Physics of Semiconductor Devices. Wiley-Interscience.
166
Thouless, D.J., 1980. The effect of inelastic electron scattering on the conductivity of
very thin wires. Solid State Comm., 34, pp.683-85.
Tien, P.K. & Gordon, J.P., 1963. Multiphoton Process Observed in the Interaction of
Microwave Fields with the Tunneling between Superconductor Films. Phys. Rev.,
129(2), pp.647-51.
Tiwari, B., Bean, J.A., Szakmany, G., Bernstein, G.H., Fay, P. et al., 2009.
Controlled etching and regrowth of tunnel oxide for antenna-coupled metal-oxide-
metal diodes. Journal of Vacuum Science & Technology B: Microelectronics and
Nanometer Structures, 27(5), pp.2153-60.
Torrey, H.C. & Whitmer, C.A., 1964. Crystal Rectifiers. Lexington: Boston Technical
Publishers, Inc.
Tucker, J.R., 1979. Quantum limited detection in tunnel junction mixers. IEEE J. of
Quantum Electronics, QE-15(11), pp.1234-58.
Tucker, J.R. & Feldman, M.J., 1985. Quantum detection at millimeter wavelengths.
Reviews of Modern Physics, 57(4), pp.1055-113.
Tucker, J.R. & Millea, M.F., 1978. Photon detection in nonlinear tunneling devices.
Appl. Phys. Lett., 33(7), pp.611-13.
Umerski, A., 1997. Closed-form solutions to surface Green's functions. Phys. Rev. B,
55(8), pp.5266-75.
Ventra, M.D., 2008. Electrical Transport in Nanoscale Systems. 1st ed. Cambridge:
Cambridge University Press.
Weiss, M.D., Eliasson, B.J. & Moddel, G., 2003. Device integrated antenna for use
in resonant and non-resonant modes and method. US Patent No. 6664562.
Wilke, I., Oppliger, Y., Herrmann, W. & Kneubhl, F.K., 1994. Nanometer thin-film
Ni-NiO-Ni diodes for 30 THz radiation. Applied Physics A: Materials Science &
processing, 58(4), pp.329-41.
Wu, Y. & Childs, P.A., 2011. Conductance of Graphene Nanoribbon Junctions and
the Tight Binding Model. Nanoscale Research Letters, 6(62), pp.1-5.
167
Yon, J.J., Mottin, E., Biancardini, L., Letellier, L. & Tissot, J.L., 2003. Infrared
Microbolometer Sensors and Their Application in Automotive Safety. In Valldorf, J.
& Gessner, W. Advanced Microsystems for Automotive Applications 2003. Springer
Berlin Heidelberg. pp.137-57.
Yoo, T. & Chang, K., 1992. Theoretical and Experimental Development of 10 and 35
GHz Rectennas. IEEE Tran. on Microwave Theory and Techniques, 40(6), pp.1259-
66.
Zhu, Z., 2010. Geometric diodes fabrication using focused ion beam etching of Al-
film. Private communication.
Zhu, Z., Grover, S., Krueger, K. & Moddel, G., 2011. Optical rectenna solar cells
using graphene geometric diodes. accepted in the 37th IEEE PVSC.
Zia, R., Selker, M.D., Catrysse, P.B. & Brongersma, M.L., 2004. Geometries and
materials for subwavelength surface plasmon modes. J. Opt. Soc. Am. A, 21(12),
pp.2442-46.
168
APPENDIX - A
A. Introduction
the nonlinear dependence of current on diode voltage. Another method for changing
voltage (VG) that modifies the band profile of the MIM diode as shown in Figure A-.
A positive gate bias causes the insulator band profile to bend downwards, which
decreases the effective distance for Fowler Nordheim (FN) tunneling of electrons. A
change in the band potential by fractions of a volt can result in a large change in the
tunnel current. A structural schematic for such a device, referred to as the MIM
169
Figure A-2 Structural schematic for a metal-insulator-metal
field effect transistor
MIMFET upto 100 kHz. Finally, I estimate its performance capability related to the
B. Background
barrier tunnel transistor (Jang, 2003) using an MIM diode. The basic requirement
current and a low thermionic-leakage. Effective coupling of the gate voltage into the
tunnel insulator (Shaker, 2003) requires that the ratio of dielectric constants g/t be
(i) anodic oxidation of a metallic strip (titanium) using an AFM tip in the presence
of moisture to produce the tunnel junction (Snow, 1998), and (ii) sidewall oxidation
of a metal strip followed by a second metal deposition (Fukushima, 1999). The AFM
Ti/TiOx/Ti tunnel junction for which results were reported only at 77 K. The second
fabrication method uses the same Ti-based barrier with a 30 nm wide tunnel
170
insulator. Even for relatively low current densities, both these devices required high
VDS.
Desirable performance of MIMFETs has not been achieved in any of the
previously reported results. In some of these works (Snow, 1998), (Fujimaru, 1999),
field effect measurements are only reported for low temperatures. Extremely low
current and a large gate-voltage that causes only a small change in tunnel current
also limit the applicability of these devices. Moreover, simulations (Snow, 1998)
predict large source-drain currents, which are far from the measured values. Two
possible explanations for the inconsistency between the simulated and experimental
results are, (i) the simulations are on a structure largely different from the
fabricated device, (ii) tunneling is not the dominant conduction mechanism due to
Al2O3 as the gate oxide since it has a higher dielectric constant than SiO2, which
leads to better coupling between the gate and the tunnel dielectric.
As shown in Figure A-2, the targeted MIMFET geometry has the source, drain,
and gate electrodes separated by the tunnel and gate insulators on a sub-10 nm
scale. Previous fabrication methods are limited in the choice and design of the
tunnel-insulator. Also, in both those methods, surface conduction paths can possibly
connect the source and the drain. I describe two alternative techniques for
fabricating the MIMFETs. These methods differ in the orientation of the MIM
tunnel diode and the gate electrode. I classify them as having a horizontal or a
171
conducttion pathss. Before proceedin
ng, I ack
knowledge the supp
port of P
Phiar
Corpora
ation in the
e developm
ment of this process.
1. Hor
rizontal ga
ate MIMF
FET
Thiss approach
h is simila
ar to the sidewall
s ox
xidation m
method (Fu
ujimaru, 19
999),
except that
t I depo
osit the tun
nnel insula
ator(s) insttead of oxiidizing the sidewall. This
techniqu
ue of creating a nanoscale-t
n thin, vertiically-orien
nted tunnel-insulatoor is
depicted
d in Figure
e A-. The tu
unnel insulator is con
nformally d
deposited oover a step
p and
anisotro
opically etcched to crea
ate a thin sidewall
s w ithout the use of lithoography.
172
The device is fabricated on an oxidized silicon wafer using the following
sequence of steps:
i. Pattern and deposit contact pads (Cr/Au) using metal
evaporation and lift-off.
ii. Pattern and deposit gate metal electrode (Al) using evaporation
and lift-off.
iii. Oxidize the gate metal to produce the gate oxide (Al2O3) using
plasma oxidation done in the sputter deposition chamber.
iv. Conformally deposit source metal (NbN) by sputtering Nb in
presence of nitrogen.
v. Conformally deposit SiO2 using PECVD or sputtering. The SiO2
serves as a spacer layer.
vi. Pattern and deposit a Cr mask using lift-off. This serves as an
etch mask for the source.
vii. Etch the SiO2 using reactive ion etching (RIE) in CHF3+Ar
plasma with endpoint detection.
viii. Etch the source metal using RIE using a mixture of CF4 and O2.
ix. Sputter deposit tunnel-insulator(s) and drain metal.
x. Pattern and etch drain metal and tunnel-insulator(s) in CF4
and O2 plasma, to obtain a narrow strip of materials
overlapping with the patterned source metal.
xi. Pattern and deposit metal (Cr/Au) to contact to the drain
electrode.
section and the top-view microscope image is shown in Figure A-5. The fabricated
device resembles the intended structure, except for the undercut and the sloped
sidewall of the SiO2 region. The undercut can be reduced by decreasing the height of
the NbN source region. The sloped SiO2 sidewall does not affect the performance of
the MIMFET.
173
Figurre A-5 Crosss-section TEM
T imagee of the M
MIMFET. Th he
variouus regions of
o the cross--section sch
hematic are marked with
yellow
w lines.
In the horizo
ontal MIM
MFET fab
bricated ab
bove, the tunnel in
nsulator llayer
separates the sou
urce and the drain electrode s, preventting a dirrect connecction
between
n them. Thus, a defect or a surface coonducting layer cann
not be forrmed
between
n the sourrce and the
t drain, ensuring that tun
nneling is the dominant
conducttion mecha
anism. How
wever, this also impliies an unw
wanted (tun
nnel) insullator
below th
he drain electrode
e th
hat increasses the sep
paration between thee gate and
d the
tunnel diode,
d whicch decrease
es the effecctiveness oof the gate. Also, as th
he MIM tu
unnel
junction
n is not forrmed in a single step
p, the deviice is pron
ne to interfface impurrities
between
n the sourrce metal and the tunnel in
nsulator. T
This can b
be avoided
d by
approprriate clean
ning or precondition
p ning beforre the deeposition of the tu
unnel
insulato
or.
Any
y technique
e for fabriicating the
e MIMFET
T, in which
h the tunn
nel-insulatoor is
deposite
ed instead of grown, gives the freedom too design th
he tunnel b
barrier. Th
his is
174
the case
e with both
h the horizzontal-gate
e MIMFET
T discussed
d above an
nd the verttical-
In the
t vertica
al gate MIM
MFET, a gate
g electrrode is pla
aced on thee sidewall of a
electrica
ally insulatted from th
he stack by
y a gate inssulator.
vacuum
m, which ma
akes it lesss likely to have
h defectts along th
he interfacees across w
which
The
e fabrication of this sttructure was done by
y Limin Ca
ao using thee diode proocess
developed at Phia
ar Corpora
ation. A CV
VD SiO2 w
was depositted as the gate insullator
followed
d by the ga
ate metal de
eposition using
u evapooration.
Cha
aracterizatiion of the
e vertical MIMFET is given in the neext section
n. As
describe
ed there, no modulatiion of the tunnel currrent was acchieved usiing the verrtical
MIMFE
ET. The prrimary reason for this is the u
undercut in
n the MIM
M stack, w
which
occurs while
w defin
ning the ve
ertical edg
ges using tthe RIE. In
n Figure A
A-7, I show
w the
175
TEM micrograph of
o an MIM
M diode, wh
here the tra
apezoidal sstructure iss a metal-m
mask
used fo
or etching. The CF4 based RIIE etch leeads to a large und
dercut and
d the
subsequ
uent deposition of in
nsulator prroduces airr voids. On
n realizing
g this prob
blem,
work on
n this struccture was discontinue
d ed.
3. Tun
nnel insula
ator desig
gn
insulato
or(s) with appropriate
a e dielectricc constant and barrieer height p
profile. To sshow
the effe
ect of dielectric consttant, I com
mpare the eelectrostattic potentia
als in a sin
ngle-
and a triple-insul
t lator barrie
er for diffe
erent gate voltages. The dimen
nsions and
d the
dielectriic constan
nt profiles for both these
t barriiers are giiven in Fiigure A-8. The
the adva
antage of a low-high--low dielecttric profile in the trip
ple-insulatoor diode.
As shown
s in th
he figure, both the tu
unnel-diod
des are held
d fixed at a bias of 0.5 V,
and the
e potential is calculatted when VG is 0 and
d 2 V. At VG = 0 V, b
both the diiodes
potentia
al of the triple-insulator is larrger than the single-insulator. This happens
176
due to the dielectric profile of the triple-insulator where the low- insulators
0.8
that most of the potential drop from the drain to the source occurs in the outer
the same change in gate voltage, a stronger gate effect is obtained by using a low-
Figure A-9.
177
Figurre A-9 Barriier profile fo
or (a) singlee- and (b) triiple-insulattor
tunneel diodes.
The
e triple-barrrier has several adv
vantages. F
First, the ttriangular quantum--well
as the gate
g voltage
e is varied.. The prese
ence or abssence of a rresonant leevel in the w
well,
shown in
i Figure A-8,
A the ch
hange in po
otential due to the ga
ate is higheest close too the
center of
o the tunn
nel insulato
or. Thereforre, the effeect of the g
gate-voltagee on the tu
unnel
current controlling
g element.
The
e high barrrier also he
elps preven
nt thermion
nic currentt between the sourcee and
the dra
ain. Therm
mionic currrent is ca
aused by eelectrons, with suffficient therrmal
regulate
ed by the maximum barrier height.
h Hav
ving a narrrow, high-barrier att the
center allows
a the
e surround
ding low-ba
arrier insu
ulator (I1) to be ma
ade even loower
withoutt affecting
g the therrmionic cu
urrent. Th
his increasses the cu
urrent driiving
In next
n sectio
on, I give the perforrmance cha
aracteristiccs of the ttriple-insullator
D. Electrical characterization
MIM diode exists and to see if the gate has any effect on the diode current. The
measurements are made using the HP4145B parameter analyzer. The devices that
showed desirable behavior at DC were also tested with an AC gate voltage to ensure
that the gate-effect was due to the modulation of the tunnel current and not due to
slow moving interface charges. For the horizontal MIMFET, the results reported are
for the triple-insulator device with the largest gate effect. All measurements are
While describing the fabrication of the vertical MIMFET, I have alluded to the
structural impossibility of a gate effect due to the undercut from the RIE. Despite
this, a change in the drain-source current occurred on varying the gate voltage as
As seen in the figure, the IDS vs. VG curves are sensitive to the direction in
which the gate-voltage is scanned. Moreover, the current increases significantly
179
Figurre A-20 Drrain-source current v vs. gate-volttage for th
he
verticcal-gate MIMMFET. Alth ge
hough theree is consideerable chang
in currrent with VG, it is not due to the m
modulation of the tunn
nel
curren nt. Courtesy
y: Limin Caao
To develop
d a better
b unde
erstanding for the dirrectional dependencee, I tested tthese
0.1Hz. For
F this fre
equency ra
ange, the ch
haracteristtics resemb
bled those seen in Fiigure
VG. How
wever, as the freque
ency was cranked
c up
p from 0.1
1 to 1 Hz, the maxim
mum
possible
e explanatiion for thiss is a surfa
ace conducttion layer a
along the v
vertical edg
ge of
the MIM
M stack th
hat starts conducting
c g above a ccertain gatte voltage. This causses a
in curre
ent only forr a slowly increasing gate-voltag
ge.
2. DC character
ristics of horizonta
h al-gate MIMFET
In Figure
F A-31
1, I show th
he variatio
on of IDS wiith VDS for step chang
ge in VG froom 0
to 2 V. The
T curve for
f VG = 0 V is identiccal to the d
diode I(V) ccurve for th
he MIM tu
unnel
diode. Even
E for th
he VG 0 cu
urves the characteris
c stics resem
mble those of a diode with
180
Figurre A-31 Dra
ain-source current
c vs. voltage currves for fixeed
gate-v
voltages.
Nex
xt, I varied the VG wh
hile keeping the VDS ffixed. The result is sh
hown in Fiigure
VG on IDS
D . At negative gate voltages, the
t changee in curren
nt appears to saturatte as
tunnel distance
d ass the gate voltage be
ecome mor e negativee. On the oother hand, the
total tun
nnel distan
nce at posittive gate vo
oltages deccreases as VG increasses, thereby
y the
percenta
age change
e in the tunnel distan
nce increasses. This leeads to a ssharper cha
ange
in tunne
el current for
f VG>0.
181
Figurre A-42 IDS vs. VG mea asured at a constant VDS. The ga ate
currennt is orderss of magnitu
ude smallerr than the d
drain curren
nt,
imply
ying low gate leakage.
to the previous
p devices,
d the
e tunnel diode
d in th
he MIMFE
ET has a h
higher currrent
compare
e the charracteristicss shown in
n Figure A
A-42 with
h the devicce reported
d by
Fukush
hima et.al. (Fukushim
( ma, 1999).
182
The
e transcond
ductance, defined
d as gm = I D / VG , for the MIMFET is larger than
Fukush
himas nea
ar VG = 0 V. The
e saturatiion in tra
ansconducttance seen
n in
Fukush
himas dev
vice is incconsistent with the characterristics obsserved for the
MIMFE
ET. Later, I show th
hat the beh
havior of the MIMF
FET is closer to wha
at is
expected
d from mod
deling.
3. AC character
c ristics of horizonta
h al-gate MIM
MFET
To confirm
c th
he gate effe
ect observe
ed at DC, I test the transistorr with a la
arge-
correspo
onding cha ved. Since tthe respon
ange in IDSS is observ nse time foor slow mooving
effect sh
hould disap
ppear at a high enou
ugh frequen
ncy. Due too the smalll magnitud
de of
the tun
nnel curren
nt and the large cap
pacitive currrent that might oveershadow it at
high fre
equencies, I use a lock
k-in ampliffier setup a
as shown in
n Figure A
A-64.
By measuring
m g the current flowing in Rsense fo
for VDS = 0 and 0.8 V
V, I extractt out
the mod
dulation in
n the tunn
nel current caused by
y the gate--voltage. In
n Figure A
A-75,
frequen
ncy. At VDSS = 0 V, the
e rD vector is zero, an
nd only the current tthrough CGD is
sensed. At a finite
e VDS, the total
t curren
nt represen
nted by Rseense is senseed. The currrent
183
correspo
onding to the
t modula
ation of th
he tunnel iinsulator iss extracted
d by the veector
differen
nce of CGD and
a Rsense.
As shown
s in Figure
F A-86
6, this sign
nal is consttant with fr
frequency. The gate-eeffect
in the MIMFET
M iss confirmed
d to exist at
a non-zeroo frequenciies and rem
mains consstant
up to th
he range off the lock-in
n amplifierr used. Thee roll-off neear 100 kH
Hz is due too the
gain-ban
ndwidth lim
mitation off the lock-in.
Figurre A-86 Am
mplitude of AC drain- source currrent (iDS) v
vs.
frequeency.
184
E. Device modeling
m
1. Sim
mulator me
ethodolog
gy and initial resultts
project the
t perform
mance of im
mproved MIMFET
M sttructures. T
The simula
ation is div
vided
into two
o parts. A 2D electro
ostatic pote
ential calc ulation is required tto estimatee the
effect of
o the gatte voltage on the potential
p in the tu
unnel insu
ulator. Thiis is
implemented usin
ng the PD
DE solver in
i MATLA
AB. Next, the quanttum tunneeling
In Figure
F A-9
97(a), I sh
how a sim
mplified MIIMFET strructure th
hat is used
d for
simulation.
Thiss structure
e is based on an MIM
M diode th
hat is 7 nm
m wide and
d has a barrier
height of
o 0.4 eV. Here
H the 7 nm width
h is expecteed from th
he processing. The 0.4 eV
185
experimental device at VG = 0 V. As shown in Figure A-97(b), the shape of the
simulated curve is similar to the experimental one. I also show one of the previously
MIMFET are quite different from those simulated earlier (Shaker, 2003). Contrary
to the previous report, the MIMFET does not show any saturation of IDS at large VG.
the basic structure of the MIMFET. In Figure A-108, I compare three structures
described below:
In the IDS-VG curve shown in Figure A-108, the structural improvements lead to
186
Figurre A-108 Simmulation of improved M MIMFET strructures. Th he
I(V) curves ind dicate thatt a signifficant imprrovement in
transcconductance e and drivee-current iss possible, w
with changges
in the existing structure. Starting w with the orriginal deviice
simullated in thee previous sub-section, I modify it by removin ng
the excess insulator betwee en the gatee and the d drain. In th
he
second improvement, the dielectric constant of the ga ate
insula
ator is incre
eased.
3. Freq
quency off operatio
on
The
e reason forr pursuing a transisto
or compatib
ble with m
metal-insula
ator electroonics
is twofo
old. First, to satisfy th
he need forr a transisttor that ca
an be fabriccated using
g the
same te
echnology that
t is emp
ployed for making
m oth
her metal-iinsulator d
devices. Seccond,
the MIM
MFET is a contender for a high
h frequency
y thin-film transistorr (Reuss, 20
005).
MIMFE
ET. I model the AC characteris
c stics of thee MIMFET
T using thee circuit m
model
shown in
i Figure A-19.
A
187
Rs Rsense Rd
S D1 D
RDS
1k 1 1k
1.58Meg
Csd 4.7p
gmVgs
Rgs Cps Cgs Cgd Cpd Rgd
+
0.038u
-
10G 1.5p .1p 0.1p 1.4p 10G
G
RG
1k
GS
VGS
1Vac
0Vdc
parallel. The parameters for these components are calculated from the measured
determine the unity current-gain frequency (fc). At fc the modulated current in the
drain-source diode becomes equal to the capacitive current flowing through the gate
electrode. In Figure A-110, I show the results from three SPICE simulations with
i. The first curve, with the lowest falloff frequency corresponds to the
experimental result shown in Figure A-53. Apart from the
transconductance (gm = 1 nS/m), a gate-source capacitance (CGS)
value of 33 fF/m is measured.
ii. The second curve corresponds to the simulated device shown in
Figure A-108 that has a modified structure and a high-K gate
oxide. This device has an improved transconductance of 0.63
S/m. The CGS in this simulation is reduced to 2.2 fF/m,
assuming that an extremely small gate width of 15 nm can be
patterned using nanolithography.
188
iii. In the third simulation, I assume a transconductance of 1 mS/m
along with the reduced capacitance. The increased
transconductance is achievable by designing a resonant tunnel
diode as described in section C.3 or by applying a larger drain-
source voltage. However, at large VDS, the breakdown of the tunnel
insulator may be problematic.
1
1 2 3
Igate modulation/Itotal
gm= 1 mS/m
& reduced cap.
Experimental
Modified structure
0.1 & g=28
& reduced cap.
1k 1M 1G 1T
Frequency (Hz)
189
as the gate voltage is increased. The transconductance of this device, at zero gate
voltage, is higher than previously reported. The device has also been demonstrated
to work when an AC signal is applied on the gate. This verifies that the gate is
controlling the tunnel barrier, and confirms the absence of slow moving charges in
The fabrication method for the horizontal MIMFET is improved compared to the
previous fabrication techniques as it allows significantly smaller tunnel-insulator
widths. This allows the MIM diode to have high-current density at low voltages. The
nanoscale junction in the MIM diode is made without using nanolithography. The
key to this process is the deposition of the tunnel-insulator on the sidewall of the
metal, where the width of the insulator is governed by the thickness of deposited
characteristics across the wafer. Due to the limited size of the sputtering source, the
thickness and the angle of deposition vary at different locations on the wafer.
thickness of the tunnel insulator, which may occur due to the insulator deposition
step or the reactive ion etching. The variation makes it extremely difficult to
compare devices on different wafers. It also led to a null result for a number of
which makes it difficult to align an extremely small gate with the sub-10 nm tunnel
diode. Thus, a self-aligned approach is the key to have a narrow gate, which will
190
make the parasitic capacitance low. Finally, low barrier-height tunnel diodes need
Apart from verifying experimental results, the device simulations have served
as a tool to evaluate improved designs for the MIMFET. Results for some of these
a thin-film transistor for analog signal processing. For digital applications, power
dissipation may be a limiting factor as the Ion/Ioff ratio is only around 104 as seen in
operate at high frequency. The cutoff for the experimental device is low. However,
References
Fujimaru, K., Sasajima, R. & Matsumura, H., 1999. Nanoscale metal transistor
control of Fowler-Nordheim tunneling currents through 16 nm insulating channel.
J. Appl. Phys., 85(9), pp.6912-16.
Fukushima, K., Ryouta, S., Kouji, F. & Hideki, M., 1999. A Novel Nanoscale Metal
Transistor Fabricated by Conventional Photolithography. Jpn. J. Appl. Phys., 38,
pp.7233-36.
Jang, M., Kang, K., Lee, S. & Park, K., 2003. Simulation of Schottky barrier tunnel
transistor using simple boundary condition. Appl. Phys. Lett., 82(16), pp.2718-20.
Reuss, R.H., Chalamala, B.R., Moussessian, A., Kane, M.G., Kumar, A. et al., 2005.
Macroelectronics: Perspectives on Technology and Applications. Proc. IEEE, 93(7),
pp.1239-56.
191
Shaker, A. & Zekry, A.H., 2003. Theoretical Investigation of Single- and Dual- Gate
Metal Insulator Tunnel Transistors. In ICM 2003., 2003.
Snow, E.S., Campbell, P.M., Rendell, R.W., Buot, F.A., Park, D. et al., 1998. A
metal/oxide tunneling transistor. Semicond. Sci. Technol., 13, pp.A75-78.
Snow, E.S., Campbell, P.M., Rendell, R.W., Buot, F.A., Park, D. et al., 1998. A
metal/oxide tunneling transistor. Appl. Phys. Lett., 72(23), pp.3071-73.
192