|
model of a solar cell has been modified to include an additional rectification diode to include the
effect of exciton recombination [60].
While work exists in development of models to describe organic photovoltaics, little work
has been done in applying existing semiconductor modeling programs to simulate these cells.
Recently, Takshi et. al. applied Medici to simulate an organic transistor using P3HT as the
semiconductor [61]. Medici is a 2-dimensional device simulation program developed by Avant!
Corporation. It is designed for the simulation of MOS and bipolar transistors, and models
potential and carrier concentrations in a device by solving Poisson's equation and the electron
and hole continuity equations.
Targeted Research
The work presented in this dissertation provides a study of hybrid photovoltaic devices
from an experimental and theoretical perspective. The device simulation program Medici is used
for the first time to provide simulations of an ordered bulk heterojunction photovoltaic device
with an array of inorganic nanorods interspersed with a semiconducting organic polymer. This
design was chosen due to its semi-regular structure that allows the device to be broken into a
representative unit cell, providing greater detail in the simulations. By modeling the cell in this
way, effective values of key parameters such as the carrier mobility, exciton diffusion length,
and energy gap of the interface can be estimated for the device during operation.
This simulation supplements experimental work focusing on process development for a
polymer-nanocrystal bulk heterojunction solar cell. The issues of charge transport and exciton
dissociation are targeted. Charge transport is improved through surface treatment of the ITO
electrode prior to active layer deposition. This generates a smoother surface and promotes
adhesion of subsequent layers. Exciton dissociation is addressed through control of the
morphology of the bulk heterojunction active layer. This is achieved through surface exchange
0.06
(E Hybrid Cell
5 0.05 P3HT Cell
E Decay- Hybrid
4 Decay- P3HT
0.04 \ -------
S\ Hybrid Cell
c 0.03 J= 2.894E-3 + 1.525 exp(-9.369E-2* t)
S\ \ P3HT Cell
0.02 \ J = 5.804E-3 + 7.506E-2 exp(-4.820E-2* t)
0
o 0.01
U) t -----f-I
S-- -- -- -- -
0.00 ---
0 50 100 150 200 250 300
Exposure Time (min)
Figure 3-29. Short-circuit current decay for hybrid (gold circles) and P3HT (green squares).
-0.2 -0.1 0.0 0.1 0.2 0.3
Voltage (V)
1000
100
0.001 ----- 8-3 Illu
-- 8-4 Illu
0.0001
-1.0 -0.5 0.0 0.5
Voltage (V)
Figure 3-8. Linear (A) and base 10 log-scale (B) J-V curves for bi-layer organic solar cells
fabricated with a single 80 nm thick layer of PEDOT:PSS.
CHAPTER 2
MEDICI SIMULATIONS OF HYBRID SOLAR CELLS
Introduction
Device modeling has lagged behind cell fabrication for organic and hybrid photovoltaics.
Cell performance has been characterized using equivalent circuit theories [60, 62], and some
work has been performed to model certain parameters such as cell lifetimes [63], charge
recombination [64], and short-circuit current [59]. The use of existing simulation programs to
provide a fully-encompassed view of the device performance has not been attempted to this
point.
Organic photoabsorbing materials differ fundamentally from most inorganic crystals in
that the absorption of a photon generates an exciton, or bound electron-hole pair, with high
binding energy. Excitons can be generated in inorganic materials as well, but this occurs only
over a very limited wavelength and temperature range, and the resulting excitons exist with a low
binding energy on the order ofkT at room temperature [57]. This low binding energy causes the
exciton to be highly unstable in the presence of elevated temperatures or strong fields, such as in
the space-charge region of a p-n junction. In organic materials, excitons are generated nearly
exclusively and exist with a binding energy or approximately 10x that of the inorganic excitons,
or 300 meV. These highly stable excitons dissociate at an interface with another material,
where one of the carriers is transferred into the neighboring material. In organic photovoltaic
cells, the electron is typically transferred into an n-type material due to the relatively low
electron mobility in most organic materials.
Because the organic exciton exists as an essentially uncharged particle unaffected by
fields, it travels through the solar cell through diffusion, with the exciton diffusion length (LD)
representing the maximum distance it can travel before self-annihilation. For many organic
tsf496_75 P HT Absorption, Affinity
15 tsf496_77 P3HT Absorption, Affinity, Permittivity
tsf496_78 P3HT Absorption, Affinity, Density of States
tsf496_79 P3HT Absorption, Affinity, Mobilities
10 -
5 -
0 -
-5
-10
-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Voltage (V)
Figure 2-41. Simulated ZnO:CIS
electron affinity.
solar cells with P3HT values for absorption coefficient and
Table 2-15. Performance measures for simulated ZnO:CIS solar cells with P3HT values for
absorption coefficient and electron affinity.
Simulation Properties Varied Voc (V) Jsc (mA/cm2) FF i (%)
Absorption
75 Absorption 0.467 3.980 0.493 0.916
Affinity
Absorption
77 Affinity 0.455 3.979 0.502 0.909
Permittivity
Absorption
78 Affinity 0.471 3.975 0.479 0.896
Density of States
Absorption
79 Affinity 0.613 3.439 0.199 0.419
Mobility
For TOPO-coated nanocrystals, the surface roughness of the films decreased as the pyridine
content increased, up to 50%. This is due to the low solubility of TOPO-coated nanocrystals in
chloroform. Pyridine concentrations of 45-50% were required to generate films with less than 10
nm rms roughness. After surface exchange with pyridine is performed, the pyridine-terminated
nanocrystals become significantly more soluble in chloroform. In this case, pyridine
concentrations of less than 10% yield rms surface roughness values of less than 10 nm.
Hybrid Films
To fabricate bulk heterojunction solar cells with organic polymers and inorganic
nanocrystals, great care must be taken to ensure proper mixing between the two phases. The
exciton diffusion length of most semiconducting polymers is in the range 5 to 20 nm. Because of
this limitation, the organic and inorganic phases in the hybrid active layer must be well-mixed so
that excitons generated in the organic phase can reach an inorganic phase that is within one
diffusion length.
In this investigation, the properties of hybrid films is studied through optical microscopy,
electron microscopy, atomic force microscopy, and surface profilometry. Solutions are
generated by dissolving blends of P3HT polymer and CdSe nanopowder into solvent mixtures of
chloroform and pyridine. Hybrid films are spin-cast from these solutions onto ITO-coated glass
substrates that were subjected to N2 plasma treatment as described previously, and the films were
dried under vacuum.
TOPO-coated CdSe
Initially, hybrid solutions were created with P3HT and TOPO-coated CdSe nanocrystals
with a radius of approximately 5 nm. The solvent used in these solutions was a 50-50 mixture of
chloroform and pyridine, based on the results shown in Figure 3-19. Three hybrid solutions were
prepared with composition shown in Table 3-7.
59. W. Geens, T. Martens, J. Poortmans, T. Aernouts, J. Manca, L. Lutsen, P. Heremans, S.
Borghs, R. Mertens, D. Vanderzande, Thin Solid Films 451 (2004) 498.
60. B. Mazhari, S. En. Mater. & Solar Cells 90 (2006) 1021.
61. A. Takshi, A. Dimopoulos, J.D. Madden, Solid-State Electronics 52 (2008) 107.
62. W.U. Huynh, J.J. Dittmer, N. Teclemariam, D.J. Million, A.P. Alvisatos, K.W.J.
Barnham, Phys. Rev., B 67 (2003) 115326.
63. R.D. Bettignies, J. Leroy, M. Firon, C. Sentein, Syn. Met. 156 (2006) 510.
64. V. Lemaur, M. Steel, D. Beljonne, J.L. Bredas, J. Cornil, J. Am. Chem. Soc. 127 (2005)
6077.
65. Medici User's Manual, Avant Corporation, Beaverton, OR, Release 2001.4, (2000).
66. W.K. Kim, Ph.D. Dissertation, University of Florida, Gainesville, FL, (2006).
67. E.J. Meijer, C. Detcheverry, P.J. Baesjou, E.V. Veenendaal, D.M. de Leeuw, T.M.
Klapwijk, J. Appl. Phys. 93 (2003) 4831.
68. H. Sirringhaus, N. Tessler, R.H. Friend, Science 280 (1998) 1741.
69. C. Tanase, E.J. Meijer, P.W.M. Blom, D.M. de Leeuw, Org. Elec. 4 (2003) 33.
70. T.A. Chen, X. Wu, R.D. Reike, J. Am. Chem. Soc. 117 (1995) 233.
71. A.J. Casico, J.E. Lyon, M.M. Beerbom, R. Schlaf, Y. Zhu, S.A. Jenekhe, App. Phys. Lett.
88 (2006) 62103.
72. A.J.C. Wen, K.L. Chen, M.H. Yang, W.T. Hsiao, L.G. Chao, M.S. Leu, Surf. Coat. Tech.
198 (2005) 362.
73. Melles Griot Inc., http://mellesgriot.com/products/optics/oc_5_1.htm (2002) July 2008.
74. Kim, Y., S.A. Choulis, J. Nelson, D.D.C. Bradley, S. Cook, and J.R. Durrant, J. Mat. Sci.
40 (2005) 1371.
75. D.C. Olson, J. Piris, R.T. Collins, S.E. Shaheen, D.S. Ginley, Thin Solid Films 496
(2006) 26.
76. GetData Graph Digitizer, http://getdata-graph-digitizer.com (2008) July 2008.
77. K.R. Amundson, B.J. Sapjeta, A.J. Lovinger, Z. Bao, Thin Solid Films 414 (2002) 143.
78. S. Li, "Semiconductor Physical Electronics," 2nd Ed., Springer, 2006.
Figure 2-58. Photogenerated carriers at the tip (left axis) and base (right axis) corner points of
the nanorod in simulated hybrid cells.
-3 -
-0.1
0.0 0.1 0.2 0.3 0.4
Voltage (V)
Figure 2-59. J-V curves for simulated solar cells using line-source carrier generation.
1.60E+24 3.249E+22
C. 'a
'F
a 1.40E+24 -- --NanorodTip 3.248E+22 0
o C
0 1.20E+24 -- -Nanorod Base 3.247E+22 z
" 1.00E+24 3.246E+22 E
E 8.00E+23 3.245E+22
--
m 6.00E+23 3.244E+22
1 4.00E+23 3.243E+22 C
c 2.00E+23 3.242E+22
O.OOE+00 3.241E+22
. 10nm 20nm 40nm Full
Exciton Diffusion Length
Future Work
Although the field of organic photovoltaics is rapidly growing and advancing, several
research directions are suggested by this work for its continuation. This section describes some
potential research directions stemming from the work presented in this dissertation.
Organic Photovoltaic Simulations
The models presented in this dissertation used the powerful modeling software package
Medici. Although the software has powerful optical and electrical simulation abilities, this
research seems to have pushed its limits by demanding nano-scale material specifications and
low levels of free carriers, carrier mobilities, and current flow in the devices. Simulation
attempts were frequently cut short due to convergence issues in the software under the specified
conditions. Additionally, a key component of the modeling work focused on the correct way to
simulate the effect of excitons in Medici, which are not explicit in the package.
In light of these difficulties, a new software package designed to simulate organic
electronic materials would be a great boost to this work. A product such as Fluxim [94] would
be able to more accurately simulate the effect of excitons in these hybrid devices.
Another direction that should be pursued is the variation of the geometry and materials of
the simulated cells. While ZnO is a well-researched material for the growth of aligned and
ordered nanowires, these structures are now being grown for other materials with stronger
absorption spectra such as CdSe and InP [66, 95]. These materials could see more use in
nanowire hybrid cells in the future, driving the need for effective simulations to study device
properties.
The most common organic cell design is currently the bulk heterojunction cell using
semiconductor nanoparticles or soluble C60 derivatives. This design presents a challenge for
simulations because little work has been done to characterize the particle distributions in these
CHAPTER 1
INTRODUCTION
Introduction
Since the discovery of the photovoltaic effect and the design of the first functional solar
cell, photovoltaic technology has consistently developed to become an increasingly viable energy
source. Photovoltaics has developed into many classes of devices and materials systems.
Photovoltaic technology can provide clean, efficient, and portable energy.
Motivation
There is currently a strong push toward alternative energy sources as the price of oil
increases and nations worldwide work to slow the emission of greenhouse gases such as COx and
NOx. Many countries have established conservation and alternative energy programs in attempts
to control the output of these gases. Recent increases in oil and gas prices and controversy
surrounding global warming have driven public recognition of the need for alternative renewable
energy sources. While small-scale steps such as hybrid cars stand to relieve a small amount of
the world's fossil fuel consumption, new technologies such as photovoltaic energy must be
developed to fulfill the world's large-scale energy needs.
Of the available candidates for alternative large-scale energy production, photovoltaic
energy conversion has many qualities that make it a leading technology, including:
Photovoltaic technology has been studied intensely for many years, initially driven by its
use in the space program to provide energy for satellites and space vehicles, to prepare it
for commercialization.
Energy generated from solar cells can be generated locally, resulting in reduced energy
distribution costs and a more reactive system.
Because of its local power generation, photovoltaics are optimal for power generation in
remote locations where it is difficult or impossible to connect to a local grid.
including surface roughness, chemical stability, and reduced work function. Electron beam and
N2 plasma treatment replace oxygen with nitrogen, reducing the electron affinity of the ITO film
and resulting in a lowering of the work function that improves the charge collection efficiency.
The effect of several ITO surface treatments on the performance of organic solar cells was
determined. It was shown that exposure to a N2 plasma was more beneficial than either the
oxygen plasma or e-beam treatment. The N2 plasma was most successful in improving the
surface characteristics, as evidenced by the extent of lowering the contact angle and decreasing
the surface roughness (AFM). This treatment incorporates nitrogen into the near surface region
and produces a slight change in the In/Sn ratio, which reduces the ITO work function. These
changes optimize the energy band diagram and improve charge collection at the ITO anode.
Bi-layer Organic Solar Cell Fabrication
A process for fabricating bi-layer organic solar cells with the cell structure
ITO/PEDOT:PSS/P3HT/C60/Al was developed. Figure 3-2 shows the energy band diagram for a
cell with this structure [45, 86]. This material system has received much attention for
applications in bulk heterojunction solar cells due after the discovery of ultrafast charge transfer
at interfaces between conjugated polymers and C60 molecules.
In these bi-layer organic cells, P3HT serves as the primary absorber layer in the cells, with
a bandgap of approximately 1.7 eV and a very strong absorption coefficient. Excitons generated
in the polymer are separated at the C60 interface, with electrons dropping to the lower energy
level of the C60 and holes returning to the P3HT. Electrons are transported to the backside
aluminum contact, while holes are transported through the P3HT layer and the hole transport
layer (HTL) of PEDOT:PSS to the transparent ITO frontside contact.
-0.8 -
-0.05
0.00 u
E -0.6 0
o 0.05 .
E -0.4 0.10 O
0.15
c -0.2 020
a, 0 ,5,5 .0 .0 I .0 6 .,io
0.0
E0 H-brid Ciark
0.2 O I H bnri Illuin ated
-B- P. HT Dark
0.4 I F'-.HT IIIiminn ated
1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5
Voltage (V)
Figure 3-28. Dark and illuminated J-V curves for hybrid bulk heterojunction solar cell deposited
from 25 mg/ml solution and P3HT polymer cell deposited from 12 mg/ml solution.
The insert shows a zoom-in on the active area of the cells.
illumination. Power conversion efficiency (rI) was calculated using Equation 3-3, with Vmax,
Imax, and Jmax representing the voltage, current, and current density at the maximum power point.
The fill factor (FF) is calculated using Equation 3-4 with Voc and Jsc being the open circuit
voltage and short circuit current density.
P,, J1mW/cm2
FF = Vmaxmax (3-4)
V J
OC SC
The rectification ratio (RR) of a diode is the ratio of forward to reverse current at some
applied bias. Higher rectification ratios show stronger diode characteristics in the I-V curve for
devices. Rectification ratios displayed in this section were calculated from dark current
measurements at 0.5 V of forward and reverse bias unless otherwise noted.
Bi-layer solar cells were fabricated using the procedure detailed previously, but with no
treatment performed on the ITO electrode. J-V curves for these cells are shown in Figure 3-12.
The cells fabricated in this set all showed a measurable photocurrent, demonstrating
functioning solar cell behavior. The best-performing cell in the set, Set I 3, showed a power
conversion efficiency of 0.04%. Despite having the lowest Voc of all cells in the set, this
champion cell showed a short-circuit current density of 1.55 mA/cm2, which was significantly
higher than any other cell in the set. Performance was low, but measurable, in all cells. With the
exception of cell I 3 with a Voc of 0.11 V, all cells showed a Voc of almost exactly 0.15 V.
The fill factor for the cells ranged from 0.15 for cell I 4 to 0.24 for cell I-1. Cells I-1 and I-4
showed a Jsc of approximately 0.5 mA/cm2. The Jsc for cell I-2 was 1.03 mA/cm2.
Another set of bi-layer cells were fabricated with the same cell structure, but applying
plasma treatment to the ITO substrate. The performance of these cells was poor, with the lack of
-10 -
-0.1
0.0 0.1 0.2 0.3 0.4
Voltage (V)
Figure 2-43. Simulated J-V curves for ZnO:P3HT solar cells with varying carrier mobility. In all
cases, carrier generation is allowed in the full P3HT region and the electron mobility
is set to 10% of the hole mobility.
Figure 2-44. Calculation method for estimated second derivatives of J-V curves.
the absorption coefficient in each region, rounded to the nearest 10% value. The input data used
for the simulations are shown in Table 2-11, as well as the resulting short-circuit current density
for that simulation. The J-V curves for these simulations are shown in Figure 2-37.
Table 2-11 shows that absorption in the cell is distributed over a wider range as the
standard deviation increases. The average multiplier to the absorption coefficient is calculated
for each situation and as targeted it is equal to 50% or 51% for all cases, with any error coming
from rounding to the nearest 10% in each region. The J-V curves in Figure 2-36 show that this
grading has virtually no effect on the final performance of the cells. The short-circuit current
density shows extremely minor variations in the curves, and this value is tabulated in Table 2-11.
The short-circuit current density increases slightly as the absorption distributions become wider.
This shows that extending the generated carrier distribution toward the back electrode improves
device performance, while limiting the carrier distribution to a narrower region, even with the
same number of carriers being generated, hinders performance.
P3HT replacement with CIS
The J-V curves generated in previous simulations shows a double-elbow shape that is not
characteristic of photovoltaic cells. To determine the root cause of this phenomenon, P3HT was
replaced with copper indium diselenide (CIS), a popular p-type thin-film photovoltaic material.
The same cell structure is used, with ZnO nanorods as the n-type material. Within this
framework, material properties of the CIS layer were adjusted to values corresponding to P3HT.
Preliminary simulations in this form were performed, and the resulting J-V curves are shown in
Figure 2-38. This set of simulations compares ZnO:P3HT cells with 10 nm and full cell
absorption regions to ZnO:CIS cells with full cell absorption, a 10 nm absorption range, and a 10
nm absorption range with an energy band gap set at 0.85 eV.
6e+21 --
5e+21-
4l+21 -
3e+21
2e+21
le+21
0
-le+21
Ar
I/ id
I2
LJ I ESi
Y coordinate (i' 0
Y Coordinate (,m)
0.00
0.01 0.02 0.03 0.04 0.05 0.06
X Coordinate (inm)
Figure 2-57. Contour plots of the photogenerated carrier difference between the full absorption
simulation and the 40 nm LD simulation.
2-19. J-V curves for simulated cells with zero absorption in the P3HT2 region and varying
reflectance from the A g electrode ......... ................. ................................ ............... 89
2-20. Simulation results showing the effect of changing charge mobilities in the P3HT
region s ............... ......... .......... ......... ........ ................................. 90
2-21. J-V curves for simulated cells with varying exciton diffusion length.............................91
2-22. Extrapolations to estimate Voc for simulated cells with varying exciton diffusion
le n g th .............. .... ........ ..... ............................................... ................ 9 1
2-23. Solar cell performance measures for simulated cells with varying exciton diffusion
lengths..................... .......................................92
2-24. Real and simulated J-V curves for cells with varying P3HT doping density. ...................93
2-25. Real and simulated J-V curves for hybrid cells with varying ZnO doping density.............93
2-26. Real and simulated J-V curves for hybrid solar cells with varying P3HT density of
state s................. .. .......... .................. ................................................ 9 4
2-27. Simulated J-V curves for hybrid solar cells with varying P3HT mobility........................94
2-28. Simulated J-V curves for hybrid cells with varying P3HT doping concentrations.............95
2-29. Energy band diagram for P3HT ZnO hybrid solar cells. .............................................95
2-30. J-V curves for simulated cells with varying energy band gap in the active layers..............96
2-31. Absorption coefficient vs. wavelength as tabulated in Medici....................... ......... 96
2-32. AM 1.5 solar spectrum. ..................................... .. .. ......... .. ............97
2-33. Carrier generation in simulated solar cells plotted with absorption coefficients for
P 3H T and Z nO ............................................................................97
2-34. J-V curves for simulated cells showing the original P3HT absorption profile and an
edited absorption profile limiting absorption between 0.2 and 0.3 m ............................. 98
2-35. J-V curves for simulated cells with exciton diffusion length of 10 nm and multi-stage
ab so rp tio n reg io n s..............................................................................................................9 8
2-36. Examples of cumulative distribution function with mean of 10 nm and a range of
standard deviation ....................................................... ................. 99
2-37. Simulated J-V curves for cells with graded absorption profiles.............. ... ................100
2-38. Simulated J-V curves with CIS replacing P3HT..................................... ...... ............... 100
The electron affinity was fixed at its P3HT value of 3.15 eV, setting the conduction band
level in the simulations to the appropriate value for P3HT. The electronic band gap of the
simulations is still set at the CIS value of 1.04 eV rather than the P3HT value of 1.7 eV or the
effective band gap of 0.85 eV at the ZnO-P3HT junction. Again, the absorption coefficient is set
at the P3HT value and absorption is allowed over a 10 nm range near the material interface.
Variations in the other materials properties resulted in the J-V curves shown in Figure 2-41 and
detailed in Table 2-15. As seen in previous simulations, there is virtually no change in the J-V
curve or device properties with the application of the P3HT values for permittivity and density of
states. The application of P3HT values for carrier mobility, however, resulted in a drastic shift in
the nature of the curve. This change resulted in a 0.5 mA/cm2 reduction in the short-circuit
current density and an increase of -0.6 V in the open-circuit voltage. Additionally, the J-V curve
takes on the double-curve shape, which drops the fill factor to approximately 0.20. This shape
will be discussed in more detail after the next set of data.
As discussed previously, the J-V curves shown in Figure 2-41 resulted from simulations
where the electron affinity was set at the P3HT value of 3.15 eV, but the electronic band gap
remained at the CIS value of 1.04 eV. Although fixing the electron affinity sets the LUMO level
of P3HT, the band gap value sets an inappropriate HOMO level and results in the generation of
carriers with higher energy than appropriate. The simulations of Figure 2-41 were modified to
include the appropriate ZnO-P3HT interfacial band gap of 0.85 eV, corresponding to the energy
gap between the conduction band of ZnO and the HOMO level of P3HT. The results, shown in
Figure 2-42, mimic those in Figure 2-41.
The curves resulting from adjustments in the permittivity and density of states show the
expected shape for a solar cell J-V curve despite having a significantly lower Voc than their
new area being included in each summation contains relatively few carriers due to lack of
absorption in these regions.
The total amount of photogeneration in the simulated cell is dependent on the exciton
diffusion length specified in the simulation. This parameter defines the size of the strongly-
absorbing area in the P3HT polymer. Figure 2-55 displays the total count of photogenerated
carriers for simulated cells with varying exciton diffusion lengths defined from 10 nm up to the
full polymer region of the cell.
In addition to dictating the total number of carriers in the cell, variations in the exciton
diffusion length also change the distribution of these carriers. Figure 2-56 displays the line
sources used in Medici to simulate photoabsorption based on the model described previously.
Note that the curves seen in Figure 2-56 are not the actual summations of carriers calculated
from the contour plots, they are the fit lines applied to Medici with the form shown in Equation
2-6. As evidence of the quality of the fit, Figure 2-56C displays the true carrier summations for
Line 2 of a cell with LD = 20 nm, along with the fit curve applied in Medici.
As described previously, photogeneration in Line 1, for 0 < x < 0.015 atm, is set as
constant in all cases. In Line 2, shown in Figue 2-56B, carrier generation increases nearly
linearly until a certain point where the exponential decay becomes the dominant effect. For Line
3, carrier generation again increases nearly linearly for x > 0.015 [tm before reaching a point
where it becomes approximately constant. Unlike Line 1, the nearly-constant region for Line 3
was not assumed to be exactly constant, and all parameters for the exponential-linear equation
were calculated. For the 10 nm and 20 nm cases, this region shows a lower slope than in the 40
nm and full cell cases.
The dotted line drawn through the data in Figure 3-35 is the best fit line for film thickness
vs. number of layers, which follows Equation 3-8. With this trend, to reach a hybrid film
thickness of 100 nm, 21 layers would be required. In the fabrication setup used for these
experiments where film deposition equipment is open to atmosphere, this would result in a
prohibitive amount of atmospheric exposure time for the P3HT. For deposition in a nitrogen or
argon environment like a glove box, however, this process is feasible.
FilmThickness = 4.903 (# Layers) (3-8)
The film surface roughness was measured with AFM, and the surface was visualized with
SEM and optical microscopes. Figure 3-36 displays images of the results of these measurements
at 1, 3, 5, and 7 layers. The films analyzed by SEM and AFM were different from the films
analyzed with profilometry due to a difference in the sample size required for these techniques.
All SEM images are shown at the same scale (at 30,000x magnification), and all AFM images
show a 3 x 3 [tm scan area with a 100 nm scale on the z-axis.
Optical microscope images of the films are shown in Figure 3-37. These images are taken
at 100x magnification and are representative of the multiple images taken of these surfaces.
The rms surface roughness was measured from the AFM images for 3 x 3 [tm and 1 x 1 tm
areas on the surface. The results are shown in Figure 3-38. Additionally, lines are fit to
determine the trends for roughness as more layers are deposited.
The trendlines follow Equation 3-9 for the large area measurement and Equation 3-10 for the
small area measurement. If these trendlines are extrapolated to the required 21 layers for a 100
nm active layer film, the rms surface roughness is predicted to be 39.9 nm for a 5 x 5 tm area
and 15.0 nm for a 1 x 1 [tm area. From previous measurements of surface roughness by
profilometry and AFM, it was found that profilometer roughness measurements were
A)
100
80
P3HT Film
C 60
Co
S40 ITO
.- 40 -
I--c t- L
20------------ - -
0
0.0 0.5 1.0 1.5 2.0 2.5
Lateral Distance (mm)
Figure 3-24. Surface profiles of P3HT films deposited from chloroform (A), 1:1
chloroform:pyridine (B), and pyridine (C) solvents. In the graphs, the bare substrate
appears on the left hand side with the film on the right hand side beyond the wide
peak. The solid teal line represents the median film thickness, the red dash-dot line
represents the mean film thickness, and the blue dashed lines represent + one standard
deviation from the mean.
* .I" .1 q ^ tfk
IL ...., -W
.:: "' .*
.4 .4
00 "o S"
:o. '. -,." -; > .'
*. :'
'- .u.& "" ." ""
'-" ,'-" ,. ,*'*,: ".'
.. ...
v: "
t.
*
S
4.
'.4'
a'
a
lb
9i
I.
S. **
'.,
Figure 3-17 continued. Optical microscope images of selected films.
10xI
A)
50
S40
30
0
20
Cn
S10 -
0
0 10 20 30 40 50 60
Pyridin e Percentage in Solvent (%)
B)
12
S10
,r-
-)
r 2 8
Co
6U 6
2
O 2 4 6 8 10 12
Pyridinre Percentage in Solvent(%J
Figure 3-19. Film surface roughness vs. pyridine concentration in the chloroform solvent for
TOPO-coated CdSe nanocrystals (A) and pyridine-coated CdSe nanocrystals (B).
175
Table 3-9. Hybrid solar cell fabrication information and performance data.
SSolution CdSe wt.% %Pyridine in (1) Vo() (2) Notes
Figure S C 0 0 Jsc Voc (V) FF 1 Notes
SConcentration in P3HT Chloroform
3-25 5 mg/ml 60% 2% 8.99 0.338 0.230 0.7
3-25 5 mg/ml 60% 0% 0.026 0.194 0.252 0.001 Pure Chloroform Solvent
3-26 10 mg/ml 60% 2% 33.9 0.256 0.275 2.4
3-28 25 mg/ml 50% 2% 49.9 0.701 0.232 17.5
3-28 12 mg/ml 0% 2% 20.3 0.523 0.310 12.5 Pure P3HT film
3-30 25 mg/ml 50% 2% 205 0.705 0.288 41.6 Minimized air exposure
Pure P3HT film
3-30 12 mg/ml 0% 2% 52.6 0.237 0.242 3.0 ureHTlm
Minimized air exposure
3-31 30 mg/ml 50% 2%(3) 138 0.391 0.292 15.8 Pyridine in Chlorobenzene Solvent
3-32 25 mg/ml 50% 2% 94.4 0.329 0.247 7.65 Commercial CdSe
(1) Jsc displayed in units of aA/cm2 (mA/cm2 x 10-3)
(2) Efficiency displayed in units of % x 10-3
(3) This solvent is 2% pyridine in chlorobenzene
from the P3HT region beyond the tip of the nanorod, as well as a small area within nanorod.
This area inside the nanorod decreases as the x-coordinate increases because portions of this
region begin mapping to Line 2. This decrease results in a negligible decrease in the number of
carriers, however, because the ZnO absorbs poorly in comparison to the active P3HT region
included in the calculation. This effect holds true for any exciton diffusion length that is
considered in this study, because the P3HT region will always absorb much more strongly than
the ZnO region. Due to this, the carrier distribution for Line 1 is approximated as a constant for
all simulations.
Along Line 2 (x = 0.015 am, 0.025 am < y < 0.26 am), photogeneration increases steadily
at low values of y because a larger section of the absorbing region is included at each point.
Seen in Figure 2-54, this occurs for approximately one LD, from 0.025 < y < 0.045 am. Beyond
that point, the new regions being added are not absorbing regions, so even though the size of the
area being summed is larger, this larger area is not contributing a significant number of carriers.
This is compounded by the fact that the LD region at this greater depth produces a lower number
of carriers due to a drop in the number of photons remaining in the cell. This exponential decay
of absorption becomes the dominant contribution to the number of carriers along the line, and
continues to the tip of the nanorod. The point where the distribution along Line 2 turns from
linear growth to exponential decay depends on the exciton diffusion length of the simulated cell.
For Line 3 (x > 0.015 a, y = 0.025 am), the number of photogenerated carriers initially
shows a linear increase with x for the same reasons as Line 2. As the line is traversed, additional
area is being mapped to this line, which increases the number of carriers contributed. After a
distance of LD along the line, the distribution becomes nearly constant. This occurs because the
4.0e-4
2.0e-4 -
0.0 -
-2.0e-4
-4.0e-4 -
-6.0e-4 -
-8.0e-4 -
-*- Dark
IlluImininaedj
-1.0e-3
_1 I
-02 -0.1 00 01 02 03 04 05
-1.5 -1.0 -0.5 0.0 0.5 1.0
Voltage (V)
Figure 3-25 continued. Dark and illuminated J-V curves for cells generated from 5 mg/ml
composite solution in pure chloroform.
-1.5 -1.0 -0.5 0.0 0.5 1.0
Voltage (V)
Figure 3-26. Dark and illuminated J-V curves for 10 mg/ml hybrid solution deposited with low-
speed spin-coating.
~----~"El
- I. -
films. Because the physical dimensions of the model are vital for accurate simulation, this
provides both a limitation and an opportunity. On the one hand, this lack of information limits
the results that can be generated through simulations. On the other hand, if an accurate, robust
model can be developed for a hybrid material system, it could be used to back-calculate
unknown physical dimensions of the system.
Development of Hybrid Photovoltaic Cells
A central theme of this dissertation is the morphology control of hybrid films for
photovoltaic applications. This study only focused on nearly spherical nanocrystals, but future
studies should expand the field to include nanoparticles with dimensionality. Shaped
nanoparticles, nanowires, tetrapods, and more exotic branched structures are being grown using
techniques similar to the one used for spherical nanocrystals in this study [35-37]. These
dimensional crystals offer the promise of directed charge transport without the multiple electron
"jumping" processes required for small spherical nanocrystals.
Alternative semiconductor nanoparticles should be considered as well. Although CdSe is
one of the easiest particles to be synthesized, other compound semiconductors such as CdS,
CdTe, PbSe, and CIS can be grown on the nano scale. Some semiconductors such as PbSe have
demonstrated multi-carrier generation on short timescales that offer the possibility of
constructing photovoltaic devices with quantum efficiencies greater than 1 if these charges can
be harvested [96-97].
Regioregular P3HT, as used in this study, is currently the most promising candidate for
polymer in electronic devices such as organic thin-film transistors and solar cells [33-34].
However, this polymer shows some limitations for solar applications. The absorption spectrum
shows a cut-off above 650 nm, limiting absorption of near-IR photons which are plentiful in the
solar spectrum. Although the hole mobility of P3HT is high compared to many conductive
counterparts in Figure 2-41. This variation is expected, due to the shift in energy band levels
causing the generated carriers to exist with more energy. The Jsc of the cells remains virtually
unchanged as compared to their counterparts in Figure 2-41, but the shift in Voc results in a
slight reduction of the fill factor as the curves cross the voltage axis at a slightly lower slope.
Curve 82, corresponding to a change in carrier mobility, also shows a reduction in Voc when
compared the curve shown in Figure 2-41. This curve shows a very small fill factor of
approximately 0.15 due to the double-elbow shape which eliminates most of the active area of
the curve.
The origin of the change in shape accompanying lower mobility values is unclear, but it
obviously occurs only when the absorption, electrical energy levels, and mobilities take on their
P3HT values. This shape was observed for all ZnO:P3HT cells simulated up to this point, but it
was not seen in other simulations using a ZnO:CIS cell as a basis. From the graphs in Figures 2-
39 2-41, it can be concluded can state that the permittivity and density of states have no effect
in causing this shape. Additionally, the application of P3HT levels of absorption and energy
levels to a ZnO:CIS cell did not cause this shape without the addition of P3HT levels for carrier
mobility. In the simulations, the carrier mobilities were set as [n = 0.001 cm2/V-s, ip = 0.01
cm2/V-s for P3HT, and Ln = 30 cm2/V-s, Lp = 300 cm2/V-s for CIS. Changing from CIS values to
P3HT values represents a severe drop of 4 orders of magnitude with no further details for
intermediate values.
Using simulations of a ZnO:P3HT cell, a wide range of mobility values were examined,
with the resulting J-V curves shown in Figure 2-43. Note that in all cases, Cp = 10*pn.
Additionally, these simulations allow absorption throughout the full P3HT region of the cell to
boost the level of current flow and more clearly show the effect of the mobility variations.
38. W.J.E. Beek, L.H. Slooff, M.M. Wienk, J.M. Kroon, R.A.J. Janssen, Adv. Funct. Mater.
15 (2005) 1703.
39. P.A.C. Quist, L.H. Slooff, H. Donker, J.M. Kroon, T.J. Savenije, L.D.A. Siebbeles,
Superlattices and Microstructures 38 (2005) 308.
40. A. Watt, H. Rubinsztein-Dunlop, P. Meredith, Mat. Lett. 59 (2005) 3033.
41. X. Peng, J. Wickham, A.P. Alivisatos, J. Am. Chem. Soc. 120 (1998) 5343.
42. Z.A. Peng, X. Peng, J. Am. Chem. Soc. 123 (2001) 183.
43. M. Kuno, J.K. Lee, B.O. Dabbousi, F.V. Mikulec, M.G. Bawendi, J. Chem. Phys. 106
(1997) 9869.
44. W.U. Huynh, J.J. Dittmer, W.C. Libby, G.L. Whiting, A.P. Alivisatos, Adv. Funct.
Mater. 13 (2003) 73.
45. H.J. Choi, J.K. Yang, H.H. Park, Thin Solid Films 494 (2006) 207.
46. K.M. Coakley, M.D. McGehee, App. Phys. Lett. 83 (2003) 3380.
47. K.M. Coakley, Y. Lui, M.D. McGeehee, K.M. Frindell, G.D. Stucky, Adv. Funct. Mater.
13 (2003) 301.
48. J.J. Wu, S.C. Liu, Adv. Mater., Vol. 14 (2002) 215.
49. W. Lee, M.C. Jeong, J.M. Myoung, Acta Mater. 52 (2004) 3949.
50. C. Geng, Y. Jiang, Y. Yao, X. Meng, J.A. Zapien, C.S. Lee, Y. Lifshitz, S.T. Lee, Adv.
Funct. Mater. 14 (2004) 589.
51. S.C. Lyu, Y. Zhang, C.J. Lee, H. Ruh, H.J. Lee, Chem Mater. 15 (2003) 3294.
52. L. Vayssieres, K. Keis, A. Hagfeldt, S. Lindquist, Chem. Mater. 13 (2001) 4395.
53. Y. Tak, K. Yong, C. Park, J. Crystal Growth 285 (2005) 549.
54. M. Wang, E.J. Kim, S.H. Hahn, C. Park, Physica Status Solidi (a) 203 (2006) 2418.
55. R. Corkish, D.S.P. Chan, M.A. Green, J. Appl. Phys. 79 (1996) 195.
56. Y. Zhang, A. Mascarenhas, S. Deb, J. Appl. Phys. 84 (1998) 3966.
57. M. Burgelman, B. Minnaert, Thin Solid Films 511 (2006) 214.
58. B. Kannan, K. Castelino, A. Majumdar, Nano Lett. 3 (2003) 1729.
-2
-3
-0.1
0.0 0.1 0.2 0.3 0.4
Voltage (V)
Figure 2-37. Simulated J-V curves for cells with graded absorption profiles.
0
-0.1 0.0 0.1 0.2 0.3 0.4
Voltage (V)
Figure 2-38. Simulated J-V curves with CIS replacing P3HT.
- ZnO:P3HT with 10 nm absorption region
- ZnO:CIS with 10 nm absorption region
ZnO:P3HT with full absorption region
ZnO:CIS with full absorption region
- ZnO:CIS with 10 nm absorption region and Eg = 0.85 eV
Figure 3-18. SEM images of selected hybrid films. The first column image was taken at 5,000x
magnification and the second column image was taken at 15,000 x magnification.
N HD2 1 5 0 k V X 5 0 6 k' C; 6'0'A" ri'l
N IA D 2 1 5 0 k V X 5 0 C;
N-HD2 1 5 0 k V X 5 0 6 k' E;
N 1-1 D 2 1 5 0 k V X 1 5
N-IAD2 15.OkV X15.OK 2.00pro
C
N HD2 1 5 Ok V X 1 5 2: 6'0'r
2 5 0 0
P E D 0 T P S S F im
P E D 0 T P S S o luto n
2 0 00 15 00 1 0 0 0
3
E
W a e n u m b er (cm )
Figure 3-7. Upper spectrum: FTIR spectrum of water from Sigma Aldrich. Lower spectrum:
PEDOT:PSS film (gray) spin-coated from solution of PEDOT:PSS in water (blue).
55
WAPEDOT:PSS
PEDOT:PSS
PEDOT:PSS in Water
3 0 AI-
4 000
3 M 3 2 M M = i 1 1M 12 I a
-- - --- -.----- ."
S11f 1
:: t/
,,-
i /"-
ii
0 -
On the right-side of the cell, light would have to pass through 790 nm of the non-absorbing
P3HT2 region before returning to the thin absorbing layer of polymer and ZnO at the surface of
the cell. Based on this, it is obvious that this simulation allows more light to be absorbed in the
generating regions of the model than would exist in reality. This was corrected by studying the
effect of reducing, reflection at the back electrode.
The absorption coefficient of P3HT ranges from 2 x 104 cm-1 at X < 350 nm and X > 680
nm to a maximum value of- 2 x 105 for X = 540 nm. At the low-end absorption value,
approximately 20% of incident light is absorbed over the 10 nm LD region, according to
Equation 2-1.
I0
_e (2-1)
Ii represents the amount of light transmitted through the film, where Io is the intensity of incident
light, a is the absorption coefficient, and x is the film thickness. Nearly 99.8% of the remaining
photons would be absorbed over the 320 nm path length consisting of the forward and backward
pass through the non-absorbing region in the simulation. Even in this case, with a low
absorption coefficient and the shortest considered path length, virtually no photons would remain
in a real cell to be absorbed after reflection from the back electrode.
With this consideration, the next line of simulations was performed with reflectance at the
Ag electrode either reduced or completely removed. The resulting J-V curves are shown in
Figure 20. Despite the removal of all reflectance from the Ag electrode, simulated cell
performance was still substantially higher than that of the real cell for all simulations. All of
these simulations prevent absorption in the P3HT2 region.
By comparing simulations with varying electrode reflectance and constant mobility and
absorption properties (mobilities as defined in Table 2-1, absorption set to zero in the P3HT2
Table 3-9. The cell fabricated from a 25 mg/ml solution of equal weights P3HT and CdSe
dissolved in 2% pyridine in chloroform with minimal air exposure showed the highest
performance by simultaneously demonstrating the highest short-circuit current density and open-
circuit voltage of all cells measured.
The top-performing bi-layer and hybrid cell J-V curves are plotted together in Figure 3-34.
The current flow through the bi-layer cell is an order of magnitude higher than that of the hybrid
cell, demonstrating the need for improved morphology control in the hybrid active layers. The
bi-layer cell shows a low Voc compared to that of the hybrid cell.
The series and shunt resistances of these cells were estimated from the J-V curves using
Equation 3-6 and Equation 3-7, respectively [62]. This calculation is good for shunt resistance,
but series resistance is more accurately calculated as the applied voltage approaches infinity. For
the bi-layer cell, J-V data was not sufficiently collected to make this calculation. From these
equations, resistances for the bi-layer cell were calculated as Rs = 1.59 x 103 Q and Rsh = 2.25 x
103 Q. For the hybrid cell, resistances were calculated as Rs = 4.5 x 104 Q and Rsh = 8.33 x 104
Q. The series resistance for the hybrid cell calculated at the maximum measured voltage point
was 1.5 x 10-2 Q. For another bi-layer cell with a more extensive set of J-V data, the series
resistance is calculated as 1.8 x 102 Q.
R (3-6)
s dl )I=)
Rh(d- (3-7)
These resistance calculations show shunt resistances approximately 5 orders of magnitude
higher than the series resistance, which should result in high-quality solar cells. Further
improvements in cell design to maximize absorption and charge separation in the hybrid cells
the encapsulant and the dessicant. The encapsulant consisted of a small glass slab to serve as the
backing and a rubber spacer to prevent contact between the substrate and encapsulant glass. The
assembly of substrate / spacer / encapsulant was sealed with epoxy.
Prior to fixing the encapsulant onto the substrate, a dessicant was added to protect the cells
from moisture exposure. The dessicant used was barium oxide. A small pouch was created from
part of a piece of weighing paper. Inside the nitrogen glove box, the pouch was filled with BaO
powder and sealed with double-sided tape. It was then affixed to the inside of the encapsulant
glass. Several tiny holes were punctured in the pouch to allow moisture to reach the dessicant,
and the final assembly was sealed to the substrate with epoxy.
Film Drying
Due to the potential impact of residual solvent in these films, care must be taken to ensure
drying is complete after each film deposition step. Residual organic solvent from the active layer
film serves as an insulator to cripple electrical performance, while excess water remaining from
the HTL film can oxidize and degrade the active layer polymer. To confirm the effectiveness of
the drying step, FTIR spectra were compared between the dry films, the solutions used for spin-
coating, and spectra obtained from Sigma-Aldrich for the pure solvent. Figure 3-6 shows the
spectra for a P3HT film spin-coated from chlorobenzene, as well as spectra for the solution and
pure solvent. The large peak at approximately 3050 cm-1 in the chlorobenzene spectra is clearly
visible in the solution spectra, but is noticeably missing in the P3HT film spectra. Also, several
sharp, narrow peaks between 1600 and 500 cm-1 match in the solution and solvent spectra, but
are missing from the film spectra. In Figure 3-7, the broad O-H stretching peak from
approximately 3700 to 3000 cm-1 is an obvious feature in the FTIR spectrum for pure water.
This wide peak is also obvious in the solution spectrum but is missing from the PEDOT:PSS film
spectrum. Similarly, the strong peak at approximately 1640 cm-1 in the water spectrum is present
-0.2 -0.1 0.0 0.1 0.2 0.3
Voltage (V)
0.001
-0.4 -0.2 0.0 0.2 0.4
Voltage (V)
Figure 3-13 continued. J-V curves under 100 mW/cm2 illumination (C and D).
region) we can see the effect of reflectance at the Ag electrode with no absorption in the P3HT2
region. These curves are shown in Figure 2-19. As expected, the short-circuit current density
decreases as the Ag reflectance is decreased. The simulation with the Ag reflectance set to zero
results in a Jsc = 3.6 mA/cm2, which is still approximately 1.5x higher than the real cell value of
Jsc = 2.2 mA/cm2.
Mobility
In addition to the removal of reflectance from the back electrode, the charge mobilities in
P3HT were reduced to restrict current flow in the device. The physical justification for this is
that the initial value of hole mobility was taken from the literature [71], and was the highest
reported value for hole mobility in P3HT. Additionally, this value is a field-effect mobility
which relies on a strong applied bias to drive current flow. In photovoltaic devices, biases are
significantly lower, and this field-effect mobility may be a serious over-estimation of the true
film properties. Also, charge mobility has been shown to be anisotropic in P3HT [77], so values
of mobility can vary depending on the alignment of the polymer chains in the film.
Based on these reasons, simulations were performed to evaluate the effect of reducing
mobility values in the polymer regions of the model, with the resulting J-V curves shown in
Figure 2-20. As predicted, simulated cells with higher mobilities resulted in stronger device
performance. The black and green curves, corresponding to no change in mobility values and a
50% reduction of mobility values in the P3HT2 region only, are nearly identical. A change from
the initial mobility values to a 50% reduction in both P3HT regions, shown by the black and red
curves, results in an efficiency drop from ir = 2.12% to ir = 2.00%, or approximately 6%.
Reducing the P3HT mobility b 90% in both regions results in an efficiency drop of
approximately 23%, from i = 2.12% to i = 1.64%. Attempts to reduce the P3HT2 mobility
1.5 1.0 0.5 0.0 -0.5 -1.0
Voltage (V)
Figure 3-31. Dark and illuminated J-V curves for hybrid solar cell fabricated from
chlorobenzene with 2% pyridine solution. Red curves represent illuminated J-V and
black curves represent dark J-V.
40
/
30 -
E /
10
10
0 1 2 3 4 5 6 7
Number of Layers
Figure 3-35. Film thickness for multi-layer hybrid films.
# SEM Image AFM Image
Figure 3-36. SEM and AFM surface images of multi-layer hybrid films.
191
P3HT
Prior to cell fabrication, P3HT films were analyzed under AFM to determine the surface
quality of the films. Measurements were made under different solution concentrations and spin-
coating speeds. The results, shown in Figure 3-11, show high-quality films with low surface
roughness, indicating their suitability for bi-layer cell construction. All films spin-cast from the
5 mg/ml solution showed an RMS surface roughness of around 1 nm. For the films cast from the
10 mg/ml solution, the RMS surface roughness was between 2.5 nm and 3.6 nm, with the higher
roughness occurring at the slowest spin speed. The data from the images shown in Figure 3-11 is
tabulated in Table 3-3.
Bi-layer Cell Fabrication
Following these film and process characterization investigations, bi-layer organic solar
cells were fabricated with the device structure ITO/PEDOT:PSS/P3HT/C60/Al. The bi-layer
device structure is commonly used in OLED devices [88] and has been explored in molecular
organic photovoltaics [89], but less work has been performed regarding bi-layer devices using
polymer active layers [90]. The devices were fabricated on ITO-coated glass substrates prepared
as described previously. A PEDOT:PSS film was deposited by spin-coating for 30 sec at 2500
rpm and drying for 30 min. The active layer of P3HT was deposited by spin-coating from a 5
mg/ml solution in 1,2-dichlorobenzene for 30 sec at 3500 rpm and dried under vacuum. The
samples were loaded into an evaporator where a 150 A film of C60 was deposited, followed by a
800 A Al electrode.
Bi-layer cell J-V measurements were performed at Busan National University in Busan,
South Korea using a Keithley I-V measurement system under illumination from a solar
simulator. Current measurements were converted to current density by dividing by the active
cell area (0.04 cm2). Illuminated measurements were performed under 100 mW/cm2
3000
2500 -
" 2000 -
(,
C,
- 1500
--
I--
E
1000
-L
2000
3000
4000
5000
Spin Speed (rpm)
Figure 3-10. Calibration curves for slow- and fast-filtered PEDOT:PSS. Slow-filtered
PEDOT:PSS is shown in the "051116" data set. Fast-filtered PEDOT:PSS is shown
in data sets "051122", "051123", and "051130".
A B C
D LE F D.2
I0 110.
Figure 3-11. AFM images of P3HT films. Images A C were spin-cast from 5 mg/ml P3HT in
chlorobenzene solutions, while images D F were spin-cast from 10 mg/ml P3HT in
chlorobenzene solutions. Images A and D were spin-cast at 2000 rpm, images B and
E at 3000 rpm, and images C and F at 4000 rpm. For all images, the scale bar for the
film height axis is 50 nm, and the scan area is 1 [tm x 1 [tm.
051116
V 051122
051123
051130
* V
*
3-2. Performance of organic solar cells on treated ITO substrates............................................155
3-3. RMS surface roughness of P3HT films shown in Figure 3-11. ........................................163
3-4 J-V data for bi-layer solar cells............................................. ......................................... 166
3-5. Solvents considered for hybrid bulk heterojunction film deposition. ...............................166
3-6. Mean rms surface roughness in nm for hybrid films deposited from selected solvents......170
3-7. Hybrid solutions of P3HT and TOPO-coated CdSe nanocrystals in a mixed solvent of
chloroform and pyridine. ........................................................................ ................... 176
3-8. Film properties for P3HT films deposited from various solvents............. ................181
3-9. Hybrid solar cell fabrication information and performance data. .......................................190
CHAPTER 3
ORGANIC AND HYBRID SOLAR CELL PROCESS DEVELOPMENT
Introduction
The experimental details for process development of organic and hybrid solar cells are
presented in this chapter. In the first section, pretreatment of the indium tin oxide anode is
studied. This work was done in collaboration with the research group of Dr. Chinho Park at
Yeungnam University in South Korea, particularly Jiyoun Seol. This work was previously
presented at the 2006 World Conference on Photovoltaic Energy Conversion sponsored by IEEE
and was published in their Conference Record [80]. The second section focuses on the
development of bilayer photovoltaic cells using absorbing polymers. The next section describes
efforts to characterize solvents appropriate for use in hybrid bulk heterojunction films. The
fourth section details work characterizing hybrid films and the fabrication of photovoltaic cells
from these films. The final section introduces Particle Induced Nanostructuring, a process for
hybrid film deposition intended to control the distribution of nanocrystals in the polymer matrix.
ITO Anode Treatment
Organic solar cells incorporate transparent conducting substrates as an anode, with indium-
tin-oxide (ITO) coated glass most often used. ITO films offer several positive characteristics as
substrates for optical devices, including a high luminous transparency, good electrical
conductivity, and good infrared reflectivity. For these reasons, ITO is widely adopted as
transparent anodes in light-emitting diodes, liquid crystal displays, and solar cells [4, 81-82].
ITO coated glass substrates are commercially produced by sputter deposition followed by
processes to improve surface roughness and microstructure. As-received substrate surfaces,
however, have to be further processed prior to application to current flowing devices such as
OLEDs and solar cells because a sputter-deposited surface microstructure and chemical
ACKNOWLEDGMENTS
I acknowledge my loving wife, Kim, for her patience, perseverance, and support. She was
a source of inspiration, and without her this could not have been possible. I thank my parents for
their continued love and support that has guided me through my personal and academic life.
I thank my advisor, Dr. Tim Anderson, for his advice and guidance in my research. I
would like to thank all of my committee members for their guidance and direction. Dr. Chinho
Park granted me the use of his laboratory and much assistance and advice in organic
photovoltaics. Dr. Kirk Ziegler provided laboratory space and assistance with nanocrystal
chemistry. Dr. Oscar Crisalle and Dr. Sheng Li provided valuable guidance through their roles
in the interdisciplinary photovoltaics team.
I thank Dr. Chinho Park's students at Yeungnam University (particularly Jiyoun Seol,
Young Wook Kim, Trong Nguyen Tam Nguyen, and Md. Azizul Hasnain) for extensive
experimental support and for making a foreign country feel like home. I thank Dr. Woo Kyoung
Kim for sharing his expertise in Medici and photovoltaics. I thank Dr. Ziegler's students,
particularly Justin Hill and Randy Wang, for assistance with their laboratory equipment and for
many helpful discussions.
I thank all of the support staff at the Chemical Engineering department, particularly Sean
Poole for helping me set up my simulation work. Special thanks go to Sherrie Jenkins, for
performing scheduling miracles on a regular basis.
Finally, I would like to thank all of my friends and family that have stood behind me and
helped me through my doctoral research and my whole life up to this point. They have helped
mold me into the person I am today, and I am grateful for it.
1 ey
Figure 3-37. Optical microscope images for multi-layer hybrid films. The number of film layers
were a) 1, b) 2, c) 3, d) 4, e) 5, and f) 6.
should result in further reduction of the series resistance and drastic improvements in current
flow.
Particle Induced Nanostructuring
A key challenge in hybrid bulk heterojunction solar cells is control of the nanocrystal
distribution throughout the active layer. The film must be well-mixed to allow efficient exciton
dissociation, but also provide percolation pathways to provide charge collection pathways. A
new concept developed to resolve this issue is particle induced nanostructuring (PIN). The PIN
concept involves depositing multiple ultra-thin layers to control nanocrystal distribution
throughout the thickness of the film.
In this study, ITO-coated glass substrates were chemically cleaned and treated with
nitrogen plasma before film deposition. Multiple layers were deposited from a weakly
concentrated 5 mg/ml solution of 60 wt. % CdSe and 40 wt. % P3HT in chloroform with 2%
pyridine by volume. Deposition occurred at a spin-coating speed of 5000 rpm to produce very
thin films. The thicknesses of the films were measured with a profilometer after removing a
portion of the film to create a step.
Film thickness measurements on samples with 1 to 6 film layers showed an interesting
phenomenon. As expected, the film thickness grew with the addition of multiple layers. The
film thickness increased, however, only after two deposition steps were performed, as shown in
Figure 3-35. On the first deposition, a film of approximately 10 nm was deposited on the
substrate. After the second, the film thickness remained constant. On the third, another film of
approximately 10 nm was deposited, followed by another spin-coating resulting in no film
growth. After the next deposition step, the film grows by another 10 nm, with the 6th resulting in
a minimal amount of growth.
rather than 10 nm. This sets the nanorod half-width in the unit cell simulation area as 2 nm. The
resulting simulated J-V curves are shown in Figure 2-6. The length listed in the legend is the
unit cell thickness, not including electrodes, so the CdSe rod length is 10 nm shorter than that
distance due to the constant 10 nm capping layer.
The cells showed an increase in Voc, Jsc, FF, and efficiency as the cell thickness increased
due to increased absorption in the devices. However, due to increasing series resistance, there
was a diminishing return as the film thickness increased. Interestingly, the short-circuit current
density continued increasing as the film thickness was increased to 1 [tm, despite P3HT's low
hole mobility of 0.01 cm2/V-s.
Figure 2-7 displays solar cell performance measures for the simulated J-V curves shown in
Figure 2-6, with additional data points that were not displayed in Figure 2-6 for clarity. The
short-circuit current density of the unit cells continues to increase with cell thickness, although it
begins to level off. The fill factor and Voc of the cells reach a maximum in this simulation, with
the value of the 1 [tm cell showing slightly lower values than the 500 nm cell. The peak values
are approximately Voc = 60 mV and FF = 0.36. The efficiency continues increasing up to the 1
[tm cell thickness, but as with the short-circuit current, the rate of increase slows dramatically. A
summary of the results from these simulations is shown in Table 2-5.
With the effects of varying cell thickness characterized, simulations were then performed
to determine the impact of nanorod width on cell performance. Similar to the thickness
variations displayed previously, these simulations maintain a constant P3HT thickness of 10 nm
on the top and side of the CdSe nanorod. The resulting J-V curves are shown in Figure 2-8. The
values displayed in the legend represent the nanorod half-width rather than the unit cell width.
The unit cell width is 10 nm higher than the listed value due to the constant 10 nm of P3HT on
LIST OF FIGURES
Figure page
1-1. Worldwide cumulative installed PV Power in Megawatts from 1992 to 2006...................26
1-2. Common organic materials used in solar cell development ............................................26
2-1. Hybrid solar cell and corresponding unit cell used for device simulation ..........................76
2-2. Wavelength-dependent absorption coefficient data used in simulations............. ...............77
2-3. J-V curves for simulated hybrid solar cells with different methods of specifying doping
density ................ ............... .........................77
2-4. Simulated J-V curves showing the effect of doping density in the CdSe nanorods..............78
2-5. Variation of open circuit voltage with doping density of CdSe nanorods...........................79
2-6. Simulated J-V curves with varying unit cell thickness............................... ............... 80
2-7. Solar cell parameters for unit 12 nm wide unit cells with varying cell thickness ................80
2-8. J-V curves for hybrid cells with varying nanorod width ............................................. 81
2-9. J-V curves for simulated hybrid cells with CdSe nanorod half-thickness between 10
an d 2 5 nm ................................................................................82
2-10. Solar cell performance measures for simulated hybrid cells with varying CdSe half-
w idth .......................................................... ................................... 82
2-11. J-V curves for hybrid solar cells with different light source specifications ......................83
2-12. Illustration of the PHOTOGEN command in Medici .......................................................84
2-13. SEM images of ZnO nanofibers and nanofiber and P3HT composite films........................ 84
2-14. J-V curve for a real ZnO:P3HT solar cell to be used for verification of Medici
sim ulations. ............................................................................... 85
2-15. J-V and P-V curves for the real solar cell fabricated by Olson et al. .................................85
2-16. Unit cell used for simulations of ZnO/P3HT hybrid cells.. .............................................86
2-17. M edici unit cell used for device simulation.......................................................... ......... 87
2-18. Simulated J-V curves for ZnO:P3HT solar cell using two P3HT regions..........................88
Surface profiles of the films generated with a profilometer are shown in Figure 3-27. For
these scans, the film was wiped clean at the left-hand side of the image to allow measurement of
the film thickness. Note that the scale for both profiles is the same. The P3HT film shows a
thickness of 95.7 + 17.8 nm with an rms roughness of 15 nm. The hybrid solution produces a
slightly thicker film at 130 16.8 nm, but with an rms roughness of 239 nm. The inclusion of
nanocrystals increases the surface roughness of the film by more than an order of magnitude.
However, the nanocrystals have only a minimal effect on the film thickness. Although the
hybrid solution had a total solute concentration that was twice as high as the P3HT solution, the
final film thickness only increased by 35%. This demonstrates that the polymer is a much
stronger factor in film thickness than the nanocrystals.
Dark and illuminated J-V curves for cells fabricated from the 25 mg/ml hybrid solution
and the 12 mg/ml P3HT solution are shown in Figure 3-28. The hybrid cell showed performance
characteristics of Jsc = 4.99 x 10-2 mA/cm2, Voc = 0.701 V, FF = 0.232, and efficiency =
0.0175%. These measurements show a short-circuit current density 150% higher than that of the
10 mg/ml hybrid cell shown in Figure 3-26. As expected, J-V curves for the pure P3HT cell
showed poorer performance than that of the hybrid cell. Performance characteristics for the
P3HT cell showed Jsc = 2.03 x 10-2 mA/cm2, Voc = 0.523 V, FF = 0.310, and efficiency = 1.25 x
10-3 %, which are similar to that of the 10 mg/ml hybrid cell. Despite the poor measurements for
the P3HT cell, the dark J-V curve shows a strong rectification ratio, signifying a strong diode.
J-V measurements in the dark for each of these cells showed a strong rectification ratio,
something that was not observed previously for thinner cells. This shows that the thicker films
limit reverse leakage current that was easily driven through the thin active films in the previous
cells.
Table 3-4. J-V data for bi-layer solar cells.
Sample RR Jsc (mA/cm2) Voc (V) FF 1 (%)
I-1 2.4 0.55 0.16 0.23 0.02
I-2 1.03 0.15 0.19 0.03
1-3 1.55 0.11 0.23 0.04
I-4 0.41 0.16 0.15 0.01
II-1 8.53 1.10 0.08 0.24 0.02
II-2 3.96 0.68 0.07 0.22 0.01
II-3 1.81 0.70 0.06 0.19 0.01
II-4 0.88 0.07 0.21 0.01
Table 3-5. Solvents considered for hybrid bulk heterojunction film deposition.
Solvent B.T. (C) Polarity P3HT Solubility
Acetone 56.2 5.1 No
2-butanol 79.6 4.0 No
DMF 153 6.4 No
Methanol 64.6 5.1 No
2-propanol 82.4 3.9 No
MEK 80.0 4.7 Poor
Pyridine 115.3 5.3 Poor
Benzene 80.1 2.7 Yes
Chlorobenzene 131.7 2.7 Yes
Chloroform 61.2 4.1 Yes
o-dichlorobenzene 180 2.7 Yes
THF 66.0 4.0 Yes
Toluene 110.6 2.4 Yes
TCE 87.2 1.0 Yes
o-xylene 144.4 2.5 Yes
-1.0
-1.5
-2.0
-2.5
-3.0
-3.5 -
-4.0
-0.2
I I I I I
0.0 0.2 0.4 0.6 0.8 1
Voltage (V)
Figure 2-20. Simulation results showing the effect of changing charge mobilities in the P3HT
regions.
Table 2-9. Performance measures for simulated cells with varying carrier mobility.
Mobility (cm2/V-s) Jsc (mA/cm2) Voc (V) FF q (%)
Cln = 0.01
=0.01 3.61 1.04 0.57 2.12
pp = 0.001
pn = 0.005
S0.005 3.58 1.04 0.54 2.00
pp = 0.0005
Cln = 0.01
p.=. 3.58 1.04 0.57 2.13
pp = 0.0005
pn = 0.001
S0.001 3.44 1.04 0.46 1.64
pp = 0.0001
-- tsf496_05
n = 0.001 cm2/V-s, Lp = 0.01 cm2N-s
- tsf496 07
i's reduced by 50% in both P3HT regions
tsf496 08
i's reduced by 50% in P3HT2 region
tsf496 09
i's reduced by 90% in both P3HT regions
in the solution spectrum but missing from the film spectra. These FTIR spectra provide
confirmation that film drying is complete under the conditions specified, and in that respect,
these films are suitable for use in organic solar cells. Both solvents analyzed, chlorobenzene and
water, produce strong peaks that are easily distinguished from the polymer spectrum, making this
an effective technique for identifying residual solvent in the films after deposition.
PEDOT:PSS
The PEDOT:PSS layer in bi-layer solar cells was deposited with a thickness in the range of
80 to 100 nm. One potential issue in this charge transport layer is that it is susceptible to
pinholes, which can lead to shorting of the devices. Pinholes can be generated during the drying
process as the solvent evaporates and rises through the drying polymer. To see the if pinholes
were problematic in this deposition process, cells were fabricated using a single-layer and
double-layers of PEDOT:PSS. The double-layer devices should eliminate the presence of layer-
spanning pinholes by providing two separate films so that any pinholes would only span half of
the final film.
Experiments were performed by fabricating bi-layer solar cells using single- or double-
layers of PEDOT:PSS. A consistent final film thickness of 80 nm was used for the PEDOT:PSS
layer: one 80 nm layer for the single-layer devices, and two 40 nm layers for the double-layer
devices. All other cleaning, preparation, deposition, and encapsulation steps were held constant.
After fabrication, the cell performance was characterized under 100 mW/cm2 light from a solar
simulator. The resulting J-V curves are shown for single-layer devices in Figure 3-8 and for
double-layer devices in Figure 3-9.
From the results, it is clear that the performance of cells using the double-layer
PEDOT:PSS film suffer dramatically. Of the four cells using the double-layer structure, two fail
to show any diode characteristics, while the other two show only a minimal photovoltaic effect.
0.05
0.10
0.15
E
0.20
0
0 0.25
M 2e+21
0.30 M 4e+21
M 6e+21
8e+21
0.35 M le+22
0.40
0.00 0.01 0.02 0.03 0.04 0.05 0.06
X-Coordinate (tm)
Figure 2-54. Photogenerated carrier distribution for a simulated cell with LD = 20 nm.
1) Mesh definition
2) Region definitions
3) Electrode definitions
4) Material property definitions
5) Illumination definition
6) J-V calculation
7) Output definition.
The simulation mesh is the collection of points where calculations are performed
throughout the simulation area. The spacing of the mesh points can be specified by the user and
is somewhat arbitrary, with the exception that mesh points should exist at boundaries between
regions to facilitate convergence of the calculations. Medici allows a maximum of 3,200 mesh
points, with more points providing a more well-defined calculation at the expense of
computational time.
Regions of the mesh can be defined to correspond to different materials in the simulation.
These are specified through ranges of (x, y) coordinates in units of [m, with y = 0 corresponding
to the top of the simulation area. The electrodes are specified in terms of their optical properties
and are not defined by (x, y) coordinates. Calculations are performed by Medici at the interface
points with the electrodes, but not within the actual electrode.
Material properties are specified to correspond to the material ranges given in the region
definition. An example would be
MATERIAL REGION=P3HT PERMITTI=3 NC300=2E18 NV300=2E19 EG300=1.7
+ EGO300=1.7 AFFINITY=3.15 ABS.FILE="abs-p3ht.txt" PR.TAB
where the properties for the area defined as "P3HT" are specified. Carrier mobilities and
impurity profiles are defined in separate statements. Further details on impurity profile
definitions are given in the following section. The absorption coefficient is defined through an
external text file, where the value can be specified by the user for a range of wavelengths.
mobilities in P3HT from the field effect mobility values found in literature resulted in a decrease
in the fill factor to levels similar to the real cell.
Although the total number of photogenerated carriers increased, there was not a uniform
increase at all points in the film. Specifically, in the column of polymer within one exciton
diffusion length of the side of the nanorod, the number of carriers decreases quickly for the
simulation with an increased absorption coefficient. This is because the number of photons
penetrating deeper into the film is reduced due to the stronger absorption. This is clearly seen in
Figure 2-62, which shows a comparison of carrier generation for the increased absorption case
and the standard absorption case. The carrier distributions were subtracted, and green regions
show no change in absorption between the two simulations. Red, orange, and yellow regions
show areas where the 150% absorption coefficient resulted in stronger absorption, and this is
primarily contained within a range of 40 nm (LD) from the upper surface of the ZnO-P3HT
interface. Blue regions represent areas where the 150% absorption coefficient resulted in less
absorption, and this is limited to the exciton diffusion length region along the nanorod edge.
This is due to strong absorption near the surface of the P3HT region, leaving fewer photons
available for absorption in the deeper region.
Although there are regions of increased and decreased carrier generation in the cell, the
overall number of generated carriers increased by approximately 3 x 1024 pairs/cm3 over the unit
cell. This additional charge generation did not translate into additional photocurrent as expected,
however. Figure 2-63 shows J-V curves for the two cells, calculated with [p = 1 x 10-4 cm2/V-s.
Despite the increased carriers density for the 150% absorption coefficient simulation, the
short-circuit current density decreased by 0.23 mA/cm2. This is very similar to the effect seen
when the full P3HT area was allowed to contribute to absorption. In fact, a comparison of those
cells shows extremely similar properties. Note that the full cell simulation was performed with
the standard values for mobility in the P3HT regions, while the two 40 nm simulations were
Figure 2-53. Carrier mapping scheme for two-stage simulations. Gray regions are ZnO and blue
regions are P3HT. Orange arrows represent the nearest interface points for
photogenerated carriers in the region defined by the dotted white lines.
CHAPTER 4
CONCLUSIONS AND FUTURE WORK
Conclusions
This dissertation presented the results of exploratory research on the processing and
performance of hybrid PV. It provided encouragement for a more complete study of hybrid
photovoltaic devices. With collaboration and assistance from several teams and individuals, the
groundwork was laid for future studies to continue this project and achieve high-performance
hybrid photovoltaic devices.
The innovations of this study include the first simulations of hybrid photovoltaics using
existing semiconductor modeling software, development of anode surface treatment processes,
solvent selection for hybrid films, and hybrid bulk heterojunction photovoltaic process
development, including an interesting multiple spjn coating process sequence to better disperse
the inorganic phase.
Hybrid Photovoltaic Simulation
Simulations of an ordered heterojunction photovoltaic cell provided interesting results. It
was found that reported values of certain parameters reported in the literature from organic field-
effect transistor fabrication were poor estimates for organic photovoltaic simulation.
The hole mobility values given in the literature [71] proved to be higher than the
simulations estimated. This value of 0.01 cm2/V-s for the hole mobility produced J-V curves
with fill factors around 0.85, which is considerably larger than published values which typically
range from 0.4 to 0.6 [27, 33-34, 44, 75]. By reducing the mobility to 1 x 10-4 cm2/V-s the fill
factor was reduced to 0.78, which is closer to the published range.
The short-circuit current density of the real cell could not be matched by the simulations.
All attempts using the two-step model gave values lower than the observed ones, including
-0.1 0.0 0.1
0.2 0.3 0.4
Voltage (V)
Figure 2-48. Simulated J-V curves for ZnO:P3HT solar cells with varying absorption
coefficients in P3HT.
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Voltage (V)
Figure 2-49. Simulated J-V curves for ZnO:P3HT solar cells with varying energy band gap.
tsf496_98 100% absorption
tsf496_103 80% absorption
-- tsf496_104 50% absorption
- tsf496_105 20% absorption
t. 7
-
1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Voltage (V)
Figure 2-42. Simulated ZnO:CIS solar cells with P3HT values for absorption coefficient,
electron affinity, and energy band gap.
Table 2-16. Performance measures for simulated ZnO:CIS solar cells with P3HT values for
absorption coefficient, electron affinity, and energy band gap.
Simulation Properties Varied Voc (V) Jsc (mA/cm2) FF (%)
Absorption
80 Affinity, E, 0.265 3.633 0.418 0.402
Permittivity
Absorption
81 Affinity, Eg 0.281 3.608 0.395 0.401
Density of States
Absorption
82 Affinity, E, 0.423 2.164 0.159 0.145
Mobility
-- tsf496_80 P3HT Absorption, Affinity, Eg, Permittivity
-- tsf496_81 P3HT Absorption, Affinity, Eg, Density of States
tsf496_82 P3HT Absorption, Affinity, Eg, Mobilities
1.5
0.4
1.0 0.2
E O0 ....*****
-.-2 02
E 0.5 -
>. -0.4
-0.2 0.0 0.2 0.4 106 0.8 1.0
0.0
-0.5
-1.0
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
Voltage (V)
Figure 3-32. Dark and illuminated J-V curves for a hybrid solar cell fabricated with commercial
CdSe nanopowder. Green curves represent illuminated J-V and black curves
represent dark J-V.
spots in the secondary electron image. Images for chlorobenzene and o-dichlorobenzene show
high-quality films with relatively few surface features. The images for TCE also shows a
minimal amount of surface features, particularly in the 15,000x image. Chloroform displays a
moderate amount of features, but less so than THF, which similarly to the optical images shows
the poorest film quality.
Attempts to identify the composition of these surface features failed to yield results.
Backscattered electron images of the films showed images of the features similar to the
secondary electron images, but with no additional bright spots to demonstrate possible clustering
of the higher density nanocrystals. Compositional scans were performed using energy-dispersive
x-ray analysis (EDX), and these scans detected cadmium and selenium at a constant
concentration in both the features and in smooth areas of the film surface.
From this study, it was determined that chloroform, chlorobenzene, and o-dichlorobenzene
are good candidates for hybrid bulk heterojunction film deposition. These solvents all produce
films that were among the lowest in surface roughness for all of the measurement techniques
used. Additionally, the film quality could be visually confirmed from optical microscopy and
SEM surface images.
Hybrid Bulk Heterojunction Cell Fabrication
Bulk heterojunction photovoltaic cells were fabricated using P3HT as the absorbing
semiconductor and nano-CdSe as the electron transporter. The cell design was the modeled after
the bi-layer organic cell design described previously. The bulk heterojunction cell structure was
ITO/PEDOT:PSS/P3HT:CdSe/Al, which is very similar to the bi-layer structure with the
exception of CdSe replacing C60 as the electron acceptor, and that acceptor is now blended into
the active layer film rather than deposited on top. Performance measurements for these cells
were performed in-house rather than remotely, so the encapsulation process was not performed.
-- 90% Reflectance
-1.550% Reflectance
0% Reflectance
o" -2.0
E
< -2.5
cn -3.0
C
(.
C)
S-3.5 /
0 -4.0 .
-5.0
-0.2 0.0 0.2 0.4 0.6 0.8 1.0
Voltage (V)
Figure 2-19. J-V curves for simulated cells with zero absorption in the P3HT2 region and
varying reflectance from the Ag electrode.
note that for the pyridine solvent, the error range for the mean film thickness and the standard
deviation of film thickness is larger than the actual measurement. This is due to the extremely
nonuniform film which shows extremely large features over a very thin base film.
Cell Fabrication
Replacing TOPO-coated nanocrystals with pyridine-coated nanocrystals is the key to
reducing the amount of pyridine needed in the solvent mixture. Because of this, pyridine-coated
nanocrystals were used for attempts at hybrid bulk heterojunction cell fabrication. The
processing steps for these cells were similar to those used for the organic bi-layer solar cells
described earlier. ITO-coated glass substrates were etched with HC1, then cleaned under
ultrasonication in TCE, acetone, and methanol. The clean surface was exposed to a N2 plasma
under specific conditions, followed by deposition of PEDOT:PSS, then the hybrid solution,
followed by evaporation of the Al electrode. Cell performance was tested in the dark and under
simulated solar illumination.
Cells deposited from chloroform solutions
Bulk heterojunction cells were fabricated from pure chloroform solution and a 2% pyridine
in chloroform solution. Nitrogen plasma treatment was performed at 50 W and 200 mTorr for 10
minutes after etching and cleaning of the substrates. A thin film of PEDOT:PSS was deposited
via spin-coating and dried under vacuum. The hybrid solution concentration was 5 mg/ml,
consisting of 60% CdSe by weight. The aluminum electrode was between 125 and 150 nm in
thickness. The resulting dark and light J-V curves for these cells are shown in Figure 3-25.
The active cell performance is low for both cells, particularly the cell deposited from pure
chloroform solvent. The first cell, deposited from the chloroform-pyridine mixed solvent, shows
Jsc = 8.99 x 10-3 mA/cm2 with a maximum efficiency = 7 x 10-4 %. The second cell, in pure
chloroform solution, shows Jsc = 2.62 x 10-5 mA/cm2 and maximum efficiency = 1.3 x 10-6 %.
-1.5 0o.o00 7 0 1
o o0.05 0 -
E /
S-1.0 010
E 0.15- o 0
.t: 0.20 o
S-05 0
( 0.25 1 0
Q 0.2 0.0 -0.2 -0.4 -0.6 -0.. O
CO -- Hybrid Dark
0.5 O-.^ 0 Hybrid Illuminated
-8- P3HT Dark
P3HT- Illuminated
1.0
1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5
Voltage (V)
Figure 3-30. Dark and illuminated J-V curves for hybrid bulk heterojunction and pure P3HT
solar cells with limited air exposure during processing.
The illumination source can be defined in multiple ways. For absorption models such as
photovoltaic cells, radiation can be defined to follow any specified profile and generated from a
point source or line source at any location outside of the simulation area. Further details are
given in a later section describing the effect of altering the origin of the illumination source. In
addition to general illumination, sources of photogenerated carriers can be specified along any
line within the simulation area.
With the defined illumination source, Medici calculates the photogenerated carrier
distribution throughout the simulation area based on the absorption coefficient data given for
each material region. The program then begins iterating over a range of applied biases specified
by the user until the current and potential has been mapped and converged throughout the
simulation area. Once a solution has been achieved, Medici provides a vast array of output
mechanisms for the user to extract data from the simulation, including plot generation and
numerical data extraction.
Impurity profile
The way in which Medici assigns impurity profiles in simulations is a bit of a mystery. A
few simulations were performed to determine the proper way to call this function. The profile
can be defined in terms of (x, y) coordinates or by the "REGION" statement which applies
values to an area defined by a specific name. The impurity profile input statements used are
listed in Table 2-4.
The file "impurity_l" specifies the doping areas by the "REGION" statement, which
generates doping in an area which has previously been defined in coordinate space and assigned
a label (P3HT or CdSe). The file "impurity_2" first applies p-type doping at a concentration of 5
x 1016 cm-2 to the entire cell area and then applies an n-type doping of 6 x 1016 cm-2 to the area
occupied by CdSe. This depends on Medici overwriting the impurity values when multiple
0.0 -
-0.5 -
-1.0
-1.5
-2.0 -
-2.5 ,,,,,,
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Voltage (V)
Figure 2-28. Simulated J-V curves for hybrid cells with varying P3HT doping concentrations.
iaculum
Figure 2-29. Energy band diagram for P3HT ZnO hybrid solar cells.
- P3HT Doping: 5x1016
- P3HT Doping: 1x1016
P3HT Doping: 5x1015
P3HT Doping: 1x1015
- P3HT Doping: 5x1014
- P3HT Doping: 1x1014
-- P3HT Doping: 5x101
-- P3HT Doping: lx101
P3HT Doping: 5x10
P3HT Doping: lx10
-- P3HT Doping: 5x10 4
-- P3HT Doping: lx10 4
derivatives have been demonstrated [30]. These cells frequently feature a hole transporting layer
of poly (3, 4-ethylenedioxythiophene):poly (styrenesulfonate) (PEDOT:PSS) to improve hole
collection. The structures of PCBM and some polymers commonly used for bulk heterojunction
solar cells are shown in Figure 1-2.
In particular, polythiophene derivatives are now popular alternatives to PPV, with poly (3-
hexylthiophene) (P3HT) being the most commonly used [29, 31]. P3HT is synthesized with a
regio-regular configuration where the polymer side chains alternate on opposite sides of the
backbone chain. This arrangement aids in aligning the polymer chains for efficient charge
transfer along the backbone, with additional chain straightening attributed to steric hindrance
from the fullerene molecules [32]. Bulk heterojunction cells fabricated from P3HT and PCBM
have reached efficiencies of 3.5% in 2003 [33] and 4.4% in 2005 [34].
The use of inorganic nanocrystals as a replacement for C60 is a relatively recent trend. In
addition to their high electron mobilities, semiconductor nanocrystals can contribute to
absorption and photocurrent generation in the heterojunction active layer. These nanocrystals
offer the possibility of bandgap engineering by material selection and tightly controlling the size
distribution, as well as the growth of different crystal shapes such as rods and tetrapods [35-37]
to create more efficient charge transport pathways. As with early work involving C60, dispersion
of the nanocrystals remains an issue in process development. One approach to enhance
nanocrystal dispersion is the use of sol-gel processing to grow nanocrystals in the polymer film
[18, 38-40]. The nanocrystals are typically grown in solution, and are formed with a surfactant
capping layer to relieve the high surface energy [41, 42]. Exchange of this surfactant has been
demonstrated to obtain improved solubility and electrical properties [43-45].
of the nanocrystal surfactant, selection of an appropriate solvent for film deposition, and the
introduction of a novel layer-by-layer deposition process.
Table 3-3. RMS surface roughness of P3HT films shown in Figure 3-11.
Label Solution Spin Speed RMS Roughness
A 5 mg/ml 2000 rpm 0.89 nm
B 5 mg/ml 3000 rpm 1.27 nm
C 5 mg/ml 4000 rpm 0.94 nm
D 10 mg/ml 2000 rpm 3.66 nm
E 10 mg/ml 3000 rpm 2.59 nm
F 10 mg/ml 4000 rpm 2.54 nm
/
/
/
0 -- -"
-4 I I I
-0.2 -0.1 0.0 0.1 0.2 0.3
Voltage (V)
Figure 3-12. J-V curves for bi-layer solar cells fabricated on untreated ITO substrates.
0--- Set I dark
........ v ........ S et I 1
Set I 1
--- --- Set 1-2
et I 3
Set I 4
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Voltage (V)
Figure 2-40. Simulated J-V
applied.
curves for ZnO:CIS solar cells with the P3HT absorption spectrum
Table 2-14. Performance measures for simulated ZnO:CIS solar cells with the P3HT absorption
spectrum.
Simulation Properties Varied Voc (V) Jsc (mA/cm2) FF (%)
72 Absorption 0.421 4.105 0.762 1.317
73 Absorption 0.427 4.065 0.765 1.328
Permittivity
74 Absorption 0.423 4.105 0.765 1.328
Density of States
75 Absorption 0.467 3.980 0.493 0.916
Affinity
76 Absorption 0.517 3.293 0.657 1.118
Mobility
- tsf496_72 P3HT Absorption
- tsf496_73 P3HT Absorption, Permittivit
tsf496_74 P3HT Absorption, Density of
tsf496_75 P3HT Absorption, Affinity
- tsf496_76 P3HT Absorption, Mobilities
-5
-10
-0.1
y
States
100
5D
1 D-
41i
P3HT
PHTF i P3HT in Chlorobenzene
P T So I j io
W a ve n u n b e r ( m -1 )
Figure 3-6. Upper Spectrum: FTIR spectrum for chlorobenzene from Sigma Aldrich. Lower
Spectrum: P3HT film (purple) spin-coated from P3HT in chlorobenzene solution
(red).
V~Chlorobenzene n _
4 0 O 0
20
P F..
surface of the cell and the definition of the ray width are unimportant, provided that the ray width
is high enough to illuminate the entire cell. Further studies will focus on the simulation of a real-
world hybrid solar cell from literature.
Simulation of a Real Cell
Model Parameters
The cell chosen for Medici simulation was reported in 2006 by Olson et al. [75]. The cell
was a hybrid cell fabricated from an array of ZnO nanofibers and P3HT, and has a cell structure
of ITO/ZnO/ZnO:P3HT/Ag. The ZnO layer was spin-coated onto the ITO substrate and the
nanofibers were grown through hydrothermal growth. P3HT was then spin-coated at a reported
thickness of 200 nm. SEM images of the ZnO nanofiber array before and after P3HT spin-
coating are shown in Figure 2-13. The performance measures reported for the device were as
follows: Voc = 440 mV, Jsc = 2.2 mA/cm2, FF = 0.56, and rI = 0.53%. The J-V curve for this
device is shown by the solid line in Figure 2-14.
From the SEM images shown in Figure 2-13 (a), the average height of the ZnO nanofiber
array is approximately 260 nm. From Figure 2-13 (b), the thickness of the full ZnO:P3HT layer
is approximately 430 nm. These values leave a P3HT capping layer of approximately 170 nm on
top of the nanofiber array. The image in Figure 2-13 (a) shows an average rod diameter of
approximately 30 nm. The authors note the rod spacing is approximately 100 nm. It is
extremely difficult to accurately estimate the spacing from the cross-section SEM image shown
in Figure 2-13 (a), but this figure appears to be reasonable and was likely verified with other
unpublished data, so it will be assumed to be accurate. From the SEM image, a thin base coat of
ZnO is visible on the ITO film. This film was estimated to be 25 nm thick from the SEM
images, and this height was included in the 260 nm rod length.
While the performance of the cells using the single-layer PEDOT:PSS were not phenomenal,
they all showed a significant photocurrent and diode characteristics.
Cells fabricated with a single 80 nm layer of PEDOT:PSS showed efficiencies as high as
0.168% (cell 8-2), with all cells showing efficiencies of at least 0.047%. In contrast, cells
fabricated with the double-layer structure failed to show a measurable efficiency, although
estimates put them in the range of 0.015% for the best cell (cell 9-1). Open circuit voltage values
for the single-layer cells were in the range of 0.15 V, while values for the double-layer cells are
less than 0.05 V. Short circuit current density values ranged from 1.5 to 2.5 mA/cm2 for the
single-layer cells and from 0.7 to 1.4 mA/cm2 for the double-layer cells.
The data show that pinholes are not a concern for the PEDOT:PSS films under these
deposition and drying conditions. In fact, there is a different effect causing the double-layer
films to perform more poorly than the single-layer films. The total thickness of the layers was
held constant to keep series resistance constant. However, the series resistance is impacted by
the film resistivity in addition to the path length. It seems that the double-layer films showed a
higher resistance to current flow due to the lower cell performance. For positive biases, the
single-layer films showed current densities 7 10 % higher than the double-layer films. This can
be attributed to two possible causes interfacial resistance and film resistance. Because of the
double-layer structure, there is an extra film interface which could cause an increase in the
overall resistance. Additionally, the inherent film resistivity could be increased due to the
deposition conditions. The films were deposited via spin-coating, with the 40 nm films
deposited at a much higher rpm than the 80 nm films. This causes faster solvent evaporation and
can inhibit the ability of the polymers to self-align in a configuration that could minimize
LIST OF REFERENCES
1. International Energy Agency, Trends in Photovoltaic Applications, Report IEA-PVPS
T1-16:2007, August 2007.
2. U.S. Department of Energy, Energy Efficiency and Renewable Energy,
http://wwwl.eere.energy.gov/solar/deployment.html (2006) July 2008.
3. Solarbuzz, LLC., www.solarbuzz.com/moduleprices.htm (2008) July 2008.
4. C.W. Tang, Appl. Phys. Lett. 48 (1986) 183.
5. Md.K.H. Bhuiyan, T. Mieno, Thin Solid Films 441 (2003) 187.
6. N.S. Sariciftci, L. Smilowitz, A.J. Heeger, F. Wudl, Synth. Met. 59 (1993) 333.
7. B. O'Regan, M. Gratzel, Nature 335 (1991) 737.
8. M.K. Nazeeruddin, P. Pechy, T. Renouard, S.M. Zakeeruddin, R. Humphry-Baker, P.
Comte, P. Liska, L. Cevey, E. Costa, V. Shklover, L. Spiccia, G.B. Deacon, C.A.
Bignozzi, M. Gratzel, J. Am. Chem. Soc. 123 (2001) 1613.
9. K. Hara, M. Kurashigo, Y. Dan-oh, C. Kasasa, Y. Ohga, A. Shinpo, S. Suga, K. Sayama,
H. Arakawa, New J. Chem. 27 (2003) 783.
10. Q.B. Meng, K. Takahashi, X.T. Zhang, I. Sutanto, T.N. Rao, O. Sato, A. Fujishima, H.
Watanabe, T. Nakamori, Uragami Langmuir 19 (2003) 3572.
11. L.J. Rothberg, M. Yan, F. Papadimitrakopoulos, M.E. Galvin, E.W. Kwock, T.M. Miller,
Synth. Met. 80 (1996) 41.
12. J.J.M. Halls, K. Pichler, R.H. Friend, S.C. Moratti, A.B. Holmes, Appl. Phys. Lett. 68
(1996)3120.
13. L.A. Pettersson, L.S. Roman, O. Inganas, J. Appl. Phys. 86 (1999) 487.
14. J.J.M. Halls, C.A. Walsh, N.C. Greenham, E.A. Marseglia, R.H. Friend, S.C. Moratti,
A.B. Holmes, Nature 376 (1995) 498.
15. T.J. Savenije, J.M. Warman, A. Goossens, Chem. Phys. Lett. 287 (1998) 148.
16. M. Theander, A. Yartsev, D. Zigmantas, V. Sundstrom, W. Mammo, M.R. Andersson, O.
Inganas, Phys. Rev., B 61 (2000) 12957.
17. H.L. Wong, L.S.M. Lam, K.W. Cheng, K.Y.K. Man, W.K.Chan, C.Y. Kwong, A.B.
Djurisic, App. Phys. Lett. 84 (2004) 2557.
18. G.D. Sharma, R. Kumar, S.K. Sharma, M.S. Roy, S. En. Mater. & Solar Cells 90 (2006)
933.
LIST OF TABLES
Table page
2-1. P3HT Properties for Device Simulations. ........................................................................76
2-2. CdSe Properties for Device Simulations. ........................................ ......................... 76
2-3. Electrode Properties for Device Simulations..................................... .............76
2-4. Impurity profile inputs for simulations .................................................... ...................79
2-5. Solar cell performance measures for unit cells of 12 nm width and varying thickness........81
2-6. Performance measures for simulated hybrid cells with varying CdSe half-width ...............83
2-7. Cell performance measures from published and digitally converted J-V curves .................85
2-8. ZnO properties used for hybrid solar cell simulation ................ .................. ............... 86
2-9. Performance measures for simulated cells with varying carrier mobility...........................90
2-10. Performance measures for simulated cells with varying exciton diffusion lengths. ...........92
2-11. Absorption data and short-circuit current for graded absorption simulations ...................99
2-12. Materials properties for P3HT and CIS used in cell simulations..................................... 101
2-13. Performance measures for ZnO:CIS cells with an individual material property set at
the P3H T value. ............................................................................101
2-14. Performance measures for simulated ZnO:CIS solar cells with the P3HT absorption
sp e ctru m ................... ............................ .......................... ................ 10 2
2-15. Performance measures for simulated ZnO:CIS solar cells with P3HT values for
absorption coefficient and electron affinity. ........................................ ...............103
2-16. Performance measures for simulated ZnO:CIS solar cells with P3HT values for
absorption coefficient, electron affinity, and energy band gap................................... 104
2-17. Cell performance measures for real and simulated solar cells. .....................................118
2-18. Performance measures for simulated cells with 40 nm LD and varying mobility
values ............. ......... ......................................... .............. ................ 118
2-19. Generated carriers and performance measures for simulated solar cells ........................121
3-1. Chem ical com position of ITO film s................................ ...................... ............... 155
Table 3-8. Film properties for P3HT films deposited from various solvents.
Solvent Mean Thickness (nm) Median Thickness (nm) Standard Deviation (nm)
Chloroform 21.2 + 8.18 23.1 + 7.95 9.58 + 5.41
Mixed 56.2 34.3 16.2 4.42 227 157
Pyridine 78.5 85.4 10.9 8.61 338 457
-0.4 4-
-1.5
-1.0 -0.5 0.0 0.5 1.0
Voltage (V)
Figure 3-25. Dark and illuminated J-V curves for cells generated from 5 mg/ml composite
solutions in (A) chloroform mixed with 2% pyridine and (B) pure chloroform. Note
the scale difference of the graphs.
-4 --
a = 100%
-5 i i
0.0 0.2 0.4 0.6 0.8 1.0
Voltage (V)
Figure 2-50. Simulated J-V curves for ZnO:P3HT solar cells with varying band gap and
absorption in the P3HT region.
0.0 0.1
0.2 0.3 0.4 0.5
Voltage (V)
Figure 2-51. Simulated J-V curves for ZnO:P3HT solar cells with P3HT energy band gap of 1.0
eV and hole mobility of 500 cm2/Vs.
-4 -
-40.1
-0.1
The semiconductor properties that are necessary for a proper simulation using Medici
include the electron and hole mobility, band gap energy, permittivity, electron affinity, density of
states in the conduction and valence band, and doping density. These properties are well-known
for most common semiconductor materials, but are less well-characterized for many organic
materials. Despite this, measurements and assumptions regarding the necessary properties for
P3HT can be found in the literature, and these values were used as starting points for device
simulation.
Initial Modeling Efforts
A dispersion of P3HT in an ordered array of CdSe nanorods was chosen as a first attempt at
hybrid solar cell modeling using Medici. The choice of materials and dimensions were
somewhat arbitrary, but represent a pseudo-realistic scenario with which to expore the
capabilities of Medici for hybrid cell modeling. This initial modeling effort serves as a precursor
to attempts to simulate a real cell from literature.
Preliminary Model Description
Hybrid solar cells consisting of an ordered array of CdSe nanorods dispersed in a P3HT
matrix were simulated using the device modeling software package Medici. The nanorods were
assumed to be vertically aligned, be in contact with the aluminum back electrode, have uniform
dimensions, and have uniform spacing. An illustration of this design is shown in Figure 2-1.
Due to the symmetry imposed in this structure, the cell can be divided into a basic repeating unit
cell, as illustrated in the figure. The active layer is fixed at 100 nm thick, while the CdSe
nanorod has dimensions of 90 nm x 10 nm. The rod spacing is fixed at 20 nm. These
dimensions define a P3HT "capping" layer of 10 nm over the tip of each of the CdSe nanorods.
Additionally, the unit cell has a polymer thickness of 10 nm to the side of the nanorod, due to the
20 nm rod spacing imposed in the model. This is chosen because it is equal to the approximate
Figure 3-18 continued. SEM images of selected hybrid films.
tl
N PID2 1 5 0 k V X 5 0 C; i3'0'p' M'
N IAD2 1 5 0 k V X 5 o 6 W 'c; We'
N-HD2 15.OkV X15.OK 200prn
N-IAD2 15.OkV X15.OK 2.eOpm
Figure 2-38 shows that a limited absorption range in the simulations does not cause the
double-elbow shape of the J-V curve, as the ZnO:CIS cell with a 10 nm absorption region shows
a similar shape to the one with absorption in the full CIS region. The same holds true for
comparisons of the two ZnO:P3HT cells, which both show the double-elbow shape. The
ZnO:CIS cells produce J-V curves with the anticipated shape and high fill factor as compared to
the ZnO:P3HT cells. This also held true for the cell in which the CIS electrical bandgap was
reduced to 0.85 eV, although this cell resulted in a low Voc of 0.25 V.
Parameters used for P3HT and CIS in the simulations are displayed in Table 2-12. Data for
CIS was obtained from previous simulations from Woo Kyoung Kim [66]. The doping density is
the only parameter that is the same in both materials. By individually adjusting parameters
between the values for CIS and P3HT, an attempt will be made to determine the origin of the
double-elbow J-V curve shape. Although not shown in the table, the absorption profile for each
material was also adjusted.
Using the same 10 nm absorption region as the ZnO:CIS cell shown in red in Figure 2-38,
the properties shown in Table 2-12 were individually changed from the CIS values to the P3HT
values. The resulting J-V curves are shown in Figure 2-39. Performance measures for these
cells are displayed in Table 2-13. These results show that the permittivity and density of states
have virtually no impact on the cell performance. The three parameters that strongly impacted
the J-V curves are electron affinity, carrier mobility, and absorption profiles.
The change in electron affinity causes a shift in Voc from 0.44 V to 0.49 V due to the
higher conduction band level for P3HT. The short circuit current is minimally affected by this
change, but there is a notable drop in the fill factor of this curve, down nearly 30% from the
values of the original ZnO:CIS cell. The variation in carrier mobilities showed a 1.5 mA/cm2
The authors note that the thickness of their P3HT film was 200 nm, but are somewhat
unclear whether this 200 nm thickness is measured from the top of the nanofiber structure or if it
represents the full film thickness. If the scale bars for the SEM images are reliable, it appears
that the 200 nm figure refers to the excess P3HT film on top of the fiber array. This seems to be
an excessive amount of capping considering that the exciton diffusion length of P3HT is on the
order of 10 nm in P3HT [15], a fact which is referenced by the authors. However, light is
incident from the ITO/ZnO side of the cell in the reported device, so photons must pass through
more than 200 nm ofZnO:P3HT film before reaching this capping layer. Additionally, the
inconsistent length of the ZnO nanofibers must be considered, as the relatively high mobility of
both electrons and holes in ZnO compared to P3HT would create a short-circuit in the device if
the wire tips were exposed. With this consideration, the additional buffer layer thickness may be
important experimentally to ensure that the device will function as a diode.
The J-V curve shown in Figure 2-14 was converted to numerical data using a graph
digitizer program [76] so it could be compared to simulated curves. The converted J-V and P-V
curves are shown in Figure 2-15. To verify the effectiveness of this conversion, the published
cell performance measures were compared to the performance measures calculated from the
digitized curve. The results are shown in Table 2-7. The results were Voc = 0.44 V, Jsc = 2.2
mA/cm2, FF = 0.57, and rI = 0.55%. This shows exact matches for Voc and Jsc and values for
FF and rI that are 0.01 and 0.02 higher than the published results. This is less than a 5% error in
both cases, and demonstrates that the curve was re-produced with a high accuracy. This new
curve can now be plotted with the results of simulated cells to find a best fit.
From previous work performed by Dr. Woo Kyoung Kim, the material properties for ZnO
were defined as shown in Table 2-8 [66].
BIOGRAPHICAL SKETCH
Matthew Lowell Monroe was born in 1978, in Marietta, Georgia, to Ronald L. and Debbie
B. Monroe. He earned a Bachelor of Science degree from the Chemical Engineering Department
at the Georgia Institute of Technology in Atlanta in 2002. He joined the Chemical Engineering
Department at the University of Florida in 2002 and joined Dr. Anderson's research group in
2003. He earned a Doctor of Philosophy in chemical engineering in 2008.
-0.2 -0.1 0.0 0.1 0.2 0.3
Voltage (V)
1000
100
0.01
0.001
-1.0 -0.5 0.0 0.5
Voltage (V)
Figure 3-9. Linear (a) and base 10 log-scale (b) J-V curves for bi-layer organic solar cells
fabricated with two 40 nm thick layers of PEDOT:PSS.
4000
3000-
(,
rS
S2000
1--
1000
0
1000 2000 3000 4000 5000
Spin Speed (rpm)
Figure 3-4. PEDOT:PSS film thickness vs. spin-coater speed.
600
500
i 400
(,
S300
-c
I-
E
'- 200
LL
1000 2000 3000
Spin Speed (rpm)
Figure 3-5. P3HT film thickness vs. spin-coating speed.
4000
-2 5E16
E 05 6E16
,o A 7E16
E -4
-6
0
-8
-10
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Voltage (V)
Figure 2-4. Simulated J-V curves showing the effect of doping density in the CdSe nanorods.
The legend shows the n-type doping density measured in cm-2
Table 2-12. Materials properties for P3HT and CIS used in cell simulations.
Property P3HT CIS
Doping Density p-type 5x1016 cm-2 p-type 5x1016 cm-2
Permittivity 3 13.6
Nc 2x1018 3x1018
Nv 2x1019 1.5x1019
Eg 1.7 eV 1.04 eV
Electron Affinity 3.15 eV 3.93 eV
[te 0.1 cm2/V-s 300 cm2/V-s
[lh 0.01 cm2/V-s 30 cm2/V-s
-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Voltage (V)
Figure 2-39. Simulated J-V curves for ZnO:CIS solar cells with an individual material property
changed to the P3HT value.
Table 2-13. Performance measures for ZnO:CIS cells with an individual material property set at
the P3HT value.
Simulation Parameter Adjusted Voc (V) Jsc (mA/cm2) FF q (%)
64 None 0.439 7.774 0.772 2.633
68 Permittivity 0.443 7.747 0.778 2.669
69 Density of States 0.441 7.773 0.774 2.655
70 Affinity 0.487 7.654 0.485 1.808
71 Mobility 0.545 6.309 0.662 2.277
72 Absorption 0.421 4.105 0.762 1.317
0.0 0.2 0.4 0.6 0.8
Voltage (V)
Figure 2-21.
J-V curves for simulated
cells with varying exciton diffusion length.
Figure 2-22. Extrapolations to estimate Voc for simulated cells with varying exciton diffusion
length.
-3
-4 -
-0.2
-0.2 C
-0.4
-0.6
-0.8
-1
-1.2
-1.4
0.9 0.95
Voltage (V)
v LD= 10 rni
E LD= 9 nm
LD= 8 nm
X LD= 7 nm
SLD= 6 nm
O LD= 5 nm
- Linear(LD = 10 nm)
-- Linear(LD= 9 nm)
SLinear (LD = 8 nm)
------------ Llnear(LD= 8nm)
- Linear (LD = 7 nm)
SLinear (LD = 6 nm)
-- Linear (LD = 5 nm)
).
1 1.05
0.5 1.0 1.5 2.0
Lateral Distance mm)
0.5 1.0 1.5 2.0
Lateral Distance (mm)
Figure 3-24 continued. Surface profiles of P3HT films deposited from chloroform (A), 1:1
chloroform:pyridine (B), and pyridine (C) solvents.
180
indicative of a lower effective doping density. The purpose of input files "impurity_4" and
"impurity_5" was to determine this difference, but these simulations failed to converge.
Further simulations demonstrated the effect of doping density in the CdSe nanorods, with
the results shown in Figure 2-4. The CdSe doping density was varied from 3 x 1016 cm-2 7 x
1016 cm-2. The graph shows a linear dependence of Voc on the CdSe doping density. This
dependence is further illustrated in Figure 2-5, along with the linear trendline fit through the
data.
The fit in Figure 2-5 predicts a Voc of 0.181 V for an undoped semiconductor nanorod and
a value of 0.193 V for a doping level of 1 x 1016 cm 2. The curve generated from the input file
"impurity_2" displayed an open-circuit voltage of 0.185 V, which corresponds to a doping
density of 3.3 x 1015 cm-2 according to the trendline in Figure 2-5. However, this linear trend
should not hold for doping levels approaching zero, because zero doping in the CdSe region
should coincide with a Voc of zero, due to the lack of a p-n junction in the device. Any error
from this value could arise from a space-charge region forming between P3HT and one of the
contacts.
From these simulations, the optimal method for defining impurity profiles was determined
to be the use of the "REGION" statement. This is simpler than defining the impurities for
multiple regions as in the "impurity_3" simulation. The "REGION" statement ensures that each
region is properly assigned the appropriate doping density.
Cell dimension adjustments
After definition of the basic model parameters, simulations were performed to determine
the effect of varying cell dimensions. The first parameter adjusted was the cell thickness, which
was accomplished by adjusting the length of the nanorod while maintaining a 10 nm polymer
capping layer. These simulations were performed with the CdSe nanorod width set to 4 nm
PROCESS DEVELOPMENT AND SIMULATION OF HYBRID PHOTOVOLTAIC CELLS
By
MATTHEW L. MONROE
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
2008
40 60 80 100
40 60 80 100
140 16 1
140 160 180
Solvent Boiling Temperature (oC)
1.0 1.5
3.5 4.0
Solvent Polarity
Figure 3-16. RMS surface roughness for 1 x 1 pm surface area samples measured with AFM.
* THF
* Chlorobenzene
o-dichlorobenzene
Chloroform
* Toluene
* o-xylene
* THF
* Chlorobenzene
o-dichlorobenzene
Chloroform
* Toluene
* o-xylene
t
El
I
4
2
0 -
0.5
The approaches to using organic for light conversion can be categorized into three general
classes of cell structures: bi-layer, dye-sensitized, and bulk heterojunction. Similar to inorganic
designs, bi-layer cells use a flat junction created by stacking p- and n-type organic layers, with
additional layers incorporated for charge transport enhancement [4-6]. Dye-sensitized solar cells
use an organic dye adsorbed on inorganic transport materials, typically TiO2, so that the dye
absorbs photons and the inorganic phase allows for efficient charge transport [7-10]. Bulk
heterojunction cells consist of donor and acceptor materials mixed together to form a blended
junction throughout the device active layer.
Organic cells are distinguished from their inorganic counterparts by exciton creation upon
photon absorption. Due to their extremely low exciton binding energy, inorganic p-n junction
cells generate free carriers upon photon absorption, and the carriers are primarily collected by the
field across the depleted junction. Organic materials, on the other hand, primarily generate
excitons that have significant binding energy and transport by diffusion until they recombine or
dissociate at an energetic interface to produce free carriers for eventual collection.
Excitons are efficiently dissociated at a p-n junction in organic devices, although exciton
dissociation occurs to a lesser extent at interfaces with electrodes, polymer chain defects,
absorbed oxygen sites, or active-layer impurities [11]. Because of this dissociation requirement,
only excitons generated within a diffusion length of the junction will contribute to the collected
current. Exciton diffusion lengths are typically on the order of 5 to 20 nm for organic
semiconductors, placing a limit on the thickness of active layers and therefore the photon
absorption extent [12-16]. The blended junction of a bulk heterojunction cell attempts to provide
an interface within an exciton diffusion length throughout the entire active layer, thus allowing
for thicker active layers with better adsorption and more efficient exciton dissociation. Once the
30
E
V)
20-
15
10 -
10 ------------------
0 1 2 3 4 5 6 7
Number of Layers
Figure 3-38. RMS surface roughness for multi-layer hybrid films. Red squares represent
measurements over a 3 x 3 tm area and blue circles represent measurements on a 1 x
1 tm area.
194
8 = 1 cm2/V-s
= 10 cm2N-s
E 6- lp = 100 cm2N-s
S --- p = 200 cm2N-s
E 4 p = 500 cm2N-s
Sip = 800 cm2N-s
S- p = 1000 cm2N-s
S -- = 5000 cm2N-s
S0- L = 10,000 cm2N-s
-2 -
-4 -
-0.1 0.0 0.1 0.2 0.3 0.4 0.5
Voltage (V)
Figure 2-46. Simulated J-V curves for ZnO:P3HT solar cells with varying carrier mobilities and
carrier generation in the 10-nm exciton diffusion length region. In all cases, the
electron mobility is set to 10% of the hole mobility.
polymers, it is several orders of magnitude lower than most inorganic semiconductors and limits
charge collection in devices. Additionally, it degrades quickly under water and oxygen
atmospheres, particularly in the presence of radiation. As new polymers are developed with a
broader absorption spectrum, improved carrier mobility, improved environmental resistance, and
the ability to deposit high-quality films with strong adhesion and low surface roughness they will
quickly find relevance in organic photovoltaics research.
Regardless of the targeted future research direction, some processing equipment changes
should be considered. The recent installation of a photolithographic patterning system will allow
for the fabrication of multiple cells on single substrates, and this will greatly improve the speed
of research and also the quality of the devices fabricated. As long as P3HT is used for the active
layer film, all processing equipment should be set up under an inert atmosphere of argon or
nitrogen in a glove box. Because of the sensitivity of this polymer to water and oxygen, high-
quality device fabrication at this scale requires that it be protected from exposure at all phases of
the process. This is not true for the HTL layer of PEDOT:PSS, which is in fact deposited from a
water solution and shows no negative effects of short-term atmospheric exposure.
All processing steps beyond the deposition of the HTL should be contained in this glove
box, including mixing for the hybrid solution, deposition and drying of the hybrid films, possible
inclusion of exciton blocking layers, back electrode deposition, and cell characterization. This is
a difficult challenge due to the size of some processing equipment, but this is the design used by
groups fabricating world-record organic photovoltaic cells. The processing equipment line
currently in use already involves an evaporation chamber opening directly into the glove box,
which is the largest piece of equipment used in the process. In order to improve these cells to the
highest performance level, this environmental protection is a step that absolutely must be taken.
200
Table 2-17. Cell performance measures for real and simulated solar cells.
Cell Jsc (mA/cm) Voc (V) FF i (%)
Real 2.20 0.44 0.56 0.53
10 nm LD 0.59 0.44 0.84 0.22
20 nm LD 1.01 0.45 0.84 0.38
40 nm LD 1.47 0.45 0.84 0.56
Full Cell 1.27 0.45 0.85 0.48
0.0 -
-0.2 -
E -0.4 -
o
E -0.6 -
c -0.8 -
o -1.0 -
1.2
-1.2 -
-1.4 -
Voltage (V)
Figure 2-60. J-V curves for simulated cells with a 40 nm LD and varying carrier mobility.
Table 2-18. Performance measures for simulated cells with 40 nm LD and varying mobility
values.
Mobility Jsc (mA/cm2) Voc (V) FF i (%)
0.01 x 1.47 0.45 0.78 0.52
0.1 x 1.47 0.45 0.84 0.56
1 x 1.47 0.45 0.84 0.56
10 x 1.47 0.45 0.84 0.56
100 x 1.47 0.45 0.84 0.56
Real Cell 2.20 0.44 0.56 0.53
considered was only between 0 and 20 nm. In a true normal distribution with a mean of 10 nm
and a standard deviation of 4 nm or greater, an appreciable area exists under the curve beyond 0
and 20 nm. To remove this area from consideration, normalization was performed by dividing
by the sum of the area between 0 and 20 to establish these boundaries. The distribution was
calculated as shown in Equation 2-2.
F(x)= ex 2 (2-2)
In Equation 2-2, x is the distance into the P3HT region and x is the mean of the distribution, set
to 10 nm in this case. By dividing this by the sum of all distribution values between x = 0 and x
= 20 nm, the normalized distribution was obtained. The cumulative distribution was calculated
using Equation 2-3.
F(x)
G(x)= 1 =0 (2-3)
x=20
x=0
Again, this cumulative distribution is normalized by the area under the distribution function
between 0 and 20 nm rather than between negative and positive infinity. This cumulative
distribution function is shown in Figure 2-36 for a range of standard deviations and a mean of 10
nm. By coupling these distribution functions with the multi-stage cell design demonstrated in
Figure 2-35, graded absorption can be generated in the simulation by applying a fractional
absorption coefficient in each stage of the absorption region.
Simulations were performed to consider the effect of this absorption distribution in the
cells. These simulations used a 10-stage absorption region with a graded absorption coefficient
in an attempt to more accurately define the absorption in the cell. Absorption profiles
corresponding to standard deviations of 0 to 10 nm were generated by using a fractional value for
O
Figure 3-23 continued. SEM images of P3HT films deposited from pure chloroform and an equal
mixture of chloroform and pyridine solvents. Images are shown at 2,000x and 10,000x
for each film.
-5
rC
for each film.
2,000 x
10,000 x
|
|
PAGE 1
PROCESS DEVELOPMENT AND SIMULATI ON OF HYBRID PHOTOVOLTAIC CELLS By MATTHEW L. MONROE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008 1
PAGE 2
2008 Matthew L. Monroe 2
PAGE 3
To my wife, Kim 3
PAGE 4
ACKNOWLEDGMENTS I acknowledge my loving wife, Kim, for her pa tience, perseverance, and support. She was a source of inspiration, and without her this could not have been possible. I thank my parents for their continued love and support that has guide d me through my personal and academic life. I thank my advisor, Dr. Tim Anderson, for hi s advice and guidance in my research. I would like to thank all of my committee member s for their guidance and direction. Dr. Chinho Park granted me the use of his laboratory and much assistance and advice in organic photovoltaics. Dr. Kirk Ziegler provided labo ratory space and assistance with nanocrystal chemistry. Dr. Oscar Crisalle and Dr. Sheng Li provided valuable guidance through their roles in the interdisciplinary photovoltaics team. I thank Dr. Chinho Parks students at Ye ungnam University (par ticularly Jiyoun Seol, Young Wook Kim, Trong Nguyen Tam Nguyen, a nd Md. Azizul Hasnain) for extensive experimental support and for making a foreign c ountry feel like home. I thank Dr. Woo Kyoung Kim for sharing his expertise in Medici and phot ovoltaics. I thank Dr Zieglers students, particularly Justin Hill and Randy Wang, for assi stance with their laboratory equipment and for many helpful discussions. I thank all of the support staff at the Chemical Engineering department, particularly Sean Poole for helping me set up my simulation work. Special thanks go to Sherrie Jenkins, for performing scheduling miracles on a regular basis. Finally, I would like to thank all of my friends and family that have stood behind me and helped me through my doctoral research and my w hole life up to this point. They have helped mold me into the person I am today, and I am grateful for it. 4
PAGE 5
TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4LIST OF TABLES ...........................................................................................................................7LIST OF FIGURES .........................................................................................................................9ABSTRACT ...................................................................................................................... .............151 INTRODUCTION ................................................................................................................ ..17Introduction .................................................................................................................. ...........17Motivation .......................................................................................................................17Photovoltaic Technologies ..............................................................................................18Organic Photovoltaics .....................................................................................................19Simulation of Hybrid Photovoltaic Devices ....................................................................23Targeted Research ...........................................................................................................242 MEDICI SIMULATIONS OF HYBRID SOLAR CELLS ....................................................27Introduction .................................................................................................................. ...........27Initial Modeling Efforts ..........................................................................................................29Preliminary Model Description .......................................................................................29Definition of model in Medici ..................................................................................30Impurity profile ........................................................................................................32Cell dimension adjustments .....................................................................................34Illumination source ...................................................................................................37Summary ..................................................................................................................38Simulation of a Real Cell ..................................................................................................... ...39Model Parameters ............................................................................................................39Initial Simulations ........................................................................................................... 42Reflectance of Ag electrode .....................................................................................43Mobility ....................................................................................................................45Exciton diffusion length ...........................................................................................46Doping density .........................................................................................................47Density of states .......................................................................................................48Open-circuit voltage examination ............................................................................48Absorption coefficient adjustment ...........................................................................50Multi-stage absorption ..............................................................................................51P3HT replacement with CIS .....................................................................................54Shortcomings of the initial model ............................................................................61Two-Step Simulation Technique ............................................................................................62Photogenerated Carrier Distribution ................................................................................63Line Source Generation ...................................................................................................64 5
PAGE 6
J-V Curves .......................................................................................................................70Summary of Results ........................................................................................................743 ORGANIC AND HYBRID SOLAR CELL PROCESS DEVELOPMENT ........................122Introduction .................................................................................................................. .........122ITO Anode Treatment ...........................................................................................................122Bi-layer Organic Solar Cell Fabrication ...............................................................................126Cell Fabrication Procedure ............................................................................................127Substrate Preparation .....................................................................................................127Spin-Coating .................................................................................................................. 128Evaporation ................................................................................................................... .128Encapsulation ................................................................................................................1 28Film Drying ................................................................................................................... 129PEDOT:PSS ..................................................................................................................130P3HT ............................................................................................................................ ..133Bi-layer Cell Fabrication ...............................................................................................133Solvent Comparisons ............................................................................................................135Hybrid Bulk Heterojunction Cell Fabrication ......................................................................138Nanocrystal Synthesis and Surfactant ...........................................................................139Hybrid Films .................................................................................................................. 140TOPO-coated CdSe ................................................................................................140P3HT solubility in chloroform and pyridine ...........................................................142Cell Fabrication .............................................................................................................144Cells deposited from chloroform solutions ............................................................144Thicker-film cells in chloroform solution ..............................................................145Chlorobenzene solvent ...........................................................................................149Commercial CdSe nanocrystals .............................................................................150Hybrid cell performance summary .........................................................................150Particle Induced Nanostructuring .........................................................................................1524 CONCLUSIONS AND FUTURE WORK ...........................................................................195Conclusions ...........................................................................................................................195Hybrid Photovoltaic Simulation ....................................................................................195Anode Surface Treatment ..............................................................................................196Solvent Selection ...........................................................................................................196Hybrid Bulk Heterojunction Photovoltaic Development ..............................................197Future Work ..........................................................................................................................198Organic Photovoltaic Simulations .................................................................................198Development of Hybrid Photovoltaic Cells ..................................................................199LIST OF REFERENCES .............................................................................................................201BIOGRAPHICAL SKETCH .......................................................................................................206 6
PAGE 7
LIST OF TABLES Table page 2-1. P3HT Properties for Device Simulations. ..............................................................................762-2. CdSe Properties for Device Simulations. ..............................................................................762-3. Electrode Properties for Device Simulations. ........................................................................762-4. Impurity profile inputs for simulations. ................................................................................ .792-5. Solar cell performance measures for unit ce lls of 12 nm width and varying thickness. .......812-6. Performance measures for simulated hybrid cells with varying CdSe half-width. ................832-7. Cell performance measures from publis hed and digitally converted J-V curves. .................852-8. ZnO properties used for hybrid solar cell simulation. ...........................................................862-9. Performance measures for simulated cells with varying carrier mobility. ............................902-10. Performance measures for simulated cells with varying exciton diffusion lengths. ...........922-11. Absorption data and short-circuit cu rrent for graded absorption simulations. ....................992-12. Materials properties for P3HT and CIS used in cell simulations. ......................................1012-13. Performance measures for ZnO:CIS cells with an individual material property set at the P3HT value. ................................................................................................................1012-14. Performance measures for simula ted ZnO:CIS solar cells with the P3HT absorption spectrum. ..................................................................................................................... .....1022-15. Performance measures for simu lated ZnO:CIS solar cells with P3HT values for absorption coefficient and electron affinity. ....................................................................1032-16. Performance measures for simu lated ZnO:CIS solar cells with P3HT values for absorption coefficient, electron affinity, and energy band gap. .......................................1042-17. Cell performance measures for real and simulated solar cells. .........................................1182-18. Performance measures for simulated cells with 40 nm LD and varying mobility values. ..............................................................................................................................1182-19. Generated carriers and performan ce measures for simulated solar cells. ..........................1213-1. Chemical composition of ITO films ....................................................................................155 7
PAGE 8
3-2. Performance of organic solar cells on treated ITO substrates .............................................1553-3. RMS surface roughness of P3HT films shown in Figure 3-11. ...........................................1633-4. J-V data for bi-layer solar cells. ...........................................................................................1663-5. Solvents considered for hybrid bulk heterojunction film deposition. .................................1663-6. Mean rms surface roughness in nm for hybrid films deposited from selected solvents ......1703-7. Hybrid solutions of P3HT and TOPO-coated CdSe nanocrystals in a mixed solvent of chloroform and pyridine. .................................................................................................1763-8. Film properties for P3HT films deposited from various solvents. .......................................1813-9. Hybrid solar cell fabrication information and performance data. .......................................190 8
PAGE 9
LIST OF FIGURES Figure page 1-1. Worldwide cumulative installed PV Power in Megawatts from 1992 to 2006. ....................261-2. Common organic materials used in solar cell development. .................................................262-1. Hybrid solar cell and correspondi ng unit cell used for device simulation. ...........................762-2. Wavelength-dependent absorption coe fficient data used in simulations. ..............................772-3. J-V curves for simulated hybrid solar cell s with different methods of specifying doping density.. ..................................................................................................................... .........772-4. Simulated J-V curves showing the effect of doping density in the CdSe nanorods.. ............782-5. Variation of open circuit voltage with doping density of CdSe nanorods. ............................792-6. Simulated J-V curves with varying unit cel l thickness. .........................................................802-7. Solar cell parameters for unit 12 nm wi de unit cells with varying cell thickness. ................802-8. J-V curves for hybrid cells with varying nanorod width. ......................................................812-9. J-V curves for simulated hybrid cells w ith CdSe nanorod half-thickness between 10 and 25 nm. .................................................................................................................... ......822-10. Solar cell performance measures for simu lated hybrid cells with varying CdSe halfwidth. ........................................................................................................................ .........822-11. J-V curves for hybrid solar cells with different light sour ce specifications. .......................832-12. Illustration of the P HOTOGEN command in Medici. .........................................................842-13. SEM images of ZnO nanofibers and nanofiber and P3HT composite films. .......................842-14. J-V curve for a real ZnO:P3HT solar cell to be used for verification of Medici simulations. .................................................................................................................. ......852-15. J-V and P-V curves for the real so lar cell fabricated by Olson et al. ..................................852-16. Unit cell used for simulations of ZnO/P3HT hybrid cells.. .................................................862-17. Medici unit cell used for device simulation. ........................................................................872-18. Simulated J-V curves for ZnO:P3HT solar cell using two P3HT regions. ...........................88 9
PAGE 10
2-19. J-V curves for simulated cells with zer o absorption in the P3HT2 region and varying reflectance from the Ag electrode. .....................................................................................892-20. Simulation results showing the effect of changing charge mobilities in the P3HT regions. ...................................................................................................................... .........902-21. J-V curves for simulated cells with varying exciton diffusion length. ................................912-22. Extrapolations to estimate VOC for simulated cells with varying exciton diffusion length..................................................................................................................................912-23. Solar cell performance measures for si mulated cells with varying exciton diffusion lengths. ...................................................................................................................... .........922-24. Real and simulated J-V curves for cells with varying P3HT doping density. .....................932-25. Real and simulated J-V curves for hybr id cells with varying ZnO doping density. ............932-26. Real and simulated J-V curves for hybrid solar cells with varying P3HT density of states. ....................................................................................................................... ...........942-27. Simulated J-V curves for hybr id solar cells with varying P3HT mobility. ..........................942-28. Simulated J-V curves fo r hybrid cells with varying P3HT doping concentrations. .............952-29. Energy band diagram for P3HT ZnO hybrid solar cells. ..................................................952-30. J-V curves for simulated cells with varying energy band gap in the active layers. .............962-31. Absorption coefficient vs. wavelength as tabulated in Medici. ...........................................962-32. AM1.5 solar spectrum. .................................................................................................. ......972-33. Carrier generation in simulated solar ce lls plotted with absorption coefficients for P3HT and ZnO. ...................................................................................................................972-34. J-V curves for simulated cells showing the original P3HT absorption profile and an edited absorption profile limiting absorption between 0.2 and 0.3 m. ............................982-35. J-V curves for simulated cells with ex citon diffusion length of 10 nm and multi-stage absorption regions. .............................................................................................................982-36. Examples of cumulative distribution f unction with mean of 10 nm and a range of standard deviation. .............................................................................................................992-37. Simulated J-V curves for cells with graded absorption profiles. .......................................1002-38. Simulated J-V curves with CIS replacing P3HT. ...............................................................100 10
PAGE 11
2-39. Simulated J-V curves for ZnO:CIS solar cells with an indivi dual material property changed to the P3HT value. ..............................................................................................1012-40. Simulated J-V curves for ZnO:CIS solar cells with the P3HT absorption spectrum applied. ...................................................................................................................... .......1022-41. Simulated ZnO:CIS solar cells with P3HT values for absorption coefficient and electron affinity. ...............................................................................................................1032-42. Simulated ZnO:CIS solar cells with P3HT values for absorption coefficient, electron affinity, and energy band gap. ..........................................................................................1042-43. Simulated J-V curves for ZnO:P3HT solar cells with varying carrier mobility. ...............1052-44. Calculation method for estimated second derivatives of J-V curves. ................................1052-45. Estimated second derivative Jest for simulated ZnO:P3HT solar cells with varying carrier mobilities and carrier generation in the full P3HT region. ...................................1062-46. Simulated J-V curves for ZnO:P3HT solar cells with varying carrier mobilities and carrier generation in the 10-nm exciton diffusion length region. ....................................1072-47. Estimated second derivative Jest for simulated ZnO:P3HT solar cells with varying carrier mobilities and carrier generation in the 10-nm exciton diffusion length region. .1082-48. Simulated J-V curves for ZnO:P3HT solar cells with varying absorption coefficients in P3HT.............................................................................................................................1092-49. Simulated J-V curves for ZnO:P3HT solar cells with varying energy band gap. ..............1092-50. Simulated J-V curves for ZnO:P3HT solar cells with varying band gap and absorption in the P3HT region. ..........................................................................................................1102-51. Simulated J-V curves for ZnO:P3HT solar cells with P3HT energy band gap of 1.0 eV and hole mobility of 500 cm2/Vs. ...................................................................................1102-52. Photogenerated carrier distribution in pairs/cm3 for the full unit ce ll and the region of 13 nm x 27 nm along the edge of the ZnO nanorod. .................................................1112-53. Carrier mapping scheme for two-stage simulations. .........................................................1122-55. Cumulative number of photogenerated carrie rs in simulated cells with varying exciton diffusion length. ...............................................................................................................1142-56. Photogenerated carrier distribution along xand ycoordinates for models with varying exciton diffusion lengths.....................................................................................1152-57. Contour plots of the photogenerated car rier difference between the full absorption simulation and the 40 nm LD simulation. ........................................................................116 11
PAGE 12
2-58. Photogenerated carriers at the tip and base corner poin ts of the nanorod in simulated hybrid cells. ......................................................................................................................1172-59. J-V curves for simulated solar cells using line-source car rier generation. ........................1172-60. J-V curves for simulated cells with a 40 nm LD and varying carrier mobility. ................1182-61. Photogenerated carrier distribution for a simulated unit cell with LD = 40 nm and LD = 40 with the P3HT absorption coefficient increased by 50%. ........................................1192-62. Difference in photogenerated carriers between 150% and 100% P3HT absorption coefficients in simulated cells with 40 nm LD. ...............................................................1202-63. J-V curves for simulated cells with varying absorption coefficient in P3HT. ...................1213-1. J-V curves for organic solar cells on treated ITO substrates ...............................................1553-2. Energy band diagram for bi -layer organic solar cells. .........................................................1563-3. Steps for bi-layer solar cell fabrication. ............................................................................... .1563-4. PEDOT:PSS film thickness vs. spin-coater speed. ...............................................................1573-5. P3HT film thickness vs. spin-coating speed. ........................................................................1573-6. FTIR spectra for P3HT film and solution. ...........................................................................1583-7. FTIR spectra for PEDOT:PSS film and solution. ................................................................1593-8. J-V curves for bi-layer or ganic solar cells fabricated with a single 80 nm thick layer of PEDOT:PSS. .................................................................................................................... 1603-9. J-V curves for bi-layer organic solar cells fabricated with two 40 nm thick layers of PEDOT:PSS. .................................................................................................................... 1613-10. Calibration curves for slow and fast-filtered PEDOT:PSS. .............................................1623-11. AFM images of P3HT films. ..............................................................................................1623-12. J-V curves for bi-layer solar cells fabricated on untreated ITO substrates. ......................1633-13. J-V curves in the dark and under 100 mW/cm2 illumination. ...........................................1643-14. Surface roughness measurements by profilometry for hybrid films deposited from various solvents. ............................................................................................................. ..1673-15. RMS surface roughness for 5 x 5 m surface area samples measured with AFM. ...........1683-16. RMS surface roughness for 1 x 1 m surface area samples measured with AFM. ...........169 12
PAGE 13
3-17. Optical microscope imag es of selected films. ...................................................................1703-18. SEM images of selected hybrid films. ...............................................................................1733-19. Film surface roughness vs. pyridine con centration in the chloroform solvent for TOPO-coated and pyridine-coated CdSe nanocrystals. ...................................................1753-20. Optical microscope images of hybrid films deposited from 20.4 mg/ml, 19 mg/ml and 5 mg/ml solutions. ............................................................................................................ 1763-21. Optical microscope im age of 19 mg/ml hybrid film deposited at 3000 rpm and subjected to a pure solvent spin -coating step at 8000 rpm. .............................................1773-22. Optical microscope images of P3HT films deposited from 5 mg/ml solutions in chloroform, 1:1 chloroform: pyridine, and pyridine. ........................................................1773-24. Surface profiles of P3HT films deposited from chloro form, 1:1 chloroform:pyridine, and pyridine solvents. ......................................................................................................17 93-25. Dark and illuminated J-V curves fo r cells generated from 5 mg/ml composite solutions in chloroform mixed with 2% pyridine and pure chloroform. .........................1813-26. Dark and illuminated J-V curves for 10 mg/ml hybrid solution deposited with lowspeed spin-coating. ...........................................................................................................1823-27. Surface profiles of film deposited from 12 mg/ml P3HT and 25 mg/ml hybrid solutions in 2% pyridine in chloroform. ..........................................................................1833-28. Dark and illuminated J-V curves for hybr id bulk heterojunction solar cell deposited from 25 mg/ml solution and P3HT polymer cell deposited from 12 mg/ml solution. .....1843-29. Short-circuit current decay for hybrid and P3HT. ..............................................................1853-30. Dark and illuminated J-V curves for hybrid bulk heterojunction and pure P3HT solar cells with limited air exposure during processing. ...........................................................1863-31. Dark and illuminated J-V curves for hybrid solar cell fabricated from chlorobenzene with 2% pyridine solution. ...............................................................................................1873-32. Dark and illuminated J-V curves for a hybrid solar cell fabricat ed with commercial CdSe nanopowder. ...........................................................................................................1883-33. Illuminated J-V curves for hybrid bul k heterojunction solar cells with various fabrication conditions.......................................................................................................1893-34. J-V curves for the best bi-layer and hybrid cells shown in this dissertation. ....................1893-35. Film thickness for multi-layer hybrid films. ......................................................................191 13
PAGE 14
3-36. SEM and AFM surface images of multi-layer hybrid films. .............................................1913-37. Optical microscope images for multi-layer hybrid films. ..................................................1933-38. RMS surface roughness for multi-layer hybrid films. .......................................................194 14
PAGE 15
Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy PROCESS DEVELOPMENT AND SIMULATI ON OF HYBRID PHOTOVOLTAIC CELLS By Matthew L. Monroe December 2008 Chair: Timothy J. Anderson Major: Chemical Engineering The fabrication and simulation of hybrid bulk heterojunction photovoltaic cells was studied to develop an understanding of these devices an d facilitate improvements in device fabrication techniques. The device simulation software package Medici was app lied to this type of device for the first time, and novel device fabricati on techniques were developed to improve device performance. Medici was adapted for a hybrid system cons isting of an ordered array of inorganic nanorods interspersed with an absorbing, se miconducting polymer. A cell taken from the literature was simulated using a two-stage simu lation technique which in dependently calculated the light absorption profiles and the cell performa nce. This technique ap plied a line source of carriers directly at the inte rface between the inorganic and organic regions of the cell, realistically imitating the exciton dissociation phy sics of real hybrid cells. The simulations showed low current densities as compared to th e real cell, due to str ong recombination at the material interface. Nitrogen plasma treatment of the transparen t anode in bi-layer and hybrid photovoltaic cells was found to reduce surface roughness, increas e the hydrophilic nature of the film, and improve cell performance. A range of solvents were tested for hybrid bulk heterojunction film 15
PAGE 16
16 deposition, with chloroform, chlo robenzene, and o-dichlorobenzen e found to provide films with low surface roughness and strong uniformity. Hybr id bulk heterojunction solar cells were fabricated, and these cells showed low performance of less than 1% efficien cy. The active layers degraded upon exposure to air, re sulting in a drop in short-circu it current density that was more pronounced for hybrid films than for pure polymer single-layer cells. These findings highlight the environmental sensitivity of these devices and the need for an inert environment for cell fabrication and testing.
PAGE 17
CHAPTER 1 INTRODUCTION Introduction Since the discovery of the photovoltaic effect and the design of the first functional solar cell, photovoltaic technology has consistently developed to become an increasingly viable energy source. Photovoltaics has developed into many classes of devices and materials systems. Photovoltaic technology can provide clea n, efficient, and portable energy. Motivation There is currently a strong push toward alte rnative energy sources as the price of oil increases and nations worldwide work to slow the emission of greenhouse gases such as COx and NOx. Many countries have established conservation and alternative energy programs in attempts to control the output of these gases. Recent in creases in oil and gas prices and controversy surrounding global warming have driven public rec ognition of the need for alternative renewable energy sources. While small-scale steps such as hybrid cars stand to relieve a small amount of the worlds fossil fuel consumption, new tec hnologies such as photovoltaic energy must be developed to fulfill the worlds large-scale energy needs. Of the available candidates for alterna tive large-scale ener gy production, photovoltaic energy conversion has many qualities that make it a leading technology, including: Photovoltaic technology has been studied intensely for many year s, initially driven by its use in the space program to provide energy fo r satellites and space vehicles, to prepare it for commercialization. Energy generated from solar cells can be ge nerated locally, resulting in reduced energy distribution costs and a more reactive system. Because of its local power generation, photovol taics are optimal for power generation in remote locations where it is difficult or impossible to connect to a local grid. 17
PAGE 18
Solar energy is plentiful. Americas energy needs could be supplied by a single solar array covering an area of 100 x 100 miles in the uninhabited deserts of the western United States. Photovoltaic sources provide peak energy during peak consumption times. Energy consumption is at its maximum during the mi ddle of the day, requiring power plants to work harder to provide the necessary energy to their consumers. Because the suns rays on earth are also at a maximu m during that time, solar en ergy provides maximum energy at the most critical portion of the day. Worldwide photovoltaic installations have begun to grow rapidly, as shown in Figure 1-1 [1]. The U.S., who at one time was the world le ader in photovoltaics, ha s fallen behind countries such as Japan and Germany who have made str ong efforts to boost their solar energy programs. However, recent initiatives such as the Mil lion Solar Roofs Initiative and Solar America Initiative, along with federal and state sponsor ed cost-sharing program s for residential and commercial solar installations, have created a ne w boom in American solar installations [2]. Photovoltaic Technologies Many photovoltaic technologies are currently in the develo pmental or production stages, including crystalline and polycrystalline si licon, thin film, concen trator arrays, space photovoltaics, and organic cells. Polycrystalline silicon is currently the dominant technology in the photovoltaics market, drawing on years of silicon processing technologies to provide a low production cost. Polycrystalline silicon modules area commercia lly available for as low as $4.29/Watt in April 2008, approximately $0.50/Watt ch eaper than the U.S. average cost of $4.81/Watt [3]. These panels are ea sily identified by thei r deep blue color and are currently in production worldwide. The next wave of phot ovoltaic production app ears to be thin-film photovoltaics. These cells employ direct-bandgap materials with strong absorption coefficients such as amorphous silicon (a-Si), cadmium te lluride (CdTe), and copper indium-gallium diselenide (CuInxGa1-xSe2). Many of these cells have demonstrated extremely promising 18
PAGE 19
laboratory efficiencies, and the thin layers reduce material costs and provide the ability to create flexible cells for a wider range of applications. Concentrator photovoltaic arrays generate a large amount of power from a single cell by using an array of mirrors to focus the suns ener gy on the cell. This tech nology has been used to reduce large-scale production costs because the cost of a mirror is lower than the cost of a photovoltaic panel. Space solar cells are the most advanced cells, using expensive growth methods to produce high quality photovoltaic material. These cells provide extremely high performance and durability through their multi-ju nction stack technology, but this comes at a high cost that makes the cells too expensive fo r standard terrestrial applications. Under concentration, however, their high efficiency (a pproaching 40%) mitigates their higher cost. Organic and hybrid solar cells provide extr emely lightweight devices and low-cost manufacturing, but currently th eir low performance and life time limits their applications. Organic Photovoltaics Compared to their inorganic counterparts, the development of organic PV is in its infancy. Drawing on advances in organic light emitting diodes, however, considerable progress has been achieved that encourag es continued exploration of organic solar cells. Organic photovoltaic technology uses organi c molecules or polymers to absorb sunlight and generate photocurrent. Many of these materials have extr emely high absorption coefficients with maxima in the visible region of the spectrum ( 105 cm-1), so an extremely th in layer (hundreds of nanometers) is sufficient for absorbing incident light. Because the films are extremely thin and flexibile as compared to inorga nic crystals, organic photovoltaics show promise for flexible and lightweight portable devices. Many organic materials can be rapidly deposited through inexpensive techniques such as thermal evapora tion at moderate vacuum, or by spin-coating, dipcoating, screen printing, ink jet printing, and spray coating at room temperature and pressure. 19
PAGE 20
The approaches to using organi cs for light conversion can be categorized into three general classes of cell structures: bi-lay er, dye-sensitized, and bulk hetero junction. Similar to inorganic designs, bi-layer cells use a flat junction create d by stacking pand n-type organic layers, with additional layers incorporated for charge transport enhancement [4-6 ]. Dye-sensitized solar cells use an organic dye adsorbed on inorgani c transport materials, typically TiO2, so that the dye absorbs photons and the inorganic phase allows fo r efficient charge transport [7-10]. Bulk heterojunction cells consist of donor and acceptor materials mixed together to form a blended junction throughout the device active layer. Organic cells are distinguishe d from their inorganic count erparts by exciton creation upon photon absorption. Due to their extremely low exciton binding energy, inorganic p-n junction cells generate free carriers upon photon absorption, and the carriers are primarily collected by the field across the depleted junction. Organic ma terials, on the other hand, primarily generate excitons that have significant binding energy and transport by diffusion until they recombine or dissociate at an energetic interface to produ ce free carriers for eventual collection. Excitons are efficiently dissoci ated at a p-n junction in or ganic devices, although exciton dissociation occurs to a lesser extent at inte rfaces with electrodes, polymer chain defects, absorbed oxygen sites, or active-layer impurities [ 11]. Because of this dissociation requirement, only excitons generated within a diffusion length of the junction will contribute to the collected current. Exciton diffusion lengt hs are typically on the order of 5 to 20 nm for organic semiconductors, placing a limit on the thickness of active la yers and therefore the photon absorption extent [12-16]. The blended junction of a bulk heterojunction ce ll attempts to provide an interface within an exciton diffusion length throughout the entire active layer, thus allowing for thicker active layers with better adsorption an d more efficient exciton dissociation. Once the 20
PAGE 21
free carriers are generated they must be collected at their re spective electrodes before they recombine, which occurs to some extent in th e bulk or more prevalently at interfaces. Bulk heterojunction devices have been fabricat ed by blending several cl asses of materials, including multiple organic small molecules, polymer and organic molecules, polymer and carbon nanotubes, polymer and inorganic nanoparticles, and polymers deposited in a prefabricated inorganic nanostructure. Heterojunctions base d on small organic molecules such as copper phthalocyanine (CuPc) receive li ttle attention compared to polymer-based heterojunctions, but have demonstrated reasonable efficiency using a variety of depositi on techniques [17-19]. Carbon nanotubes have very recently received consid eration as a solar cell material, both as an ntype conductor in a heterojunction [20] and as a structured electrode [21]. Bulk heterojunctions fabricated from conjugated polymers and C60 represent the most widely-studied class of bulk heterojunction solar cell. Sin ce the discovery of ultra-fast photoinduced charge transfer from conducting polymers to C60 [22], polymer-C60 bi-layer and bulk heterojunctions have been studied exte nsively. The first cells using semiconducting polymers and C60 were fabricated by the same group [23] Early bulk heterojunction cells using C60 employed poly (pheneylene vinylene) (PPV) deri vatives as the polymer material, and issues centered on co-dispersion of the two materials to create a wellblended structure [24]. This issue was solved with the synthesis a nd application of [6, 6]-phenyl C61 butyric acid methyl ester (PCBM), a highly-soluble C60 derivative, [25, 26], and led to a record efficiency of 2.5% in 2000 [27]. Since this development, other fullerene derivatives have been s ynthesized and evaluated, but PCBM remains the most widely used [28, 29]. The most popular PPV-based polymers are poly [2-methoxy,5-( 2-ethylhexyloxy) -1,4-phe nylene-vinylene] (MEH-PPV) and poly [2methyl,5-(3*,7** dimethyloctyl oxy)]-p-phenylene vinylene (M DMO-PPV), although other 21
PAGE 22
derivatives have been demonstrated [30]. These cells frequently feature a hole transporting layer of poly (3, 4-ethylenedioxythi ophene):poly (styrenesulfonate) (PEDOT:PSS) to improve hole collection. The structures of PCBM and some polymers commonly used for bulk heterojunction solar cells are shown in Figure 1-2. In particular, polythiophene derivatives are now popular alternatives to PPV, with poly (3hexylthiophene) (P3HT) being the most commonly used [29, 31]. P3HT is synthesized with a regio-regular configuration wher e the polymer side chains alte rnate on opposite sides of the backbone chain. This arrangement aids in ali gning the polymer chains for efficient charge transfer along the backbone, with additional chai n straightening attributed to steric hindrance from the fullerene molecules [32]. Bulk heterojunction cells fabricated from P3HT and PCBM have reached efficiencies of 3.5% in 2003 [33] and 4.4% in 2005 [34]. The use of inorganic nanocrystals as a replacement for C60 is a relatively recent trend. In addition to their high electron mobilities, se miconductor nanocrystals can contribute to absorption and photocurrent generation in the hete rojunction active layer. These nanocrystals offer the possibility of bandgap engineering by mate rial selection and tightly controlling the size distribution, as well as the growth of different crystal shapes such as rods and tetrapods [35-37] to create more efficient charge transport pathways. As with early work involving C60, dispersion of the nanocrystals remains an issue in pro cess development. One approach to enhance nanocrystal dispersion is the use of sol-gel proc essing to grow nanocrystals in the polymer film [18, 38-40]. The nanocrystals are typically grown in solution, a nd are formed with a surfactant capping layer to relieve the high surface energy [41, 42]. Exchange of this surfactant has been demonstrated to obtain improved solubility and electrical properties [43-45]. 22
PAGE 23
Ordered heterojunctions have been fabricat ed from a variety of materials including mesoporous TiO2 [46, 47], but ZnO nanostructures have b een the most popular. Regular arrays of well-aligned ZnO nanowires have been gr own using a variety of techniques including MOCVD [48, 49], evaporation [50, 51], and soluti on-based thermal growth [52]. Tak et al. demonstrated selective MOCVD growth of Zn O nanoneedles on a patterned buffer layer [53], allowing for substrate patterning with nanomateria l coverage. Low-temperature growth has been achieved using a sol-gel precursor method [ 54]. Device performance of these ordered heterojunction devices has been limited by rod sp acings several times larger than the exciton diffusion length of the active polymer. Simulation of Hybrid Photovoltaic Devices The incorporation of excitons in device phys ics models has been done both inside and outside the organic photovoltaics world. Models of silicon solar cells show that the inclusion of excitons causes a decrease in dark current but an increase in photocurre nt [55], while further studies showed that this eff ect is only substantial when the exciton diffusion length is significantly greater than the electron diffusion le ngth [56]. This work was further expanded by Burgelman and Minnaert to include exciton diss ociation at surfaces, showing that even purely excitonic devices such as organic solar cells ca n be effective as long as the rate of exciton dissociation is high [57]. Simulations of polymer-inorganic hybrid ce lls have been performed with various assumptions. Studies have proven that effectiv e charge transport only takes place in these devices when the dimensions of the polymer regi on are less than or equal to the exciton diffusion length [58]. Additional m odels have validated this effect in polymer-C60 bulk heterojunction cells [59], highlighting the importance of excitons in these types of cells. The traditional circuit 23
PAGE 24
model of a solar cell has been modified to includ e an additional rectificat ion diode to include the effect of exciton recombination [60]. While work exists in development of models to describe organic phot ovoltaics, little work has been done in applying existing semiconductor modeling programs to simulate these cells. Recently, Takshi et. al. applied Medici to simulate an organic transistor using P3HT as the semiconductor [61]. Medici is a 2-dimensiona l device simulation program developed by Avant! Corporation. It is designed for the simulati on of MOS and bipolar tr ansistors, and models potential and carrier concentrations in a devi ce by solving Poissons equation and the electron and hole continuity equations. Targeted Research The work presented in this dissertation provides a study of hybrid photovoltaic devices from an experimental and theoretic al perspective. The device si mulation program Medici is used for the first time to provide simulations of an ordered bulk heteroj unction photovoltaic device with an array of inorganic nanorods interspersed with a semi conducting organic polymer. This design was chosen due to its semi-regular structur e that allows the device to be broken into a representative unit cell, providing greater detail in the simulations. By modeling the cell in this way, effective values of key parameters such as the carrier mobility, exciton diffusion length, and energy gap of the interface can be esti mated for the device during operation. This simulation supplements experimental work focusing on process development for a polymer-nanocrystal bulk heterojunction solar cell. The issues of charge transport and exciton dissociation are targeted. Charge transport is improved through surface treatment of the ITO electrode prior to active laye r deposition. This generates a smoother surface and promotes adhesion of subsequent layers Exciton dissociation is a ddressed through control of the morphology of the bulk heterojunction active layer. This is achieved through surface exchange 24
PAGE 25
of the nanocrystal surfactant, se lection of an appropriate solv ent for film deposition, and the introduction of a novel layer-by-layer deposition process. 25
PAGE 26
26 Figure 1-1. Worldwide cumulative installed PV Power in Megawatts from 1992 to 2006 [1]. Figure 1-2. Common organic mate rials used in solar cell de velopment regioregular P3HT (a), PCBM (b), MDMO-PPV (c), MEH-PPV (d), and PEDOT:PSS (e). S S O O OSO3H S O O * O O a) b) O c) e) d)
PAGE 27
CHAPTER 2 MEDICI SIMULATIONS OF HYBRID SOLAR CELLS Introduction Device modeling has lagged behind cell fabri cation for organic and hybrid photovoltaics. Cell performance has been characterized using e quivalent circuit theories [60, 62], and some work has been performed to model certain para meters such as cell lif etimes [63], charge recombination [64], and short-circuit current [59] The use of existing simulation programs to provide a fully-encompassed view of the device performance has not been attempted to this point. Organic photoabsorbing materials differ fundame ntally from most i norganic crystals in that the absorption of a photon generates an exciton, or bound electron-hole pair, with high binding energy. Excitons can be generated in inor ganic materials as well, but this occurs only over a very limited wavelength and temperature range and the resulting excitons exist with a low binding energy on the order of k T at room temperature [57]. Th is low binding energy causes the exciton to be highly unstable in th e presence of elevated temperatur es or strong fields, such as in the space-charge region of a p-n junction. In organic materials, excitons are generated nearly exclusively and exist with a bi nding energy or approximately 10x th at of the inorganic excitons, or ~ 300 meV. These highly stable excitons diss ociate at an interface with another material, where one of the carriers is transferred into th e neighboring material. In organic photovoltaic cells, the electron is typically transferred into an n-type material due to the relatively low electron mobility in most organic materials. Because the organic exciton exists as an e ssentially uncharged particle unaffected by fields, it travels through the so lar cell through diffusion, with the exciton diffusion length (LD) representing the maximum distan ce it can travel before self-a nnihilation. For many organic 27
PAGE 28
photoabsorbers, including P3HT, LD has been measured to be on the order of 10 nm [15]. This restricts the current in organic photovoltaics because only excitons generated within LD of an electron acceptor can generate free carriers, and any photons abso rbed by the organic absorber outside of this distance are lost to recombination. This inhere nt limitation is the motivation behind the bulk heterojunction cell design, in wh ich the organic absorber and electron acceptor are blended on a scale that approximates LD. The hybrid bulk heterojunction desi gn considered for simulations in this dissertation is an ordered array of inorganic nanorods with an organic absorbing pol ymer interspersed between and on top of the rods. This inorganic array is a ssumed to be uniform in terms of rod length, rod width, and rod spacing, and the polymer is assume d to fully occupy vacant areas between the rods. The simulation program chosen for this effort was Medici [65]. Medici is a device modeling software package from Synopsis designed to simulate diodes, MOS transistors, and bipolar transistors, as well as emissive devices. The program is provides a two-dimensional simulation area and solves Po issons equation and the electron and hole continuity equations to generate two-dimens ional distributions of potential and carrier concentrations [65]. Medici is designed for the simulation of crystalline inorganic materials, but has been used in certain rare instances for the modeling of organic semiconductors. Recently, Takshi et al. used Medici to simulate a dual gate organic transistor using P3HT as the organic semiconductor [61]. The simulation applied literature values for P3HT properties and measured the performance of singleand dual-gate OFET de signs with varying film thickness. No effort was made by the authors to account for exciton d ynamics in the device because the polymer was functioning as a transistor rather than a solar absorber, so the influence of excitons would be minimal. 28
PAGE 29
The semiconductor properties that are nece ssary for a proper simulation using Medici include the electron and hole mobility, band gap en ergy, permittivity, electron affinity, density of states in the conduction and valence band, and doping density. These properties are well-known for most common semiconductor materials, but are less well-characterized for many organic materials. Despite this, measurements and assu mptions regarding the necessary properties for P3HT can be found in the literature, and these va lues were used as starting points for device simulation. Initial Modeling Efforts A dispersion of P3HT in an ordered array of CdSe nanor ods was chosen as a first attempt at hybrid solar cell modeling using Medici. The choice of materials and dimensions were somewhat arbitrary, but represent a pseudo-re alistic scenario with which to expore the capabilities of Medici for hybrid cell modeling. This initial modeli ng effort serves as a precursor to attempts to simulate a real cell from literature. Preliminary Model Description Hybrid solar cells consisting of an ordered array of CdSe nanorods dispersed in a P3HT matrix were simulated using the device modeling software package Medici. The nanorods were assumed to be vertically aligned, be in contact with the aluminum back electrode, have uniform dimensions, and have uniform spacing. An illustra tion of this design is shown in Figure 2-1. Due to the symmetry imposed in this structure, the cell can be divided in to a basic repeating unit cell, as illustrated in the figure. The active layer is fixed at 100 nm thick, while the CdSe nanorod has dimensions of 90 nm x 10 nm. The rod spacing is fixed at 20 nm. These dimensions define a P3HT capping layer of 10 nm over the tip of each of the CdSe nanorods. Additionally, the unit cell ha s a polymer thickness of 10 nm to the side of the nanorod, due to the 20 nm rod spacing imposed in the model. This is chosen because it is equal to the approximate 29
PAGE 30
exciton diffusion length in P3HT [15]. The unit cell used for simulation has dimensions of 100 nm thick (the thickness of the device, not includ ing electrodes) by 15 nm wide (half of the 10 nm rod width plus half of the 20 nm rod spacing). The top contact is indium tin oxide (ITO), and the bottom contact is aluminum. The unit cell occupies 0.1 m in the y direction and 0.015 m in the x direction. The CdSe nanorod occupies an area from 0.01 y 0.1 m and 0 x 0.005 m. P3HT occupies the other regions, for y < 0.01 m and x > 0.005 m. The materials properties assigned for the simula tion were taken from the literature and are displayed in Tables 2-1 2-3. Material properties for CdSe were taken from previous simulations from Dr. Woo Kyoung Kim [66]. Figur e 2-2 shows the absorp tion coefficient data used for several different materials in simulatio ns, with the sources of the data noted in the caption. Note that the electrical band gap and op tical band gap can be i ndividually specified in MEDICI, but they are assumed to be equal in all simulations unless otherwise noted. The doping density was calculated to depe nd on the amount of time that P3HT is exposed to air, but falls in the narrow range of 1 x 1016 to 2 x 1016 cm-2 for air exposure times on the order of 100 hr [67]. This value was further extended to 5 x 1016 cm-2 by Takshi et al. by as suming extended exposure times [61], and this value was used as a starting point for simulations. The P3HT hole mobility of 0.1 cm2/V-s is the highest reported mobility for this material [68]. Definition of model in Medici This section gives a brief de scription of the steps require d to perform a simulation in Medici. Medici runs based on use r-defined input files that are easil y edited in text format with any text editing program, typically Notepa d or Wordpad on Windows computers. For photovoltaic simulations, these input files contain the following in formation, typically in this order: 30
PAGE 31
1) Mesh definition 2) Region definitions 3) Electrode definitions 4) Material property definitions 5) Illumination definition 6) J-V calculation 7) Output definition. The simulation mesh is the collection of points where calculations are performed throughout the simulation area. The spacing of the mesh points can be specified by the user and is somewhat arbitrary, with the exception that mesh points sh ould exist at boundaries between regions to facilitate convergen ce of the calculations. Medici allows a maximum of 3,200 mesh points, with more points pr oviding a more well-defined cal culation at the expense of computational time. Regions of the mesh can be defined to correspond to different materials in the simulation. These are specified through ranges of (x, y) coordinates in units of m, with y = 0 corresponding to the top of the simulation area. The electrodes are specified in terms of their optical properties and are not defined by (x, y) coordinates. Calcul ations are performed by Medici at the interface points with the electrodes, but not within the actual electrode. Material properties are specifi ed to correspond to the materi al ranges given in the region definition. An example would be MATERIAL REGION=P3HT PERMITTI=3 NC300=2E18 NV300=2E19 EG300=1.7 + EGO300=1.7 AFFINITY=3.15 AB S.FILE="abs-p3ht.txt" PR.TAB where the properties for the area defined as P3HT are specified. Carrier mobilities and impurity profiles are defined in separate statements. Further details on impurity profile definitions are given in the following section. The absorption coefficient is defined through an external text file, where the value can be sp ecified by the user for a range of wavelengths. 31
PAGE 32
The illumination source can be defined in multiple ways. For absorption models such as photovoltaic cells, radiation can be defined to follow any specified profile and generated from a point source or line source at a ny location outside of the simula tion area. Further details are given in a later section describing the effect of altering the origin of the illumination source. In addition to general illumination, sources of phot ogenerated carriers can be specified along any line within the simulation area. With the defined illumination source, Medici calculates the photogenerated carrier distribution throughout the simula tion area based on the absorption coefficient data given for each material region. The program then begins ite rating over a range of applied biases specified by the user until the current and potential has been mapped and converged throughout the simulation area. Once a solution has been achie ved, Medici provides a vast array of output mechanisms for the user to extract data fr om the simulation, including plot generation and numerical data extraction. Impurity profile The way in which Medici assigns impurity profile s in simulations is a bit of a mystery. A few simulations were performed to determine th e proper way to call this function. The profile can be defined in terms of (x, y) coordinate s or by the REGION statement which applies values to an area defined by a specific name. The impurity profile input statements used are listed in Table 2-4. The file impurity_1 specifies the doping areas by the REGION statement, which generates doping in an area which has previously been defined in coordinate space and assigned a label (P3HT or CdSe). The file impurity_2 fi rst applies p-type doping at a concentration of 5 x 1016 cm-2 to the entire cell area and then applies an n-type doping of 6 x 1016 cm-2 to the area occupied by CdSe. This depends on Medici overwriting the impurity values when multiple 32
PAGE 33
definitions are made in the same region, in this case the CdSe region. In impurity_3 the cell is sectioned into three rectangl es, two which encompass the P3HT region and one which defines the CdSe region. The P3HT region consists of the regions y 0.01 m and (0.01 y 0.1 m, 0.005 x 0.015 m). The CdSe region is defined of the area from (0.01 y 0.1 m, 0 x 0.005 m). The file impurity_4 considers the possibility that MEDICI sums doping specifications rather than overwriti ng them. It takes the same fo rmat as impurity_3 but assigns the n-type doping concentration as 1 x 1016 cm-2 to account for the pos sibility that 5 x 1016 cm-2 is cancelled out by the p-type impurities when th e statements are superimposed. In impurity_5 this concept is again tested by breaking the cell in to the same three sectio ns as in impurity_3, but assigning each section a p-type doping of 5 x 1016 cm-2. Then, the CdSe section is respecified as n-type with a concentration of 6 x 1016 cm-2. The files impurity_4 and impurity_5 failed to conv erge. It is intere sting to note that impurity_5 failed, but impurity_2 did not. Both appear to perform the same actions of filling the entire cell with p-type dop ing at a concentration of 5 x 1016 cm-2, and then applying an n-type doping of 6 x 1016 cm-2 to the CdSe region. The file impurity_4 directly applied an impurity profile of 1 x 1016 cm-2 in the CdSe region. J-V curves from the three successful simulati ons are shown in Figure 2-3. From the graph, it is obvious that impurity_2 was significantly different than impurity_1 or impurity_3. Both impurity_1 and impurity_3 show nearly identical J-V curves, suggesting that the calculation is very similar. It is interesting, however, that th e results were not identical. The curve for file impurity_2 shows an open circuit voltage of 0.185 V, approximately 30% lower than the values of 0.246 and 0.253 V for the other curves. This would seem to be 33
PAGE 34
indicative of a lower effective doping density. The purpose of input files impurity_4 and impurity_5 was to determine this difference, but these simulations failed to converge. Further simulations demonstrat ed the effect of doping density in the CdSe nanorods, with the results shown in Figure 2-4. The Cd Se doping density was varied from 3 x 1016 cm-2 7 x 1016 cm-2. The graph shows a linear dependence of VOC on the CdSe doping density. This dependence is further illustrated in Figure 2-5, along with the linear tre ndline fit through the data. The fit in Figure 2-5 predicts a VOC of 0.181 V for an undoped semiconductor nanorod and a value of 0.193 V for a doping level of 1 x 1016 cm-2. The curve generated from the input file impurity_2 displayed an open-circuit voltage of 0.185 V, which corresponds to a doping density of 3.3 x 1015 cm-2 according to the trendline in Figure 2-5. However, this linear trend should not hold for doping levels approaching zero, because zero doping in the CdSe region should coincide with a VOC of zero, due to the lack of a p-n junction in the device. Any error from this value could arise from a space-charge region forming between P3HT and one of the contacts. From these simulations, the optimal method for defining impurity profiles was determined to be the use of the REGION statement. This is simpler than defining the impurities for multiple regions as in the impurity_3 simulatio n. The REGION statement ensures that each region is properly assigned the appropriate doping density. Cell dimension adjustments After definition of the basic model parameters simulations were performed to determine the effect of varying cell dimensions. The first parameter adjusted was the cell thickness, which was accomplished by adjusting the length of th e nanorod while maintaining a 10 nm polymer capping layer. These simulations were perfor med with the CdSe nanorod width set to 4 nm 34
PAGE 35
rather than 10 nm. This sets the nanorod half-w idth in the unit cell simulation area as 2 nm. The resulting simulated J-V curves are shown in Fi gure 2-6. The length liste d in the legend is the unit cell thickness, not including electrodes, so the CdSe rod lengt h is 10 nm shorter than that distance due to the constant 10 nm capping layer. The cells showed an increase in VOC, JSC, FF, and efficiency as the cell thickness increased due to increased absorption in the devices. Howe ver, due to increasing se ries resistance, there was a diminishing return as the film thickness in creased. Interestingly, th e short-circuit current density continued increasing as the film thickness was increased to 1 m, despite P3HTs low hole mobility of 0.01 cm2/V-s. Figure 2-7 displays solar cell performance meas ures for the simulated J-V curves shown in Figure 2-6, with additional data points that were not displayed in Figure 2-6 for clarity. The short-circuit current density of the unit cells c ontinues to increase with cell thickness, although it begins to level off. The fill factor and VOC of the cells reach a maximu m in this simulation, with the value of the 1 m cell showing slightly lower values th an the 500 nm cell. The peak values are approximately VOC = 60 mV and FF = 0.36. The efficiency continues increasing up to the 1 m cell thickness, but as with the short-circuit curre nt, the rate of increase slows dramatically. A summary of the results from these s imulations is shown in Table 2-5. With the effects of varying cell thickness ch aracterized, simulations were then performed to determine the impact of nanorod width on cel l performance. Similar to the thickness variations displayed prev iously, these simulations maintain a constant P3HT thickness of 10 nm on the top and side of the CdSe nanorod. The resulting J-V curves are shown in Figure 2-8. The values displayed in the legend represent the nano rod half-width rather than the unit cell width. The unit cell width is 10 nm higher than the li sted value due to the constant 10 nm of P3HT on 35
PAGE 36
the side of the nanorod. The results show that performance increases with increasing nanorod width, which is counterintuitive. The absorption coefficient for CdSe is significantly lower than that of P3HT over most wavelengths, so the photocurrent generation increase with the wider cell is expected to be minimal. Additionally, the current density calculation involves dividing the total current by the cross-sectional area of th e cell, which increases as the unit cell width increases. One interesting feature of the simulation is th at nearly all the cell widths showed a shortcircuit current density of approximately 9 mA/cm2, with the exception of two. The curves for 15 nm and 20 nm CdSe nanorod half-widths showed a short-circuit current de nsity of approximately 6 mA/cm2. To further explore this strange behavior at 15 and 20 nm nanorod half-widths, a full range of half-widths was explored between 10 and 25 nm, with the resultin g J-V curves shown in Figure 2-9. The results show that the drop in s hort-circuit current dens ity is a collection of seemingly arbitrary thickness values that result in a reduced current density rather than a trend in the range from 15 20 nm of CdSe half-thickness.. The cells with half-thicknesses of 15, 20, and 23 nm showed a short-ci rcuit current density in the range 5.4 to 6.4 mA/cm2, the cell with a half-thickness of 10 nm resulted in a short-circuit current density of 8.5 mA/cm2 while all other cells were slightly above 9.0 mA/cm2. Figure 2-10 shows the collection of cell performance data generated through the simulations involving variations in the width of the CdSe nanowires, and this data is tabulated in Table 2-6. The odd behavior of th e short-circuit current density is a prominent feature in the graph, with the cells at 15, 20, and 23 nm ha lf-thickness showing a significant drop in JSC. This results in a drop in cell efficiency at these point s, from over 3% to approximately 2%. With the exception of these outlying data points, the shortcircuit current density shows a minor but steady 36
PAGE 37
decrease as the CdSe width grows from 5 to 30 nm. This is to be expected as series resistance and the voltage across the junction increase and the cell current is bei ng divided by increasing cell areas to calculate current density. The cells fill factors remain steady for CdSe half-widths between 10 and 25 nm. At 30 nm, the VOC become very large (> 0.9 V), which causes a significant drop in the fill factor (< 0.5). These values were unable to calculated directly, so they are not included in Table 2-6. Illumination source Simulations were performed to determine the ef fect of altering the illumination source. Medici generates photons from an illumination source at a specified location that illuminates the sample at a specified angle and beam width. The illumination used for the cell dimension study used the form PHOTOGEN RAYTRACE BB.RADI BB.TEMP= 5800 WAVE.START=0.2 + WAVE.END=1.0 WAVE.NUM=@WL X.ORG=0.006 Y.ORG=-2.5 ANGLE=90 + INT.RATIO=1E-2 N.INTEG=10 RAY.N=1 RAY.W=1.0 This code generates radiation from a bl ackbody source at 5800K, which simulates the AM0 radiation spectrum. The statements X.ORG= and Y.ORG= define the origin of the light source. In the example above, the source is 6 nm to the right of the origin (centered on a 12 nm unit cell) and 2.5 microns off the surface of the cell. The Y.ORG value is negative because the surface of the device is at y = 0, so the area above the top of the devi ce is in the negative y range. The light is incident at 90 on the su rface, with a ray width (RAY.W) of 1 m. The results of varying the coordinates of the light origin are shown in Figure 2-11. Three of the four data sets are identical, with onl y one showing a difference. The perfect match between the two simulations with a centered light source shows that the distance away from the surface is not a factor. Additionally, placing the source at the left edge of the cell shows no change. However, placing the light source at the right edge of the cell results in a total loss of 37
PAGE 38
photocurrent in the cell. The J-V curve show s the same exponential growth as the other simulations and matches the other curves at large voltages where the current is primarily driven by applied voltage rather than photogenerated carriers. The lack of photocurrent generation when the sample is illuminated from over the right edge is puzzling. According to the MEDICI ma nual, the ray will generate on each side of the specified origin [65]. In this simulation, that should mean that the beam will generate from a range of 0.5 m on each side of the origin. Because the cell is 12 nm wide, this should easily illuminate the entire cell. This is illustrated in Figure 2-12. Because the angle of incidence specified is 90, generated rays wi ll be perpendicular to the top su rface of the cell. As shown in the illustration, the beam width of 1 m would be more than sufficient to illuminate the 12 nm cell width. The simulations shown in Figure 2-11 were performed a second time with the input file modified to remove the RAY.W statement. This allows MEDICI to set the beam width, which by default is 2x the greatest dimension of the subs trate [65]. The results of this simulation are not shown because they were id entical to the curves shown in Figure 2-11. There was no photogeneration when the illumination source was placed at the right edge of the unit cell. The cause of this phenomenon is uncl ear, but it is obvious that placing the beam origin at the right edge of the simulation area results in a total lack of photocurrent. Summary These preliminary simulations serve as a guide for how to appropriately define commands in Medici, as well as to create a basic understanding of the effect of certain model properties on the cell performance. It was found that impurity profiles should be defined using the REGION command to ensure the desired dopin g density is applied in each region. It was also found that the illumination source should be pl aced in the center of the cell, but that the distance from the 38
PAGE 39
surface of the cell and the definition of the ray wi dth are unimportant, provided that the ray width is high enough to illuminate the entire cell. Further studies will focus on the simulation of a realworld hybrid solar cell from literature. Simulation of a Real Cell Model Parameters The cell chosen for Medici simulation was repor ted in 2006 by Olson et al. [75]. The cell was a hybrid cell fabricated from an array of ZnO nanofibers and P3HT, and has a cell structure of ITO/ZnO/ZnO:P3HT/Ag. The ZnO layer was spin-coa ted onto the ITO substrate and the nanofibers were grown thr ough hydrothermal growth. P3HT was then spin-coated at a reported thickness of 200 nm. SEM images of the ZnO nanofiber array before and after P3HT spincoating are shown in Figure 2-13. The performa nce measures reported for the device were as follows: VOC = 440 mV, JSC = 2.2 mA/cm2, FF = 0.56, and = 0.53%. The J-V curve for this device is shown by the solid line in Figure 2-14. From the SEM images shown in Figure 2-13 (a), the average height of the ZnO nanofiber array is approximately 260 nm. From Figur e 2-13 (b), the thickness of the full ZnO:P3HT layer is approximately 430 nm. These values leave a P3HT capping layer of approximately 170 nm on top of the nanofiber array. Th e image in Figure 2-13 (a) shows an average rod diameter of approximately 30 nm. The authors note the rod spacing is approximately 100 nm. It is extremely difficult to accurately estimate the spacing from the cross-section SEM image shown in Figure 2-13 (a), but this figure appears to be reasonable and was likely verified with other unpublished data, so it will be assumed to be accurate. From the SEM image, a thin base coat of ZnO is visible on the ITO film. This film wa s estimated to be 25 nm thick from the SEM images, and this height was in cluded in the 260 nm rod length. 39
PAGE 40
The authors note that the thickness of their P3HT film was 200 nm, but are somewhat unclear whether this 200 nm thickness is measured from the top of the nanof iber structure or if it represents the full film thickness. If the scale bars for the SEM images are reliable, it appears that the 200 nm figure refers to the excess P3HT film on top of the fiber array. This seems to be an excessive amount of capping considering that the exciton diffusion length of P3HT is on the order of 10 nm in P3HT [15], a fact which is referenced by the authors. However, light is incident from the ITO/ZnO side of the cell in the reported device, so photons must pass through more than 200 nm of ZnO:P3HT film before reaching this capping layer. Additionally, the inconsistent length of the ZnO nanofibers must be considered, as the rela tively high mobility of both electrons and holes in ZnO compared to P3HT would create a short-ci rcuit in the device if the wire tips were exposed. With this consider ation, the additional buffer layer thickness may be important experimentally to ensure that the device will function as a diode. The J-V curve shown in Figure 2-14 was conve rted to numerical data using a graph digitizer program [76] so it coul d be compared to simulated curv es. The converted J-V and P-V curves are shown in Figure 2-15. To verify th e effectiveness of this conversion, the published cell performance measures were compared to th e performance measures calculated from the digitized curve. The results are show n in Table 2-7. The results were VOC = 0.44 V, JSC = 2.2 mA/cm2, FF = 0.57, and = 0.55%. This shows exact matches for VOC and JSC and values for FF and that are 0.01 and 0.02 higher than the published results. This is less than a 5% error in both cases, and demonstrates that the curve was re-produced with a high accuracy. This new curve can now be plotted with the results of simulated cells to find a best fit. From previous work performed by Dr. Woo K young Kim, the material properties for ZnO were defined as shown in Table 2-8 [66]. 40
PAGE 41
The unit cell used for these simulation effort s is shown in Figure 2-16, consisting of half the width of a nanorod and half of the space between the nanorod in the x direction and the full cell thickness in the y direction. The unit cell dimensions were set based on the published cell dimensions observed in Figure 2-13. In the unit cell, the transparent cond ucting oxide of ITO is shown as the top contact. There is a thin 25 nm layer of ZnO co ating the ITO electrode, and the nanorods extend out from this structure. The P3HT region exists beside and above the nanorod. The back electrode of Ag completes the structur e. From the SEM images in Figure 2-13, the nanorod width was approximated as 30 nm and the average rod spacing was approximated as 100 nm. Because the unit cell consis ts of half of a nanorod and ha lf of a rod spacing distance, the rod width shown in the unit cell is 15 nm and the P3HT width is 50 nm, resulting in a unit cell width of 65 nm. The model contains two regions of P3HT, a photocurrent generating region and a nongenerating region. For convenience, the region shown in Figure 2-16 c) will be listed as P3HT1, and the region is Figure 2-16 d) will be listed as P3HT2. The P3HT1 region is the region within one exciton diffusion length (LD) that genera tes photocurrent. Any references to the P3HT1 region, the LD region, or the P3HT absorbing region refer to this area. The P3HT2 region is outside of the diffusion length region, and c onducts holes to the Ag elecrode but does not contribute to photocurrent in the cel l. This division is included to account for the effects of the exciton diffusion length in the polymer. The P3HT2 region is interchangeably referred to as the P3HT bulk region. The effective diffusion length (LD) is a parameter that can be adjusted in the simulations, but has been shown experimenta lly to be on the order of 10 nm [15]. Materials properties will be adjusted in the non-generating region to prevent photocurrent from originating in this region. Several pote ntial adjustments will be considered, including 41
PAGE 42
reduction of the electron mobility to zero, setting the electron free carrier lifetime to zero, or setting the absorption coefficient in the regi on to zero for all wavelengths. For simulation purposes, it is assumed that the ITO electrode is perfectly transp arent and that the Ag electrode has a reflectance of 90%. The unit cell structure as seen in Medici is shown in Figure 2-17. In this image, the blue region is ZnO, the red region is the P3HT1 region and encompasses all areas within 10 nm of the ZnO, and the green region is the P3HT2 region which includes all areas outside of 10 nm from the ZnO. Note the difference in scale between the xand yaxes. Initial Simulations Simulations were performed using the mate rials properties found in Table 2-1 for P3HT and Table 2-8 for ZnO. The J-V curves for th ese initial simulations are shown in Figure 2-18, along with the curve for the literature cell. These simulations de monstrate the effect of some variation in the properties for the P3HT2 region. In simulation tsf496_1, the absorption coefficient for the P3HT2 region is set to zero for all wavelengths. In simulation tsf496_2, no changes are made to the film properties, so all parts of the P3HT region are allowed to generate and transport free carriers. In simulation ts f496_4, the electron mob ility is set to 0.0001 cm2/Vs in the P3HT2 region, representing a reduction by one order of magnitude from the original value. The simulation tsf496_2 is a case where the entire P3HT region is treated as a photocurrent generating layer. This represents a case wher e the exciton diffusion lengt h is infinite. This should represent the maximum possible performance of a cell constricted by the materials properties shown in Table 2-1 for P3HT. The efficiency of that simulated cell was 5.55% with a short-circuit current of 11.66 mA/cm2. The simulation data range was stopped at 0.9 V to limit the size of output files and the simulati on time, so a direct measurement of VOC was unable to be obtained. From the shape of th e curve, it appears that the VOC would be approximately 1.0 V. 42
PAGE 43
Those numbers give an estimated fill factor of approximately 0.5, which is relatively low considering that this repres ents a best case scenario. In simulation tsf496_4, the electr on mobility is reduced by an order of magnitude in the P3HT2 region. This was an attempt to allow real istic absorption in the region but strictly limit the ability of the electrons generated in this area to reach the ZnO and contribute to the overall photocurrent. The resulting J-V curve showed a strange double-curve behavior, with the maximum power point occurring in the sec ond elbow at approximately 0.75 V. The doublecurve behavior is not physically r ealistic, but it is interesting to note that the performance of this simulation was higher than the performance of ts f496_1, showing that even with the low electron mobility, current generated in the P3HT2 region was able to make a contribution to cell performance. Simulation tsf496_1 shows the case where the abso rption coefficient is set to zero in the non-generating P3HT2 region. Of the three curves shown in Figure 2-18, this is the most similar to the real curve. Although the VOC and JSC were significantly higher th an that of the real cell, the shape of the curve seems to be very similar. The current density in creases very slowly over the low-voltage regime until approximately 0.6 V, wh ere there is a distinctive elbow in the curve and the current begins increasing more rapidly. The slope of the curve as it approaches VOC appears to approximate th at of the real curve. Reflectance of Ag electrode Because the incident light angle is set at 90 in simulation tsf496_1 and the boundaries between layers are set at right angles, one physical inconsistency in this model is the amount of reflected light that returns to the P3HT1 and ZnO regions. On the le ft-hand side of the cell where the nanorod resides, reflected light in the real cell would need to pass through 320 nm of the nonabsorbing P3HT2 region between its first and s econd passes through the absorbing P3HT1 region. 43
PAGE 44
On the right-side of the cell, light would ha ve to pass through 790 nm of the non-absorbing P3HT2 region before returning to the thin absorbing layer of poly mer and ZnO at the surface of the cell. Based on this, it is obvious that this si mulation allows more light to be absorbed in the generating regions of the model th an would exist in reality. Th is was corrected by studying the effect of reducing, reflection at the back electrode. The absorption coefficient of P3HT ranges from ~ 2 x 104 cm-1 at < 350 nm and > 680 nm to a maximum value of ~ 2 x 105 for = 540 nm. At the low-end absorption value, approximately 20% of incident light is absorbed over the 10 nm LD region, according to Equation 2-1. xe I I0 1 (2-1) I1 represents the amount of light transmitted through the film, where I0 is the intensity of incident light, is the absorption coefficient, and x is the film thickness. Nearly 99.8% of the remaining photons would be absorbed over the 320 nm path length consisting of the forward and backward pass through the non-absorbing region in the si mulation. Even in this case, with a low absorption coefficient and the shortest considered path length, virtually no photons would remain in a real cell to be absorbed afte r reflection from the back electrode. With this consideration, the next line of simu lations was performed with reflectance at the Ag electrode either reduced or completely removed. The resulting J-V curves are shown in Figure 20. Despite the removal of all reflec tance from the Ag electrode, simulated cell performance was still substantiall y higher than that of the real cell for all simulations. All of these simulations prevent absorption in the P3HT2 region. By comparing simulations with varying electrode reflectance and constant mobility and absorption properties (mobilities as defined in Table 2-1, absorption set to zero in the P3HT2 44
PAGE 45
region) we can see the effect of reflectance at the Ag electrode with no absorption in the P3HT2 region. These curves are shown in Figure 2-19. As expected, the short-circuit current density decreases as the Ag reflectance is decreased. Th e simulation with the Ag reflectance set to zero results in a JSC = 3.6 mA/cm2, which is still approximately 1.5x hi gher than the real cell value of JSC = 2.2 mA/cm2. Mobility In addition to the removal of reflectance from the back electrode, the charge mobilities in P3HT were reduced to restrict curre nt flow in the device. The physical justification for this is that the initial value of hole mobility was take n from the literature [71], and was the highest reported value for hole mobility in P3HT. Additionally, this value is a field-effect mobility which relies on a strong applied bi as to drive current flow. In photovoltaic devices, biases are significantly lower, and this field-effect mobility may be a serious over-estimation of the true film properties. Also, charge mobility has been shown to be anisotropic in P3HT [77], so values of mobility can vary depending on the alignment of the polymer chains in the film. Based on these reasons, simulations were performed to evaluate th e effect of reducing mobility values in the polymer regions of the mo del, with the resulting J-V curves shown in Figure 2-20. As predicted, simulated cells with higher mobilities resulte d in stronger device performance. The black and green curves, corres ponding to no change in mobility values and a 50% reduction of mobility values in the P3HT2 region only, are nearly id entical. A change from the initial mobility values to a 50% reduction in both P3HT regions, shown by the black and red curves, results in an efficiency drop from = 2.12% to = 2.00%, or approximately 6%. Reducing the P3HT mobility b 90% in both regions results in an efficiency drop of approximately 23%, from = 2.12% to = 1.64%. Attempts to reduce the P3HT2 mobility 45
PAGE 46
values by 90% while holding the P3HT1 values at their original values failed to converge in Medici. Although the J-V curves generated on this gr aph stop at a value of 0.9 V before reaching the open-circuit voltage, extrapolations estimate that the VOC is virtually unchanged in these simulations, with a value of approximately 1.04 V. Using this estimation, cell performance measures were calculated and are displayed in Table 2-9. The short-circuit current density decreases with each drop in the mobility. For the original mobility values, JSC = 3.61 mA/cm2. For 50% and 90% reductions in mobility, the JSC drops to 3.58 mA/cm2 and 3.44 mA/cm2, respectively. Using the extrapolated value of VOC = 1.04 V, fill factors for each cell can be calculated as 0.57, 0.54, and 0.46 for mobility values of 100%, 50%, and 10%, respectively. This reduced fill factor is obvious from Figur e 2-19, as the curve corresponding to a 90% reduction in mobility shows an obvious reduction in the sharpness of the elbow shape of the curve. Based on these values, a 50% decrease in mobility resulted in a 6% drop in efficiency, a < 1% drop in JSC, and an estimated 5% drop in fill factor. The 90% decrease in mobility resulted in a 23% drop in efficiency, a 5% drop in JSC, and an estimated 19% drop in fill factor. Exciton diffusion length Even with a reduction in charge mobilities by 90% in the P3HT regions, complete removal of absorption in P3HT outside of a diffusion length, and elimination of all reflection from the silver electrode, simulated cell performance greatly exceeds that of the real cell. The simulation with the lowest performance showed a JSC of 3.44 mA/cm2 and of 1.64%, along with estimated values for a VOC of 1.04 V and a fill factor of 0.49. To match the published experimental data, the JSC must be reduced by another 1.2 mA/cm2, the VOC by 0.6 V, and the efficiency by 1.1%. The exciton diffusion length was adjusted dow nward from the literature value of 10 nm [15] to compensate for this inconsistency. As noted previously, the ex citon does not exist in 46
PAGE 47
Medici, so the definition of an exciton diffusion length in this study is ta ken to be the width of the P3HT1 region. Since this is the only P3HT region generating photocurrent in the model, it is taken to be an accurate approximation of the exci ton diffusion length in a real cell. For this study, with results shown in Figur e 2-21, the absorption coefficient was again set to zero in the P3HT2 region and the carrier mobilities in both P3HT regions were set to th eir original values of n = 0.001 cm2/Vs and p = 0.01 cm2/Vs. As predicted, the simulated cell performance decreased as the exciton diffusion length decreased. This was a re sult of decreasing short-circuit current density, which decreased linearly with exciton diffusion le ngth, as shown in Figure 2-20. Estimates for open-circuit voltage actually increase d slightly as the diffusion length decreased, as shown in Figure 2-22. This trend was very slight, however, so it showed virtually no impact on cell performance. Based on these simulations, an exciton diffusion length of 6 nm was chosen for future simulations. Doping density From previous simulations of CdSe:P3HT hybrid cells, changing the doping level of CdSe showed a strong impact on the VOC of the cells. Based on thes e results, the doping density of P3HT was varied from the initial value of 5 x 1016 cm-3 to 1 x 1014 cm-3 in an attempt to reduce the VOC of the simulated cells. The simulated J-V cu rves are shown in Figure 2-24. Contrary to expectations, the VOC of the cells did not appear to change with the reduction in P3HT doping. A key difference between this study and the simu lations performed on the test cell of CdSe and P3HT is that CdSe and P3HT showed very similar doping densities: 6x1016 cm-3 for CdSe and 5x1016 cm-3 for P3HT. The doping density of ZnO is 5x1017 cm-3, so it is possible that the VOC is dominated by this wide discrepancy. To test this hypothesis, simulations were performed with reduced ZnO doping density. These tests fixed the P3HT doping density at 1x1016 cm-3, based on the results shown in Figure 247
PAGE 48
24. The results from this test are shown in Fi gure 2-25. Oddly, this variation fails to show a decrease in VOC. The VOC of a cell should depend on fact ors such as doping density and minority carrier lifetime [78], but the doping density variations here doesnt show any impact. Density of states Simulations were performed to determine the effect of the density of states in the conduction and valence band of P3HT, and the resulting J-V curves are shown in Figure 2-26. The curves show a decrease in performance as the density of valence band states increases, although extrapolated values for open-circuit voltage remain relatively constant. The curves are independent of the density of conduction band stat es, as shown by the overlap of the black and red data points and the green and yellow data points. Open-circuit voltage examination Simulations focused on the effect of carrier mobility were again performed, but with an expanded voltage range so that th e open-circuit voltage could be observed rather than estimated. These resulting J-V curves from these simula tions are shown in Figure 2-27. As predicted through previous extrapolations, the open-circuit voltage is independent of the carrier mobility and the short-circuit current de nsity is strongly impacted by it. Extrapolated values for VOC in previous simulations predicted values near 1.04 V, but these simulations show that VOC = 1.27 V for these cells. As noted earlier, the cells s how a double-elbow effect that was unable to be explained. The effect of P3HT doping density was also reexamined with the expanded voltage range, as shown in Figure 2-28. As predicte d through previous ex trapolations, the VOC is independent of the doping density of P3HT. Again, the VOC = 1.27 V for these simulations, and the shortcircuit current density decreases as the P3HT doping density decreases. This mimics the behavior seen in previous studies on this material system. 48
PAGE 49
Although the short-circuit current density and fill factor can be altered by variations in carrier mobility, doping density, and vale nce band density of states in P3HT, these tools offer no control over the open circuit voltage of the simulated cells. From a theoretical perspective, the open circuit voltage of organic a nd hybrid cells is controlled by the energy difference between the HOMO level of the electron donor and the LU MO level of the electron acceptor, or the conduction band level if the acceptor is an inorga nic [79]. In the case of a hybrid cell, the LUMO level of the organic el ectron acceptor is replaced by th e conduction band energy of the inorganic material. In other words, carriers do no t exist with energy equal to the band gap of the material where they are generated; instead they exist with energy equal to the spacing between the bands of the junction materials. The energy band diagram for the P3HT ZnO system displayed in Figure 2-29 shows th is band offset to be 0.85 eV. Based on this, free carriers resulting from photoabsorption in P3HT should exist at an energy of 0.85 eV rather than 1.7 eV. Medici allows independent specification of the energy band gap and optical band gap in each material. The optical band gap should remain fixed at the appropriate level to dictate the location of the absorption cutoff for each materi al. The energy band gap, however, presents an additional tuning control to desc ribe the energy of photogenerated carriers in the device. Figure 2-30 shows the results of simulations with variations in the energy band gap of P3HT. In the curve tsf496_45, the energy band gap is set to 0.85 eV in the P3HT regions of the cell. In curve tsf496_53, both th e energy band gap of both the P3HT and ZnO regions is fixed at 0.85 eV. From these curves, fixing the energy band gap of P3HT accurately simulates the VOC of the literature cell. Altering the ZnO band gap in addition drops the VOC to below 0.4 V, beyond the published cell performance data. 49
PAGE 50
Absorption coefficient adjustment Under close inspection of a Medici output file, it was noticed that the program automatically assigns absorption coefficient va lues for wavelengths outside of the range specified in the input f ile if this range is smaller than th e full spectrum used for calculation. These values are not set to zero, but assigned a default value based dependent on the wavelength. Figure 2-31 shows the absorption coefficient curves as assigned by Medici for three absorption input files. The first two files, abs_none and abs_p3ht, have an input range specified from 0.3 to 0.7 m and correspond to zero absorption and the absorption profile of P3HT, respectively. The range of 0.3 to 0.7 m was chosen because this is the range of data shown in the absorption curve from literature [74]. The file abs_i_ zno was borrowed from Dr. Woo Kyoung Kim in his simulations of CIS solar cells [65]. This input file ranges from approxima tely 0.2 to 1.0 m. From Figure 2-31, it is obvious th at Medici assigns identical va lues to all absorption files in the range from 0.12 to 0.2 m. Similarly, it assi gns an identical data point to all files at 1.24 m. Although absorption in the high wavelength region is set to zero by the program, absorption in the short wavelength region is set to a very high value in the range of 106 cm-1. Although the photon flux is low in this region, as shown in the AM1.5 spectrum in Figure 2-32, this extremely high absorption coefficient leads to a significant amount of carrier generation in the cell. This leads to an over-estimation of th e number of free carriers in the ce ll and artificially enhances cell performance in the simulations. This effect ha s no impact on the ZnO absorption, because in the simulation photons are generated from 0.2 to 1.0 m. Figure 2-33 shows the number of photogenerated carriers as a function of wavelength. Figure 2-33 shows the carriers generated by this absorption region between 0.2 and 0.4 m, but does not show the exact effect on the J-V curves. In Figure 2-34, J-V curves are compared for cells simulated with the P3HT absorption profiles shown in Figure 2-31 and an 50
PAGE 51
absorption profile with all absorption between 0.2 and 0.3 m set to 2.24x104 cm-1. The curve with the corrected ab sorption coefficient, shown in blue, shows a drop of 0.14 mA/cm2 in the short-circuit current when compared to the curv e with Medici-assigned va lues for the absorption coefficient, shown in purple. This difference is minimal, but should be included in simulations for improved accuracy in the simulations. Multi-stage absorption The technique used for simulations up to this point involved a 2-layer P3HT region with a thin absorbing region near the ZnO region and a large non-absorbing region. The width of the absorbing region was set equal to the assumed diffusion length of an exciton. Because the diffusion length of an exciton is not explicitly included in the program, Me dici assumes that all photogenerated carriers within this absorbing region are free car riers, which are driven across the space-charge region by the built-in el ectric field. This is only a first-order approximation of the true device physics of a hybrid solar cell. The exciton diffusion length is an average distance that an exciton can travel before it recombin es, not a firm line beyond which all excitons are doomed to recombination. The excitons do not se lectively move toward the inorganic organic interface, they simply travel by diffusion until they recombine or dissociate at the interface. To approximate this distinction, the model was modified to include multi-stage absorption in the P3HT region where the probability of exc itons reaching the interface is included. A framework for a simulation involving multip le layers in the absorbing region was developed with a multi-layer region extending to 20 nm from the ZnO region. J-V curves shown in Figure 2-35 correspond to simulations in which the first 10 nm of this region is set as an absorbing region using the P3HT absorption profile. The second 10 nm is set as a non-absorbing region, identical to the P3HT2 region. These cells all have identical structures, with the only difference being the number of stages within this 20 nm region. For example, the simulation 51
PAGE 52
with 10 stages has 10 layers that are each 2 nm th ick. The first 5 stages are set to absorb based on the P3HT absorption profile, and the final 5 stages are set as non-absorbing. Similarly, the simulation with 4 stages contains 4 layers with 5 nm thickness in each layer, with the first two set to absorb and the final two set as non-absorbing. In the case of the single stage cell, the absorption region was set from 0 to 10 nm away from the ZnO interface, and the P3HT2 region began at 10 nm. Because the cell structures ar e identical in terms of exc iton diffusion length, simulations were expected to produce identical J-V curves. Although the curves are all similar in shape and performance data, they are not identical, due to slight differences in the iterative calculation in Medici. In all the simulations shown in Fi gure 2-35, absorption was set to 100% of the P3HT value in all stages betwee n 0 and 10 nm into the P3HT region, and it was se t to 0 in all stages between 10 and 20 nm. This result is interesting, but is not the goal for this multi-stage model design. The purpose is to create a graded carrier generation profile that will average to the sp ecified LD, but allow for some carriers inside of LD to fail to reach the interface and allow some carriers outside of LD to succeed in reaching the interface. The structures used to generate the J-V curves in Figure 2-35 can be thought of as normal curves with standard deviations of zero. Ab sorption is 100% of the P3HT value up to the specified diffusion length of 10 nm, and then drops as a step function to 0% of the P3HT value for the next 10 nm. Using this multi-stage framework, distribution curves with nonzero values for standard deviation can be imposed on the simulation. In this case the property being distributed is the ab sorption coefficient, as this is the property chosen to simulate the exciton diffusion length. It should be noted that these curves are not exactly equal to cumulative distribution functions for a true nor mal distribution because in this case the area 52
PAGE 53
considered was only between 0 and 20 nm. In a true normal distribution with a mean of 10 nm and a standard deviation of 4 nm or greater, an appreciable area exists under the curve beyond 0 and 20 nm. To remove this area from consid eration, normalization was performed by dividing by the sum of the area between 0 and 20 to esta blish these boundaries. The distribution was calculated as shown in Equation 2-2. 2 22 exp 2 1 xx xF (2-2) In Equation 2-2, x is th e distance into the P3HT region and x is the mean of the distribution, set to 10 nm in this case. By dividing this by the sum of all distribution values between x = 0 and x = 20 nm, the normalized distribution was obtained. The cumulative distribution was calculated using Equation 2-3. 20 0 0)( 1x x x xxF xF xG (2-3) Again, this cumulative distribu tion is normalized by the area under the distribution function between 0 and 20 nm rather than between nega tive and positive infinity. This cumulative distribution function is shown in Figure 2-36 for a range of standard deviations and a mean of 10 nm. By coupling these distribution functions with the multi-stage cell design demonstrated in Figure 2-35, graded absorption can be genera ted in the simulation by applying a fractional absorption coefficient in each stage of the absorption region. Simulations were performed to consider the effect of this absorption distribution in the cells. These simulations used a 10-stage absorp tion region with a graded absorption coefficient in an attempt to more accurately define the absorption in the cell. Absorption profiles corresponding to standard deviations of 0 to 10 nm were generated by using a fractional value for 53
PAGE 54
the absorption coefficient in each region, rounded to the nearest 10% value. The input data used for the simulations are shown in Table 2-11, as we ll as the resulting short-circuit current density for that simulation. The J-V curves for th ese simulations are shown in Figure 2-37. Table 2-11 shows that absorpti on in the cell is distribute d over a wider range as the standard deviation increases. The average multiplier to the absorption coefficient is calculated for each situation and as targeted it is equal to 50% or 51% for all cases, with any error coming from rounding to the nearest 10% in each region. The J-V curves in Figure 2-36 show that this grading has virtually no effect on the final performance of the cel ls. The short-circuit current density shows extremely minor variations in the cu rves, and this value is tabulated in Table 2-11. The short-circuit current density increases slightly as the absorption distributions become wider. This shows that extending the generated carrier distribution toward the back electrode improves device performance, while limiting the carrier di stribution to a narrower region, even with the same number of carriers being ge nerated, hinders performance. P3HT replacement with CIS The J-V curves generated in previous simula tions shows a double-elbow shape that is not characteristic of photovoltaic cells. To determin e the root cause of this phenomenon, P3HT was replaced with copper indium diselenide (CIS), a popular p-type thin-film photovoltaic material. The same cell structure is used, with ZnO nanor ods as the n-type material. Within this framework, material properties of the CIS laye r were adjusted to values corresponding to P3HT. Preliminary simulations in this form were perf ormed, and the resulting J-V curves are shown in Figure 2-38. This set of simulations compares ZnO:P3HT cells with 10 nm and full cell absorption regions to ZnO:CIS cells with full ce ll absorption, a 10 nm absorption range, and a 10 nm absorption range with an energy band gap set at 0.85 eV. 54
PAGE 55
Figure 2-38 shows that a limited absorption ra nge in the simulations does not cause the double-elbow shape of the J-V curve, as the Zn O:CIS cell with a 10 nm absorption region shows a similar shape to the one with absorption in the full CIS region. The same holds true for comparisons of the two ZnO:P3HT cells, which both show the double-elbow shape. The ZnO:CIS cells produce J-V curves with the anticipa ted shape and high fill factor as compared to the ZnO:P3HT cells. This also held true for the cell in which the CIS electrical bandgap was reduced to 0.85 eV, although this cell resulted in a low VOC of 0.25 V. Parameters used for P3HT and CIS in the simulations are displayed in Table 2-12. Data for CIS was obtained from previous simulations fr om Woo Kyoung Kim [66]. The doping density is the only parameter that is the same in both mate rials. By individually adjusting parameters between the values for CIS and P3HT, an attempt will be made to determine the origin of the double-elbow J-V curve shape. Although not shown in the table, the absorption profile for each material was also adjusted. Using the same 10 nm absorption region as th e ZnO:CIS cell shown in red in Figure 2-38, the properties shown in Table 2-12 were individually changed fr om the CIS values to the P3HT values. The resulting J-V curves are shown in Figure 2-39. Performance measures for these cells are displayed in Table 2-13. These results show that the permittivity and density of states have virtually no impact on the cell performance. The three parameters that strongly impacted the J-V curves are electron affinity, carri er mobility, and absorption profiles. The change in electron affinity causes a shift in VOC from 0.44 V to 0.49 V due to the higher conduction band level for P3HT. The short circuit current is minimally affected by this change, but there is a notable drop in the fill fa ctor of this curve, down nearly 30% from the values of the original ZnO:CIS cell. The va riation in carrier mob ilities showed a 1.5 mA/cm2 55
PAGE 56
reduction in short circuit current paired with a 1 V increase in open-circuit voltage. The reduced mobility results in fewer carriers escaping the cell, causing a lower short-circuit current density. The strong increase in VOC is more difficult to explain, as adjustments to carrier mobility in previous simulations did not show this type of response. There is also a s light decrease in the fill factor for the cell with P3HT mobility values, but this effect was minor. When the absorption profile is changed from CIS values to P3HT values, the short-circuit current density of the simulated cell drops by nearly 50%, down to 4.1 mA/cm2. The VOC of that cell is slightly reduced, but the fill factor remains nearly unchanged. Based on the J-V curves shown in Figure 2-38 and detailed in Table 2-13, simulation 72 was chosen as a basis for further efforts. In this simulation, the absorption coefficient values were assigned values corresponding to P3HT rather than CIS, while maintaining a 10 nm absorption range in the simulation. Variati ons around this base case were performed by individually varying the other material pr operties from their CIS value to their P3HT value. The results are shown in Figure 2-40, with performance measure data shown in Table 2-14. As in the previous set of simulations, the addition of P3HT values for permittivity or density of states showed virtually no change from the control cell. The values for open-circuit voltage, shortcircuit current density, and fill factor of these cells were within 2% of the values for the control cell. As seen in the previous set of simulations, when the P3HT value for electron affinity is applied the J-V curve shifts to a higher VOC with a 30% drop in fill factor and a virtually unnoticeable drop in JSC. The application of P3HT values for charge mobility results in a 0.8 mA/cm2 decrease in JSC and a 0.1 V increase in VOC with an approximately 10% drop in fill factor. Despite reductions in fill factor for simulated cells with P3HT values for electron affinity and mobility, none of the J-V curves in Figure 2-40 displayed the double-elbow shape. 56
PAGE 57
The electron affinity was fixed at its P3HT value of 3.15 eV, setting the conduction band level in the simulations to the appropriate value for P3HT. The electronic band gap of the simulations is still set at the CIS value of 1.04 eV rather than the P3HT value of 1.7 eV or the effective band gap of 0.85 eV at the ZnO-P3HT junction. Again, the abso rption coefficient is set at the P3HT value and absorption is a llowed over a 10 nm range near the material interface. Variations in the other materials properties resu lted in the J-V curves shown in Figure 2-41 and detailed in Table 2-15. As seen in previous si mulations, there is virtua lly no change in the J-V curve or device properties w ith the application of the P3HT values for permittivity and density of states. The application of P3HT values for carrier mobility, howev er, resulted in a drastic shift in the nature of the curve. This change resulted in a 0.5 mA/cm2 reduction in the short-circuit current density and an increase of ~0.6 V in the open-circuit voltage. Additionally, the J-V curve takes on the double-curve shape, which drops the fill factor to approximately 0.20. This shape will be discussed in more detail after the next set of data. As discussed previously, the J-V curves show n in Figure 2-41 resulted from simulations where the electron affinity was set at the P3HT value of 3.15 eV, but the electronic band gap remained at the CIS value of 1.04 eV. Although fi xing the electron affinity sets the LUMO level of P3HT, the band gap value sets an inappropriate HOMO level and results in the generation of carriers with higher energy than appropriate. The simulations of Figure 2-41 were modified to include the appropriate ZnO-P3HT interfacial band gap of 0.85 eV, corresponding to the energy gap between the conduction band of ZnO and the HOMO level of P3HT. The results, shown in Figure 2-42, mimic those in Figure 2-41. The curves resulting from adjustments in th e permittivity and density of states show the expected shape for a solar cell J-V curv e despite having a significantly lower VOC than their 57
PAGE 58
counterparts in Figure 2-41. This variation is ex pected, due to the shift in energy band levels causing the generated carriers to exist with more energy. The JSC of the cells remains virtually unchanged as compared to their counterparts in Figure 2-41, but the shift in VOC results in a slight reduction of the fill factor as the curves cross the voltage axis at a slightly lower slope. Curve 82, corresponding to a change in carri er mobility, also shows a reduction in VOC when compared the curve shown in Figure 2-41. This curve shows a very small fill factor of approximately 0.15 due to the double-elbow shape wh ich eliminates most of the active area of the curve. The origin of the change in shape accompanying lower mobility values is unclear, but it obviously occurs only when the ab sorption, electrical energy levels and mobilities take on their P3HT values. This shape was observed for all ZnO:P3HT cells simulated up to this point, but it was not seen in other simulations using a ZnO:CIS cell as a basis. From the graphs in Figures 239 2-41, it can be concluded can state that the permittivity and density of states have no effect in causing this shape. Additionally, the application of P3HT levels of absorption and energy levels to a ZnO:CIS cell did not cause this shape without the addition of P3HT levels for carrier mobility. In the simulations, the carrier mobilities were set as n = 0.001 cm2/Vs, p = 0.01 cm2/Vs for P3HT, and n = 30 cm2/Vs, p = 300 cm2/Vs for CIS. Changing from CIS values to P3HT values represents a severe drop of 4 orders of magnit ude with no further details for intermediate values. Using simulations of a ZnO:P3HT cell, a wide range of mobility values were examined, with the resulting J-V curves shown in Fi gure 2-43. Note that in all cases, p = 10*n. Additionally, these simulations a llow absorption throughout the full P3HT region of the cell to boost the level of current flow a nd more clearly show the effect of the mobility variations. 58
PAGE 59
These simulations demonstrate that at highe r mobility values, the J-V curve shows the expected shape for a solar cell. At lower mob ility values, the curve inverts to the double-elbow shape. An ideal solar cell J-V curve should ha ve a positive second deri vative over the entire active range, from V = 0 to V = VOC. It is difficult to calculate a second derivative for these curves because they are not easily fit to empirical equations. However, the sign and approximate magnitude of the second derivative over a short range of the data can be determined by using Equation 2-4. Jest = (y3 y1) / 2 y2 (2-4) A visual representation of this calculati on is shown in Figure 2-44. The values y1, y2, and y3 represent current density values for equally sp aced applied voltages. For these J-V curves, y is current density and x is voltage. The first te rm of Equation 2-4 is noted in Figure 2-44 as y2 0, and is the midpoint of a line drawn through the points (x1, y1) and (x3, y3). If this value is greater than the actual value of y2, Jest calculated from Equation 2-4 will be positive and will correspond to a curve that is con cave up, as shown in the figures. Although this is not an exact measure of the second derivativ e, it will result in a posit ive value (the value for y2 0 is greater than the value y2) for curves that are concave up and a negative value for curves that are concave down. Additionally, th e magnitude of Jest will give a relative idea of how close to linear the curve is over the range from x1 to x3. This calculation is applied to the simulated J-V curves for ZnO:P3HT cells shown in Figure 2-43, and the calculated Jest values are plotted in Figure 2-45. Note that these calculations correspond to cells with absorption in the full P3HT region. Of all the curves shown, only those corresponding to p = 5,000 and p = 10,000 cm2/V-s showed second derivative values over the entire active region. 59
PAGE 60
When these same simulations are performed on ZnO:P3HT cells featuring carrier generation only within a 10 nm exciton diffusion length, the curves shift significantly, as shown in Figure 2-46. The most obvious and expected feature is that the current density drops significantly due to the lower number of carriers being generated in the limited diffusion length region. There is also a slight drop in the open-ci rcuit voltage for these ce lls, but an increase in the fill factor as the curves show less of an inve rted shape. This is shown more clearly in Figure 2-47, which displays the Jest values for these simulated cells. Cells with mobility values as low as p = 500 cm2/Vs show positive Jest values for the entire active re gion in this case, a full order of magnitude smaller than the threshold mobility in the full cell absorption simulations. Lower values of carrier mobility generate JV curves with consistently positive second derivatives in the 10 nm absorption region simula tions as compared to the full cell absorption region simulations. This refutes the theory th at a second p-n junction region arises at the interface of the absorbing and non-absorbing P3HT regions. Instead, the charges generated at a greater distance from the ZnO-P3HT interface drive this inversion of the J-V curves. This is due to the low mobility of electrons in P3HT and their difficulty in traveling through larg e regions of the polymer to reach the ZnO regions. The comparisons of ZnO:CIS cells demonstr ated that three parameters are key to controlling the shape of the J-V curve in these simu lations. Carrier mobility dictates the shape of the curve, as demonstrated in Figures 2-43 and 246. Altering absorption in the cell, such as by changing the exciton diffusion length or the absorp tion coefficient, has a direct impact on the number of charge carriers and th erefore on the current density of th e cell. This is demonstrated in Figure 2-48 for cells with the P3HT hole mobility set to 500 cm2/V-s and a 10 nm absorption 60
PAGE 61
region. The energy band gap of the P3HT region in the cell alters th e open-circuit voltage of the cell. This is shown in Figure 2-49 for cells with p = 500 cm2/V-s and LD = 10 nm. The curves shown in the two figures above demo nstrate the level of independent control of Jsc and Voc that is available from variations in ab sorption and energy band gap, respectively. To accurately simulate the experimental data, however, an array of combinations must be examined. Figure 2-50 shows a set of curves spanning four values for each of these parameters. In the graph, colors are constant for constant band gap, while line style is constant for constant absorption coefficient values. The experimental data falls somewhere in the range between 20% and 50% absorption with a band gap of 1.0 eV. This set of curves was expanded to s how a wider range of absorption, all with the band gap held constant at 1.0 eV and the hole mobility set at 500 cm2/Vs. The results are shown in Figure 2-51. Although th e curve for 30% absorp tion produces a nearperfect match for JSC and a close match for VOC, the simulation produces a curve with a lower fill factor than the real data. Based on this, the search must be re-expande d to three parameters: carrier mobility to control fill factor, band gap to control VOC, and absorption coefficients to control JSC. However, it must be considered that each of these controls impacts all three target parameters, not just the intended one. Shortcomings of the initial model This simulation strategy was abandoned as the physics of the model were more closely analyzed. All simulations up to this point suffered from one distinct inconsistency the existence of electrons in the P3HT region. This is demonstrated in the strong impact of electron mobility on the shape of the J-V curve for ZnO:P3HT cells. The polymer absorbs photons to generate excitons rather than free carriers, but th e model is not aware of th is distinction. As a result, the electron mobility of P3HT is an important factor in th ese simulations, but it is of minor 61
PAGE 62
importance in real cells. Because of this dis tinction, a new strategy has been employed to more accurately represent exciton physics in Medici. Two-Step Simulation Technique The important feature of excitonic solar cells, as has been discussed, is the generation of excitons rather than electron-ho le pairs. One important consid eration, addressed in simulation efforts described previously, is the limited carrier generation due to the exciton diffusion length. However, another important effect, not directly c onsidered in previous s imulations, is that the exciton dissociates at the material interface. This means electrons should not be generated in the P3HT region; instead, free carriers should be generated directly at the material interface. A new modeling strategy is needed to incorporate this effect into the simulations. This new strategy involves a two-stage simula tion. In the first st age, the photogenerated carrier distribution is measured over the entire de vice area under simulated solar illumination. In the second stage, this distributi on is compressed to a line source of carriers generated directly at the organic-inorganic material interface by mapping the carriers generated at all points to their nearest interfacial point. Specifying a line source of carriers in Medici requires specifying th e origin and endpoints of the line with X and Y coordinates, as well as parameters to describe the carrier distribution along that line, as shown in Equations 2-5 2-8. (2-5) (2-6) (2-7) (2-8) 62
PAGE 63
Gn and Gp are generated electr ons and holes, respectiv ely. This electron-hole pair generation equation consists of th ree components: lateral, radial, and time-dependent. The physical dimensions l and r represent lateral distance along the line and radial distance from the line, respectively. The lateral dependence can be specified as an exponential-linear function that requires the definition of parameters A1, A2, A3, and A4. These parameters are determined through regression for each scenario studied, and are discu ssed in further detail in the following sections. The radial dependence follows an exponentia l decay using the decay constant R.CHAR, which is set to 0.0001 m for these simulations. This d ecays the radial component of the generation by 44 orders of magnitude within 1 nm of the line, so that the generation occurs only at the material interface and not in the individual materials. Carrier generation is assumed to be uniform in time. The time dependence is set to the default value of T(t) = 1. Photogenerated Carrier Distribution The distribution of photogenerate d carriers in the unit cell was determined through Medici. The unit cell was exposed to simulated solar illu mination as described in previous sections. Medici calculates photoabsorption and carrier generation at each point in the simulation mesh, and this data can be displayed as a contour pl ot. Unfortunately, it seems to be impossible to extract the numerical source data from contour plots in the program This is possible, however, for line plots, as this is the technique used in to extract J-V data from the simulations. Line plots were generated at each mesh point along the x-axis of the unit cell, and these lines extended through the thickness of the device so that the car rier generation data for every mesh point in the unit cell was collected. Using this (x, y, z) data set, co ntour plots were created in Sigmaplot to show the dist ribution across the device area. 63
PAGE 64
A contour plot of the photogene rated carrier distribution in the full cell and a surface plot focused on the area within LD from the side of the nanorod are shown in Figure 2-52. The exciton diffusion length for the cells depicted is 10 nm. Note that in the plots, y = 0 represents the top surface of the cell, bordering on the ITO electrode. There is a 0.025 m surface layer of ZnO, with the ZnO nanorod extending to y = 0.26 m and being 0.015 nm wide. This region appears as dark blue to purp le in Figure 2-52A, as ZnO is nearly transparent to most wavelengths. The 10 nm absorbing region of P3HT appears as the brightly-colored region in Figure 2-52A, and is the focus of Figure 2-52B Although absorption is strong in the strips above the nanorod tip and just beyond the ZnO base layer, a large portion of the carrier generation in the unit cell occurs in the area on the side of the nanorod. It is interesting to note from Figure 2-52B that the strong absorption fr om the region above the nanorod tip extends slightly into the area to the side of the nanorod. It is unclear from simulations if this occurs due to refraction at the interface or if it simply a numerical anomaly to facility convergence. The P3HT bulk region is seen in blue in Figure 2-52A. Although the ab sorption coefficient is set to zero in this region, a low leve l of carriers still exist. Line Source Generation Line sources of carriers we re imposed along the ZnO-P3HT interface, consisting of three lines due to the shape of the in terface. The first line is at the tip of the ZnO nanorod. For simplicity, this is noted as Line 1. The line ex tending along the edge of the nanorod is referred to as Line 2, and the line running along th e top of the ZnO base layer is Line 3. To determine the density of photogenerated carri ers to be imposed along these line sources, all carriers generated by the simulated solar radiation are mapped to their nearest ZnO-P3HT interface point. This mapping scheme is illustra ted in Figure 2-53. Gray areas represent ZnO, while blue areas represent P3HT. Carriers in a particular region are directed to the nearest line as 64
PAGE 65
illustrated with the orange arrows. Dotted white lines show the edges of regions that map to a specific line. As one of these lines is crosse d, the carriers begin mapping to a different line. There are two blocks, in the top-left and bottom-ri ght corners, that map to a single point at the intersection of two lines. Once the carrier densities from each mesh point in the unit cell are mapped to their corresponding point along one of th e three lines, these carrier dist ributions are fit to exponentiallinear curves to determine the carrier vs line distance profile along each line. The mapping process is identical for all simu lations using the same unit cell. In other words, the exciton diffusion length does not imp act the mapping process. Because carriers are mapped to the nearest line, the (x, y) coordinate s of the point are the onl y factor not how many carriers are generated there. However, change s in the exciton diffusion length do impact the number of carriers mapped to each point on the li nes, and therefore impact the shape of the carrier distribution along those lines. To clarify the mapping process, Figure 2-54 displays the photogenerated carrier distribution for a simulated cell with an exciton diffusion length of 20 nm. The dotted lines shown in the illustration of Figure 2-53 are overlayed on this contour plot, seen as dashed white lines. Line 1, Line 2, and Line 3 are shown as solid white lines and represent the borders between the ZnO and P3HT regions. Note that the xand y-ax es are not on the same scale. The diagonal dashed lines on the plot are at 45 a ngles through the unit cell, although scaling makes them appear to be less sloped. The behavior of the three carrier generation lines depends on the ex citon diffusion length of the simulated cell and the lateral di stance along the line. For Line 1 (x 0.015 m, y = 0.26 m), the carrier distribution is nearly constant. As seen in Figure 2-54, this line collects carriers 65
PAGE 66
from the P3HT region beyond the tip of the nanorod, as well as a small area within nanorod. This area inside the nanorod decreases as the x-coordinate increases because portions of this region begin mapping to Line 2. This decrease results in a negligible decrease in the number of carriers, however, because the ZnO abso rbs poorly in comparison to the active P3HT region included in the calculation. This effect holds true for any exciton diffusion length that is considered in this study, because the P3HT region will always absorb much more strongly than the ZnO region. Due to this, the carrier distributi on for Line 1 is approximated as a constant for all simulations. Along Line 2 (x = 0.015 m, 0.025 m y 0.26 m), photogeneration increases steadily at low values of y because a larger section of the absorbing region is included at each point. Seen in Figure 2-54, this occurs for approximately one LD, from 0.025 y 0.045 m. Beyond that point, the new regions being added are not absorbing regions, so even though the size of the area being summed is larger, this larger area is not contributing a significant number of carriers. This is compounded by the fact that the LD region at this greater depth produces a lower number of carriers due to a drop in the number of photons remaining in the cell. This exponential decay of absorption becomes the dominant contribution to the number of carrie rs along the line, and continues to the tip of the nanorod. The point where the distribution along Line 2 turns from linear growth to exponential decay depends on the exciton diffusion length of the simulated cell. For Line 3 (x 0.015 y = 0.025 m), the number of photogenerated carriers initially shows a linear increase with x for the same reasons as Line 2. As the lin e is traversed, additional area is being mapped to this line, which increases the number of carrier s contributed. After a distance of LD along the lin e, the distribution becomes nearly c onstant. This occurs because the 66
PAGE 67
new area being included in each summation contai ns relatively few carriers due to lack of absorption in these regions. The total amount of photogeneration in the simulated cell is dependent on the exciton diffusion length specified in the simulation. Th is parameter defines the size of the stronglyabsorbing area in the P3HT polymer. Figure 2-55 displays the total count of photogenerated carriers for simulated cells with varying excit on diffusion lengths defined from 10 nm up to the full polymer region of the cell. In addition to dictating the to tal number of carriers in the ce ll, variations in the exciton diffusion length also change the distribution of these carriers. Figure 2-56 displays the line sources used in Medici to simulate photoabsor ption based on the model described previously. Note that the curves seen in Figure 2-56 are not the actual summations of carriers calculated from the contour plots, they are the fit lines a pplied to Medici with the form shown in Equation 2-6. As evidence of the quality of the fit, Figur e 2-56C displays the true carrier summations for Line 2 of a cell with LD = 20 nm, along w ith the fit curve applied in Medici. As desccribed previously, photogeneration in Line 1, for 0 x < 0.015 m, is set as constant in all cases. In Line 2, shown in Figue 2-56B, carrier genera tion increases nearly linearly until a certain point wher e the exponential decay becomes th e dominant effect. For Line 3, carrier generation again incr eases nearly linearly for x 0.015 m before reaching a point where it becomes approximately constant. Unlike Line 1, the nearly-const ant region for Line 3 was not assumed to be exactly constant, and a ll parameters for the e xponential-linear equation were calculated. For the 10 nm and 20 nm cases, this region shows a lower slope than in the 40 nm and full cell cases. 67
PAGE 68
Interestingly, the curves in Figure 2-56 s how regions where the carrier generation is higher for the 40 nm scenario than for the full cell scenario. This effect is observed to a very small degree near x = 0.035 m in Figure 2-56A and more noticeably near y = 0.06 m in Figure 2-56B. This is not simply an inconsistency due to curve fitting, this shift occurs in the absorption profiles that were summed in the cells. To clearly see the regions where the 40 nm a nd full cell simulations produced inconsistent carrier generation, the generation pr ofile for the 40 nm absorption case was subtracted from the profile for the full cell case, and contour plots of this difference are shown in Figure 2-57. In the figure, positive values represent areas wher e the full absorption simulation produced more carriers than the 40 nm simulation. Figure 2-57A shows that the positive-valued regions have extremely large values compared to the negative-valued regions, accounting for the increase in total absorption in the cell. Fi gure 2-57B is shown with a sma ller scale so that positive and negative-valued areas are clearly displayed. Th e green areas show no equal carrier generation for the two simulations, reds and yellows show areas where the full absorption case produced more carriers, and blues and pur ples show areas where the 40 nm simulation produced more carriers. In general, the ZnO phase shows no difference between the two simulations, the P3HT region beyond the 40 nm LD shows greater carrier generation in the full absorption simulation, and portions of the diffusion length region show greater generation in the 40 nm case. In the region beyond the tip of the nanorod, the plot sh ows a zero value for 4 nm, then a negative area with a value of approximately -2 x 1020 pairs/cm3 stretches for 16 nm before jumping to a positive value of nearly 6 x 1021 pairs/cm3. To the side of the nanorod, the contour plot shows a zero value for nearly 10 nm into the P3HT, but then turns negative for 26 nm. In the P3HT bulk 68
PAGE 69
region, there is no difference betw een the simulations up to x = 0.022 m. Beyond the interface for approximately 10 nm, values are all negati ve, indicating the 40 nm simulation generated more carriers. Above the interface in the ZnO phase there is a thin strip of 2 4 nm where there is an inconsistency. From x = 0.05 to 0.06 m, this strip shows a negative value. For x = 0.06 to 0.065 m, the strip shows a positive value. The full cell absorption simulation showed stronger carrier generation in the P3HT region beyond 40 nm from the ZnO interface. This is to be expected, as the full cell simulation allows this region to absorb any remaining carriers, wh ile the 40 nm simulation does not. However, the increased carrier generation for the 40 nm simulati on within the LD region is puzzling. Photons entering this region have the ex act same absorption history in both simulations. They have passed through a weakly-absorbing ZnO layer and entere d a strongly-absorbing P3HT region. For both simulations, this P3HT region has identical absorption properties for a path length of at least 40 nm. In the area within 40 nm of the side of the nanorod, the two simulations have identical absorption properties for 300 nm into the P3HT region. There is never a point in the simulation area where the 40 nm simulation displays a stronger absorption coefficient than the full cell simulation. With that in mind, this in consistency is considered to be a numerical anomaly in Medici during the convergence process. It is a strange occurance, but due to the difference in value between the negative regions and positive regions, as seen in Figure 2-57A, the total number of carriers in the cells is st ill increasing as the absorption area increases. There are two data points ignored in the graphs in Figure 2-56 the points at the base and tip corners of the nanorod. The carrier concentra tion at these points is pl otted against the exciton diffusion length in Figure 2-58. The point designated as the nanorod tip has coordinates of (0.015, 0.260) and shows a growth as the exciton di ffusion length increases. This point collects 69
PAGE 70
all generated carriers in the P3HT region for x 0.015 m and y 0.260 m. As the exciton diffusion length increases, this region produces a larger number of carriers because a larger portion of the region contains a strong absorpti on coefficient. The point designated as the nanorod base is the point at (0.015, 0.025) where the edge of the nanorod meets the base ZnO layer. This point colle cts all photogenerated carri ers for the region of x 0.015 m and y 0.025 m, which is composed entirely of ZnO. Because the absorption coefficient of ZnO is held constant in all simulations, the number of carriers collected at this point remains constant regardless of the exciton diffusion length. With this procedure established for measuri ng carrier generation across the simulated unit cell and compressing this distributi on to line sources in Medici, th e J-V response of cells can be determined. J-V Curves The carrier generation line sour ces shown in Figure 2-56 and 258 were input to Medici to generate J-V curves for the simulated cells. As described previously, these line sources of carriers restrict the existence of free electrons and holes to the inte rface between ZnO and P3HT, which matches the physics of excitonic solar cells. The resulting J-V curves were compared with the published data and are shown in Figure 2-59. As anticipated, the shortcircuit current density increases with increases in the exciton diffu sion length. Interestingly, this trend does not continue to the fully absorbing unit cell, wher e the short-circuit curre nt density drops by 0.2 mA/cm2 from the level of th e LD = 40 nm cell. All simulations failed to achieve the s hort-circuit current density of 2.2 mA/cm2 reported for the published cell. This is interesting cons idering that the simula ted cell was allowed to generate photocurrent for speci fied diffusion lengths ranging to the full cell area. The opencircuit voltage of the real cell was well appr oximated by the experiments due to the 0.85 eV 70
PAGE 71
energy band gap specified in P3HT. Fill factors for the simulated cells were higher than for the real cell. Table 2-17 displays a summary of the performance characteristics of the real and simulated cells shown in Figure 2-59. The simulated cell using a 40 nm exciton diffu sion length showed the best match to the real cell from literature. The VOC was nearly a perfect match, and the efficiency was approximately 5% higher than the literature value. However, the fill factor and JSC for the two cells did not match. The simulated cell showed a fill factor that was 50% higher and a JSC that was 33% lower than the real cell. Further simulations were performed using the 40 nm LD simulation as a basis. The effect of changing the carrier mobilitie s by up to two orders of magnit ude is shown in Figure 2-60, and the resulting performance measures are shown in Table 2-18. While four of the curves appeared as nearly identical, the curve corresponding to a 100x reduction in mobility began to show a noticeable drop in fill factor. Attempts to reduce the mobility by an extra order of magnitude failed due to an inability of Medici to converg e around such small values. Because Medici is designed for the simulation of inorganic semiconductors, mobility values of p = 1 x 10-5 and n = 1 x 10-6 cm2/V-s are more than 6 orders of magnitude lower than the program was designed to handle. This leads to convergence difficulties due to the low current values in the cell. The simulation using a 100x reduction in mobilit y values shows a better match to the real cell than any other simulation. The VOC and efficiency differ by only about 2% between the two cells. The JSC is again off by 33% from the real cell, as that value did not change with the variation in mobility. With the lower mobility, however, the fill factor dropped to within 40% of the published value. 71
PAGE 72
Because these simulations fail to produce a short-circuit current on par with the real cell, the P3HT absorption coefficient was adjusted to a hi gher value. This gene rates more carriers in the cell, which is expected to generate more photocurrent at all applied biases. The absorption coefficient is a parameter that is easily measur ed experimentally, but th ere is some physical of justification for this adjustment. The unit cell in Medici is defined by perfectly vertical nanowires with flat tips and perfect spacing between them. From the image in Figure 2-14, this is far from reality. Although care was taken to properly estimate the nanowire le ngth, width, and inter-wir e spacing, these values are purely averages taken from th e visible portion of the cross-se ctional view. The real wire array contains overlapping non-vertical wires that create non-uniform spaces between them where P3HT exists. In addition, these randomly angled structures would cr eate light reflection and refraction patterns that are not considered in Medici. It is not unreas onable to expect that this disordered array of nanowires could create light-trapping effects in the film, where incident rays are refracted in such a way to increase th eir residence time in the film and increase the degree to which they are absorbed. From this, it is not unreasonable to exp ect that this increased absorption occurs to some degree in the polymer regions within an exciton diffusion length from the material interfaces. For these reasons, the absorption coefficient of P3HT was increased by 50% in a simulation. The photogenerated car rier distribution for this simulation is compared to the previous 40 nm LD simulation, with an absorp tion coefficient of 100%, with the resulting surface plots shown in Figure 2-61. As expected this resulted in a larger number of photogenerated carriers in the film. 72
PAGE 73
Although the total number of phot ogenerated carriers increase d, there was not a uniform increase at all points in the film. Specificall y, in the column of polym er within one exciton diffusion length of the side of the nanorod, the number of carriers decreases quickly for the simulation with an increased absorption coeffici ent. This is because the number of photons penetrating deeper into the film is reduced due to the stronger absorption. This is clearly seen in Figure 2-62, which shows a comparison of carrier generation for the increased absorption case and the standard absorption case. The carrier distributions were subtra cted, and green regions show no change in absorption between the two simulations. Red, ora nge, and yellow regions show areas where the 150% absorption coefficien t resulted in stronger ab sorption, and this is primarily contained within a range of 40 nm (LD) from the upper surface of the ZnO-P3HT interface. Blue regions represent areas where the 150% absorption coefficient resulted in less absorption, and this is limited to the exciton di ffusion length region along the nanorod edge. This is due to strong absorption near the surface of the P3HT region, leaving fewer photons available for absorption in the deeper region. Although there are regions of increased and d ecreased carrier genera tion in the cell, the overall number of generated carriers increased by approximately 3 x 1024 pairs/cm3 over the unit cell. This additional charge generation did not tr anslate into additional photocurrent as expected, however. Figure 2-63 shows J-V curves for the two cells, calculated with p = 1 x 10-4 cm2/V-s. Despite the increased carriers density for th e 150% absorption coefficient simulation, the short-circuit current density decreased by 0.23 mA/cm2. This is very similar to the effect seen when the full P3HT area was allowed to contribute to absorption. In fact, a comparison of those cells shows extremely similar properties. Note that the full cell simulation was performed with the standard values fo r mobility in the P3HT regions, while the two 40 nm simulations were 73
PAGE 74
performed with a these mobility values reduced by two orders of magnitude. This explains the difference in fill factor shown in Table 2-19, bu t Figure 2-60 demonstrated that this mobility reduction does not impact the short-circui t current density in these simulations. It is believed that the reason for the reduced performance associated with this increase in carrier generation is an effect of increased annihilation near the material interface. This large number of carriers is being produced along a very narrow regi on directly at the material interface. This results in many free electrons and holes in a small area, which could cause attractive forces between them to promote anni hilation immediately follo wing their generation. There seems to be a breaking point between 3.14 x 1025 and 3.23 x 1025 pairs/cm3 of total carrier density, as this is the limit where the re duced charge transport seems to take hold. The short-circuit current density drops by about 15%, the VOC remains constant, and the fill factor drops by less than 10%. Summary of Results The device modeling program Medici was used for simulation of hybrid solar cells. Several nuances of the program were probed with a test model before attempts were made to simulate an existing cell from the literature. After probing the effect of various parameters, a two-step simulation strategy was adopted that separated photon absorption and charge transport. This m odel more accurately approximates exciton dynamics by eliminating free electrons in the polymer regions of the cell and applying free carrier generation directly at the ma terial interface lines. Simulations of this type initially showed lo w performance, but increases in the exciton diffusion length up to 40 nm provided increased charge generation and cu rrent flow. Further increases in the number of carriers generated re duced the current in the simulation, presumably due to increased charge attract ion and annihilation. It was found that decreasing the carrier 74
PAGE 75
mobilities in P3HT from the field effect mobility values f ound in literature resulted in a decrease in the fill factor to levels similar to the real cell. 75
PAGE 76
A B Figure 2-1. Hybrid solar cell (A) and correspond ing unit cell (B) used fo r device simulation. Table 2-1. P3HT Properties for Device Simulations. Property Value Reference Doping Density p-type 5x1016 cm-2[67] Permittivity 3 [61] NC 2x1018 [69] NV 2x1019 [69] E g 1.7 eV [70] Electron Affinity 3.15 eV [71] e 0.01 cm2/Vs 10% of h h 0.1 cm2/Vs [68] Table 2-2. CdSe Properties for Device Simulations. Property Value Doping Density n-type 6x1016 cm-2 Permittivity 10.2 NC 2x1018 NV 2x1019 E g 1.74 eV Electron Affinity 3.75 eV e 650 cm2/Vs h 30 cm2/Vs Table 2-3. Electrode Propert ies for Device Simulations. Electrode Optical Properties Reference ITO Transparent [72] Al 90% Reflectance [73] 76
PAGE 77
Wavelength ( m) 0.2 0.4 0.6 0.8 1.0 Absorption Coefficient (cm-1) 0.0 5.0e+4 1.0e+5 1.5e+5 2.0e+5 2.5e+5 P3HT ZnO CdSe Figure 2-2. Wavelength-de pendent absorption coefficient data used in simulations [66, 74]. Figure 2-3. J-V curves for simulated hybrid sola r cells with different methods of specifying doping density. See Table 2-4 for a description of the differences. 77
PAGE 78
Voltage (V) 0.000.050.100.150.200.250.30 Current Density (mA/cm2) -10 -8 -6 -4 -2 0 3E16 4E16 5E16 6E16 7E16 Figure 2-4. Simulated J-V curves showing the e ffect of doping density in the CdSe nanorods. The legend shows the n-type doping density measured in cm-2. 78
PAGE 79
Table 2-4. Impurity profile inputs for simulations. Filename Impurity Profile Input Statement impurity_1 PROFILE P-TYPE REGI ON=P3HT UNIFORM N.PEAK=5E16 PROFILE N-TYPE REGION=Cd Se UNIFORM N.PEAK=6E16 impurity_2 PROFILE P-TYPE Y.MIN=0 Y.MAX=0.100 UNIFORM N.PEAK=5E16 PROFILE N-TYPE Y.MIN=0.01 Y. MAX=0.100 X.MIN=0 X.MAX=0.005 + UNIFORM N.PEAK=6E16 impurity_3 PROFILE P-TYPE Y.MIN=0 Y.MAX=0.01 UNIFORM N.PEAK=5E16 PROFILE P-TYPE Y.MIN=0.01 Y.MAX=0.100 X.MIN=0.005 X.MAX=0.015 + UNIFORM N.PEAK=5E16 PROFILE N-TYPE Y.MIN=0.01 Y. MAX=0.100 X.MIN=0 X.MAX=0.005 + UNIFORM N.PEAK=6E16 impurity_4 PROFILE P-TYPE Y.MIN=0 Y.MAX=0.01 UNIFORM N.PEAK=5E16 PROFILE P-TYPE Y.MIN=0.01 Y.MAX=0.100 X.MIN=0.005 X.MAX=0.015 + UNIFORM N.PEAK=5E16 PROFILE N-TYPE Y.MIN=0.01 Y. MAX=0.100 X.MIN=0 X.MAX=0.005 + UNIFORM N.PEAK=1E16 impurity_5 PROFILE P-TYPE Y.MIN=0 Y.MAX=0.01 UNIFORM N.PEAK=5E16 PROFILE P-TYPE Y.MIN=0.01 Y.MAX=0.100 X.MIN=0.005 X.MAX=0.015 + UNIFORM N.PEAK=5E16 PROFILE P-TYPE Y.MIN=0.01 Y.MAX=0.100 X.MIN=0 X.MAX=0.005 + UNIFORM N.PEAK=5E16 PROFILE N-TYPE Y.MIN=0.01 Y.MAX=0.100 X.MIN=0 X.MAX=0.005 + UNIFORM N.PEAK=6E16 Figure 2-5. Variation of open circuit volta ge with doping density of CdSe nanorods. 79
PAGE 80
Voltage (V) -0.10 -0.05 0.00 0.05 0.10 Current Density (mA/cm2) -20 -15 -10 -5 0 5 10 15 20 20 nm 30 nm 40 nm 50 nm 100 nm 200 nm 300 nm 500 nm 1 m Figure 2-6. Simulated J-V curves with varying unit cell thickness. Figure 2-7. Solar cell parameters for unit 12 nm wide unit cells with varying cell thickness. 80
PAGE 81
Table 2-5. Solar cell performance measures for un it cells of 12 nm width a nd varying thickness. Thickness (nm) (%) VOC (V) JSC (mA/cm2) FF 11 0.0000 0.0000 1.251 0.0003 20 0.0000 0.0009 2.257 0.0000 30 0.0052 0.0067 3.106 0.2533 40 0.0172 0.0163 3.899 0.2699 50 0.0335 0.0252 4.627 0.2873 60 0.0507 0.0317 5.297 0.3019 70 0.0669 0.0361 5.916 0.3131 80 0.0817 0.0393 6.491 0.3203 90 0.0954 0.0418 7.023 0.3250 100 0.1082 0.0437 7.513 0.3295 120 0.1310 0.0466 8.368 0.3360 150 0.1598 0.0497 9.394 0.3422 200 0.1966 0.0531 10.625 0.3484 300 0.2456 0.0568 12.150 0.3559 500 0.2975 0.0597 13.791 0.3614 1000 0.3275 0.0588 15.594 0.3572 Voltage (V) 0.0 0.2 0.4 0.6 0.8 Current Density (mA/cm2) -10 -8 -6 -4 -2 0 2 4 2 nm 5 nm 10 nm 15 nm 20 nm 25 nm 30 nm Figure 2-8. J-V curves for hybrid cells with varying nanorod width. 81
PAGE 82
Voltage (V) 0.00.10.20.30.40.50.60.70.8 Current Density (mA/cm2) -10 -8 -6 -4 -2 0 2 4 10 nm 11 nm 12 nm 13 nm 14 nm 15 nm 16 nm 17 nm 18 nm 19 nm 20 nm 21 nm 22 nm 23 nm 24 nm 25 nm Figure 2-9. J-V curves for simulated hybrid ce lls with CdSe nanorod half-thickness between 10 and 25 nm. Figure 2-10. Solar cell performance measures for simulated hybrid cells w ith varying CdSe halfwidth. 82
PAGE 83
Table 2-6. Performance measures for simulated hybrid cells with varying CdSe half-width. CdSe Width (nm) Voc (V) Jsc (mA/cm2) FF (%) 2 0.044 7.513 0.330 0.11 5 0.253 9.658 0.623 1.52 10 0.410 8.575 0.661 2.32 11 0.435 9.488 0.670 2.77 12 0.454 9.453 0.669 2.87 13 0.470 9.421 0.668 2.96 14 0.485 9.391 0.666 3.03 15 0.487 6.461 0.675 2.12 16 0.512 9.338 0.662 3.16 17 0.523 9.315 0.659 3.21 18 0.533 9.295 0.658 3.26 19 0.542 9.276 0.656 3.30 20 0.535 5.745 0.677 2.08 21 0.558 9.243 0.654 3.37 22 0.564 9.229 0.653 3.40 23 0.553 5.414 0.679 2.03 24 0.576 9.204 0.652 3.45 25 0.580 9.194 0.652 3.48 30 9.071 4.93 Voltage (V) 0.000.050.100.150.200.250.300.350.40 Current Density (mA/cm2) -10 0 10 20 30 40 50 Center, Y=-2.5 Center, Y=-5.0 Left Edge, Y=-2.5 Right Edge, Y=-2.5 Figure 2-11. J-V curves for hybrid solar cell s with different light source specifications. 83
PAGE 84
Figure 2-12. Illustration of the PHOTOGEN command in Medici. Figure 2-13. SEM images of (a) Zn O nanofibers and (b) nanofiber and P3HT composite films. Reprinted with permission from D.C. Olson, J. Piris, R.T. Collins, S.E. Shaheen, D.S. Ginley, Thin Solid Films 496 (2006) 26 Figure 2 (a), (b) p. 28. 84
PAGE 85
Figure 2-14. J-V curve for a real ZnO:P3HT solar cell (solid line) to be used for verification of Medici simulations. Reprinted with perm ission from D.C. Olson, J. Piris, R.T. Collins, S.E. Shaheen, D.S. Ginley, Thin Solid Films 496 (2006) 26 Figure 3 p. 28. Voltage (V) -0.2-0.10.00.10.20.30.40.5 Current Density (mA/cm2) -3 -2 -1 0 1 2 Power Density (mW/cm2) -0.9 -0.6 -0.3 0.0 0.3 0.6 Current Density Power Density Figure 2-15. J-V and P-V curves for the real solar cell fabricated by Olson et al. [75] Table 2-7. Cell performance measures from pu blished and digitally converted J-V curves. Curve VOC (V) JSC (mA/cm2) FF (%) Published 0.44 2.2 0.56 0.53 Converted 0.44 2.2 0.57 0.55 85
PAGE 86
Table 2-8. ZnO properties used for hybrid solar cell simulation. Property Value Doping Density n-type 5 x 1017 cm-2 Permittivity 9.0 Nc 2 x 1018 Nv 2 x 1019 E g 3.3 eV Affinity 4.0 eV e 50 cm2/V-s h 5 cm2/V-s Figure 2-16. Unit cell used for simulations of ZnO/P3HT hybrid cells. Device areas area a) ITO electrode, b) ZnO, c) P3HT photocurrent generating region, d) P3HT non-generating region, and e) Ag electrode. 86
PAGE 87
Figure 2-17. Medici unit cell used for device simu lation. The blue area is ZnO, the red area is the photogenerating region of P3HT, and the green area is the non-generating region of P3HT. 87
PAGE 88
Voltage (V) -0.20.0 0.2 0.4 0.6 0.8 Current Density (mA/cm2) -12 -10 -8 -6 -4 -2 0 2 Real Cell tsf496_1 tsf496_2 tsf496_4 Figure 2-18. Simulated J-V curves for ZnO:P3HT solar cell using two P3HT regions. The first region is the exciton diffusion length region, where all properties are set as shown in Table 2-1. The second region is the non-ge nerating region outside of the diffusion length, with varying properties In tsf496_1, the absorption co efficient is set to zero. In tsf496_2, the properties remain the same as the exciton diffusion length region. In tsf496_4 the electron mobility is reduced by an order of magnitude (0.0001 cm2/Vs). 88
PAGE 89
Voltage (V) -0.20.00.20.40.60.81.0 Current Density (mA/cm2) -5.0 -4.5 -4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 90% Reflectance 50% Reflectance 0% Reflectance Figure 2-19. J-V curves for simulated cells with zero absorption in the P3HT2 region and varying reflectance from the Ag electrode. 89
PAGE 90
Voltage (V) -0.20.00.20.40.60.81.0 Current Density (mA/cm2) -4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 tsf496_05n = 0.001 cm2/V-s, p = 0.01 cm2/V-s tsf496_07's reduced by 50% in both P3HT regions tsf496_08's reduced by 50% in P3HT2 region tsf496_09's reduced by 90% in both P3HT regions Figure 2-20. Simulation results showing the effect of changing charge mobilities in the P3HT regions. Table 2-9. Performance measures for simulated cells with varying carrier mobility. Mobility (cm2/V-s) JSC (mA/cm2) VOC (V) FF (%) n = 0.01 p = 0.001 3.61 1.04 0.57 2.12 n = 0.005 p = 0.0005 3.58 1.04 0.54 2.00 n = 0.01 p = 0.0005 3.58 1.04 0.57 2.13 n = 0.001 p = 0.0001 3.44 1.04 0.46 1.64 90
PAGE 91
Voltage (V) -0.20.00.20.40.60.81.0 Current Density (mA/cm2) -4 -3 -2 -1 0 1 2 Real Cell tsf496_09, LD = 10 nm tsf496_11, LD = 9 nm tsf496_12, LD = 8 nm tsf496_13, LD = 7 nm tsf496_14, LD = 6 nm tsf496_15, LD = 5 nm Figure 2-21. J-V curves for simulated cel ls with varying exciton diffusion length. Figure 2-22. Extrapolations to estimate VOC for simulated cells with varying exciton diffusion length. 91
PAGE 92
Figure 2-23. Solar cell performance measures fo r simulated cells with varying exciton diffusion lengths. Table 2-10. Performance measures for simulate d cells with varying exciton diffusion lengths. LD (nm) Voc (V) Jsc (ma/cm2) FF (%) 10 1.04 3.44 0.459 1.64 9 1.04 3.17 0.468 1.54 8 1.05 2.90 0.473 1.44 7 1.05 2.61 0.482 1.32 6 1.06 2.32 0.487 1.20 5 1.07 2.02 0.492 1.06 92
PAGE 93
Voltage (V) -0.20.00.20.40.60.81.0 Current Density (mA/cm2) -3 -2 -1 0 1 2 Real Cell P3HT Doping 5x1016 cm-3 P3HT Doping 1x1016 cm-3 P3HT Doping 5x1015 cm-3 P3HT Doping 1x1015 cm-3 P3HT Doping 5x1014 cm-3 P3HT Doping 1x1014 cm-3 Figure 2-24. Real and simulated J-V curves for cells with varying P3HT doping density. Voltage (V) -0.20.00.20.40.60.81.0 Current Density (mA/cm2) -2.5 -2.0 -1.5 -1.0 -0.5 0.0 Real Cell ZnO Doping Density 5x1017 ZnO Doping Density 1x1017 ZnO Doping Density 5x1016 ZnO Doping Density 2x1016 ZnO Doping Density 1x1016 ZnO Doping Density 9x1015 ZnO Doping Density 7x1015 ZnO Doping Density 5x1015 ZnO Doping Density 1x1015 Figure 2-25. Real and simulated J-V curves fo r hybrid cells with varying ZnO doping density. 93
PAGE 94
Voltage (V) -0.20.00.20.40.60.81.0 Current Density (mA/cm2) -2.5 -2.0 -1.5 -1.0 -0.5 0.0 Real Cell P3HT NC=2x1018 NV=2x1019 P3HT NC=2x1019 NV=2x1019 P3HT NC=2x1019 NV=2x1020 P3HT NC=2x1020 NV=2x1020 P3HT NC=2x1020 NV=2x1021 Figure 2-26. Real and simulated J-V curves for hybrid solar cells with varying P3HT density of states. Voltage (V) -0.20.00.20.40.60.81.01.21.4 Current Density (mA/cm2) -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 n=0.001 cm2/V-s, p=0.01 cm2/V-s n=0.0001 cm2/V-s, p=0.001 cm2/V-s n=0.00001 cm2/V-s, p=0.0001 cm2/V-s Figure 2-27. Simulated J-V curves fo r hybrid solar cells with varying P3HT mobility. 94
PAGE 95
Voltage (V) -0.20.00.20.40.60.81.01.21.4 Current Density (mA/cm2) -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 P3HT Doping: 5x1016 P3HT Doping: 1x1016 P3HT Doping: 5x1015 P3HT Doping: 1x1015 P3HT Doping: 5x1014 P3HT Doping: 1x1014 Figure 2-28. Simulated J-V curves for hybrid cells with varying P3HT doping concentrations. Figure 2-29. Energy band diagram for P3HT ZnO hybrid solar cells. 95
PAGE 96
Voltage (V) -0.2-0.10.00.10.20.30.40.50.6 Current Density (mA/cm2) -4 -2 0 2 4 6 8 10 Real Cell tsf496_45 tsf496_53 Figure 2-30. J-V curves for simulated cells w ith varying energy band ga p in the active layers. Wavelength ( m) 0.00.20.40.60.81.01.21.4 Absorption Coefficient (cm-1) 0.0 5.0e+5 1.0e+6 1.5e+6 2.0e+6 2.5e+6 abs_none abs_p3ht abs_i_zno Figure 2-31. Absorption coeffici ent vs. wavelength as tabulated in Medici for input files corresponding to zero absorption (red), P3HT (green), and ZnO (teal). 96
PAGE 97
Wavelength ( m) 0.20.40.60.81.01.21.41.61.8 Normalized Flux 0.0 0.2 0.4 0.6 0.8 1.0 1.2 AM1.5 Solar Spectrum Figure 2-32. AM1.5 solar spectrum. Wavelength ( m) 0.0 0.2 0.4 0.6 0.8 1.0 Photogenerated Carriers (C m-1s-1) 0.0 5.0e-14 1.0e-13 1.5e-13 2.0e-13 2.5e-13 3.0e-13 3.5e-13 Absorption Coefficient (cm-1) 0.00 1.00e+5 2.00e+5 1.25e+6 1.50e+6 1.75e+6 2.00e+6 2.25e+6 Generated Carriers abs_p3ht abs_zno Figure 2-33. Carrier generation (left axis) in simulated solar cells plotted with absorption coefficients for P3HT and ZnO (right axis). 97
PAGE 98
Voltage (V) -0.10.00.10.20.30.40.5 Current Density (mA/cm2) -4 -2 0 2 4 Medici-assigned values Fully specified values Figure 2-34. J-V curves for simula ted cells showing the original P3HT absorption profile and an edited absorption profile limiting absorption between 0.2 and 0.3 m. Voltage (V) -0.10.00.10.20.30.40.5 Current Density (mA/cm2) -3 -2 -1 0 1 tsf496_45 1 stage tsf496_47 2 stages tsf496_48 4 stages tsf496_49 6 stages tsf496_50 8 stages tsf496_51 10 stages Figure 2-35. J-V curves for simulated cells w ith exciton diffusion length of 10 nm and multistage absorption regions. 98
PAGE 99
P3HT Thickness 0 5 10 15 20 Fraction of Maximum Absorption Coefficient 0.0 0.2 0.4 0.6 0.8 1.0 = 0.001 = 2 = 4 = 6 = 8 = 10 Figure 2-36. Examples of cumu lative distribution function with mean of 10 nm and a range of standard deviation. Table 2-11. Absorption data a nd short-circuit current for graded absorption simulations. 0 nm 1 nm 2 nm 3 nm 4 nm 5 nm 6 nm 10 nm 0-2 nm 100% 100% 100% 100% 100% 100% 100% 100% 2-4nm 100% 100% 100% 100% 100% 90% 90% 90% 4-6 nm 100% 100% 100% 100% 90% 90% 80% 90% 6-8 nm 100% 100% 90% 80% 80% 70% 70% 70% 8-10 nm 100% 80% 70% 60% 60% 60% 60% 60% 10-12 nm 0% 20% 30% 40% 40% 40% 40% 40% 12-14nm 0% 0% 10% 20% 20% 30% 30% 30% 14-16 nm 0% 0% 0% 10% 10% 10% 20% 20% 16-18 nm 0% 0% 0% 0% 0% 10% 10% 10% 18-20 nm 0% 0% 0% 0% 0% 0% 0% 0% Average 50% 50% 50% 51% 50% 50% 50% 51% JSC (mA/cm2) 1.62 1.62 1.66 1.67 1.68 1.69 1.69 1.71 99
PAGE 100
Voltage (V) -0.10.00.10.20.30.40.5 Current Density (mA/cm2) -3 -2 -1 0 1 2 3 = 0 nm = 1 nm = 2 nm = 3 nm = 4 nm = 5 nm = 6 nm = 10 nm Figure 2-37. Simulated J-V curves for cells with graded absorption profiles. Voltage (V) -0.10.00.10.20.30.40.5 Current Density (mA/cm2) -30 -20 -10 0 10 20 30 ZnO:P3HT with 10 nm absorption region ZnO:CIS with 10 nm absorption region ZnO:P3HT with full absorption region ZnO:CIS with full absorption region ZnO:CIS with 10 nm absorption region and Eg = 0.85 eV Figure 2-38. Simulated J-V curves with CIS replacing P3HT. 100
PAGE 101
Table 2-12. Materials properties for P3HT and CIS used in cell simulations. Property P3HT CIS Doping Density p-type 5x1016 cm-2 p-type 5x1016 cm-2 Permittivity 3 13.6 NC 2x1018 3x1018 NV 2x1019 1.5x1019 E g 1.7 eV 1.04 eV Electron Affinity 3.15 eV 3.93 eV e 0.1 cm2/Vs 300 cm2/Vs h 0.01 cm2/Vs 30 cm2/Vs Voltage (V) -0.10.00.10.20.30.40.50.6 Current Density (mA/cm2) -10 -5 0 5 10 tsf496_64 ZnO:CIS tsf496_68 P3HT Permittivity tsf496_69 P3HT Density of States tsf496_70 P3HT Affinity tsf496_71 P3HT Mobilities tsf496_72 P3HT Absorption Figure 2-39. Simulated J-V curves for ZnO:CIS solar cells with an individual material property changed to the P3HT value. Table 2-13. Performance measures for ZnO:CIS cells with an individual ma terial property set at the P3HT value. Simulation Parameter Adjusted Voc (V) JSC (mA/cm2) FF (%) 64 None 0.439 7.774 0.772 2.633 68 Permittivity 0.443 7.747 0.778 2.669 69 Density of States 0.441 7.773 0.774 2.655 70 Affinity 0.487 7.654 0.485 1.808 71 Mobility 0.545 6.309 0.662 2.277 72 Absorption 0.421 4.105 0.762 1.317 101
PAGE 102
Voltage (V) -0.10.00.10.20.30.40.50.6 Current Density (mA/cm2) -10 -5 0 5 10 15 20 tsf496_72 P3HT Absorption tsf496_73 P3HT Absorption, Permittivity tsf496_74 P3HT Absorption, Density of States tsf496_75 P3HT Absorption, Affinity tsf496_76 P3HT Absorption, Mobilities Figure 2-40. Simulated J-V curves for ZnO:CIS solar cells with the P3HT absorption spectrum applied. Table 2-14. Performance measures for si mulated ZnO:CIS solar cells with the P3HT absorption spectrum. Simulation Properties Varied VOC (V) JSC (mA/cm2) FF (%) 72 Absorption 0.421 4.105 0.762 1.317 73 Absorption Permittivity 0.427 4.065 0.765 1.328 74 Absorption Density of States 0.423 4.105 0.765 1.328 75 Absorption Affinity 0.467 3.980 0.493 0.916 76 Absorption Mobility 0.517 3.293 0.657 1.118 102
PAGE 103
Voltage (V) -0.10.00.10.20.30.40.50.60.7 Current Density (mA/cm2) -10 -5 0 5 10 15 20 tsf496_75 P3HT Absorption, Affinity tsf496_77 P3HT Absorption, Affinity, Permittivity tsf496_78 P3HT Absorption, Affinity, Density of States tsf496_79 P3HT Absorption, Affinity, Mobilities Figure 2-41. Simulated ZnO:CIS solar cells with P3HT values for absorption coefficient and electron affinity. Table 2-15. Performance measures for simulated ZnO:CIS solar cells with P3HT values for absorption coefficient and electron affinity. Simulation Properties Varied VOC (V) JSC (mA/cm2) FF (%) 75 Absorption Affinity 0.467 3.980 0.493 0.916 77 Absorption Affinity Permittivity 0.455 3.979 0.502 0.909 78 Absorption Affinity Density of States 0.471 3.975 0.479 0.896 79 Absorption Affinity Mobility 0.613 3.439 0.199 0.419 103
PAGE 104
Voltage (V) -0.10.00.10.20.30.40.50.6 Current Density (mA/cm2) -10 -5 0 5 10 15 20 tsf496_80 P3HT Absorption, Affinity, Eg, Permittivity tsf496_81 P3HT Absorption, Affinity, Eg, Density of States tsf496_82 P3HT Absorption, Affinity, Eg, Mobilities Figure 2-42. Simulated ZnO:CIS solar cells with P3HT values for absorption coefficient, electron affinity, and energy band gap. Table 2-16. Performance measures for simulated ZnO:CIS solar cells with P3HT values for absorption coefficient, electr on affinity, and energy band gap. Simulation Properties Varied VOC (V) JSC (mA/cm2) FF (%) 80 Absorption Affinity, Eg Permittivity 0.265 3.633 0.418 0.402 81 Absorption Affinity, Eg Density of States 0.281 3.608 0.395 0.401 82 Absorption Affinity, Eg Mobility 0.423 2.164 0.159 0.145 104
PAGE 105
Voltage (V) -0.10.00.10.20.30.40.5 Current Density (mA/cm2) -10 0 10 20 p = 0.1 cm2/V-s p = 1 cm2/V-s p = 10 cm2/V-s p = 100 cm2/V-s p = 200 cm2/V-s p = 500 cm2/V-s p = 800 cm2/V-s p = 1000 cm2/V-s p = 5000 cm2/V-s p = 10,000 cm2/V-s Figure 2-43. Simulated J-V curves for ZnO:P3HT solar cells with varying carrier mobility. In all cases, carrier generation is allowed in the full P3HT region and the electron mobility is set to 10% of the hole mobility. Figure 2-44. Calculation method for estim ated second derivatives of J-V curves. 105
PAGE 106
Voltage (V) 0.00.10.20.30.40.50.60.7 J"est -4e-4 -2e-4 0 2e-4 4e-4 p = 0.1 cm2/V-s p = 1 cm2/V-s p = 10 cm2/V-s p = 100 cm2/V-s p = 200 cm2/V-s p = 500 cm2/V-s p = 800 cm2/V-s p = 1000 cm2/V-s p = 5000 cm2/V-s p = 10,000 cm2/V-s Figure 2-45. Estimated second derivative Jest for simulated ZnO:P3HT solar cells with varying carrier mobilities and carrier generation in the full P3HT region. In all cases, the electron mobility is set to 10% of the hole mobility. 106
PAGE 107
Voltage (V) -0.10.00.10.20.30.40.5 Current Density (mA/cm2) -4 -2 0 2 4 6 8 10 p = 0.1 cm2/V-s p = 1 cm2/V-s p = 10 cm2/V-s p = 100 cm2/V-s p = 200 cm2/V-s p = 500 cm2/V-s p = 800 cm2/V-s p = 1000 cm2/V-s p = 5000 cm2/V-s p = 10,000 cm2/V-s Figure 2-46. Simulated J-V curves for ZnO:P3HT solar cells with varying carrier mobilities and carrier generation in the 10-nm exciton di ffusion length region. In all cases, the electron mobility is set to 10% of the hole mobility. 107
PAGE 108
Voltage (V) 0 00 20 40 6 J"est -2e-4 -1e-4 0 1e-4 2e-4 3e-4 4e-4 5e-4 p = 0.1 cm2/V-s p = 1 cm2/V-s p = 10 cm2/V-s p = 100 cm2/V-s p = 200 cm2/V-s p = 500 cm2/V-s p = 800 cm2/V-s p = 1000 cm2/V-s p = 5000 cm2/V-s p = 10,000 cm2/V-s Figure 2-47. Estimated second derivative Jest for simulated ZnO:P3HT solar cells with varying carrier mobilities and carrier generation in the 10-nm exciton diffusion length region. In all cases, the electr on mobility is set to 10% of the hole mobility. 108
PAGE 109
Voltage (V) -0.10.00.10.20.30.40.5 Current Density (mA/cm2) -4 -2 0 2 4 tsf496_98 100% absorption tsf496_103 80% absorption tsf496_104 50% absorption tsf496_105 20% absorption Figure 2-48. Simulated J-V curves for ZnO:P3HT solar cells with varying absorption coefficients in P3HT. Voltage (V) -0.20.00.20.40.60.81.01.21.4 Current Density (mA/cm2) -4 -2 0 2 4 tsf496_98 Eg = 0.85 eV tsf496_106 Eg = 1.0 eV tsf496_107 Eg = 1.2 eV tsf496_108 Eg = 1.5 eV tsf496_109 Eg = 1.7 eV Figure 2-49. Simulated J-V curves for ZnO:P3HT solar cells with varying energy band gap. 109
PAGE 110
Voltage (V) 0 00 20 40 60 81 0 Current Density (mA/cm2) -5 -4 -3 -2 -1 0 1 Eg = 0.85 eV Eg = 1.0 eV Eg = 1.2 eV Eg = 1.5 eV = 100% = 80% = 50% Figure 2-50. Simulated J-V curves for ZnO:P3HT solar cells with varying band gap and absorption in the P3HT region. Voltage (V) -0.10.00.10.20.30.40.5 Current Density (mA/cm2) -4 -3 -2 -1 0 1 2 Real Data 100% absorption 80% absorption 50% absorption 40% absorption 30% absorption 20% absorption Figure 2-51. Simulated J-V curves for ZnO:P3HT solar cells with P3HT energy band gap of 1.0 eV and hole mobility of 500 cm2/Vs. 110
PAGE 111
A) X-Coordinate ( m) 0.000.010.020. 030.040.050.06 Y-Coordinate ( m) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 1e+20 1e+21 1e+22 B) 1e+19 1e+20 1e+21 1e+22 1e+23 0.014 0.016 0.018 0.020 0.022 0.024 0.026 0.05 0.10 0.15 0.20 0.25P h o t o g e n e r a t e d C a r r i e r s ( p a i r s / c m3)X C o o r d i n a t e ( m )Y C o o r d i n a t e ( m ) 1e+19 1e+20 1e+21 1e+22 1e+23 Figure 2-52. Photogenerated ca rrier distribution in pairs/cm3 for A) the full unit cell and B) the region of 13 nm x 27 nm along the edge of the ZnO nanorod. Note that the x and y axes do not follow the same scale. 111
PAGE 112
Figure 2-53. Carrier mapping scheme for two-stag e simulations. Gray regions are ZnO and blue regions are P3HT. Orange arrows represent the nearest interface points for photogenerated carriers in the region de fined by the dotted white lines. 112
PAGE 113
X-Coordinate ( m) 0.000.010.020.030.040.050.06 Y-Coordinate ( m) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 2e+21 4e+21 6e+21 8e+21 1e+22 Line 1 Line 2 Line 3 Figure 2-54. Photogenerated ca rrier distribution for a simu lated cell with LD = 20 nm. 113
PAGE 114
Figure 2-55. Cumulative number of photogenerated carriers in si mulated cells with varying exciton diffusion length. 114
PAGE 115
A) B) C) Figure 2-56. Photogenerated carr ier distribution along xand ycoordinates for models with varying exciton diffusion lengths. 115
PAGE 116
A) -1e+21 0 1e+21 2e+21 3e+21 4e+21 5e+21 6e+21 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40D i f f e r e n c e i n P h o t o g e n e r a t e d C a r r i e r s ( p a i r s / c m3)X C o o r d i n a t e ( m )Y C o o r d i n a t e ( m ) -1e+21 0 1e+21 2e+21 3e+21 4e+21 5e+21 6e+21 B) X Coordinate ( m) 0.000.010.020.030.040.050.06 Y Coordinate ( m) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 -4e+18 -2e+18 0 2e+18 4e+18 Figure 2-57. Contour plots of the photogenerate d carrier difference between the full absorption simulation and the 40 nm LD simulation. 116
PAGE 117
Figure 2-58. Photogenerated carriers at the tip (left axis) and base (right axis) corner points of the nanorod in simulated hybrid cells. Voltage (V) -0.10.00.10.20.30.40.5 Current Density (mA/cm2) -3 -2 -1 0 1 2 3 4 Real Data 10 nm 20 nm 40 nm Full Cell Figure 2-59. J-V curves for simulated sola r cells using line-sour ce carrier generation. 117
PAGE 118
Table 2-17. Cell performance measures for real and simulated solar cells. Cell JSC (mA/cm2) VOC (V) FF (%) Real 2.20 0.44 0.56 0.53 10 nm LD 0.59 0.44 0.84 0.22 20 nm LD 1.01 0.45 0.84 0.38 40 nm LD 1.47 0.45 0.84 0.56 Full Cell 1.27 0.45 0.85 0.48 Voltage (V) 0.0 0.1 0.2 0.3 0.4 0.5 Current Density (mA/cm2) -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 x 0.01 x 0.1 x 1 x 10 x 100 Figure 2-60. J-V curves for simulated cells with a 40 nm LD and varying carrier mobility. Table 2-18. Performance measures for simulated cells with 40 nm LD and varying mobility values. Mobility JSC (mA/cm2) VOC (V) FF (%) 0.01 x 1.47 0.45 0.78 0.52 0.1 x 1.47 0.45 0.84 0.56 1 x 1.47 0.45 0.84 0.56 10 x 1.47 0.45 0.84 0.56 100 x 1.47 0.45 0.84 0.56 Real Cell 2.20 0.44 0.56 0.53 118
PAGE 119
A) 0.0 2.0e+21 4.0e+21 6.0e+21 8.0e+21 1.0e+22 1.2e+22 1.4e+22 1.6e+22 1.8e+22 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40P h o to g e n e r a t e d C a r r i e r s ( p a i r s /c m3)X C o o r d i n a t e ( m )Y C o o r d i n a t e ( m ) 0.0 2.0e+21 4.0e+21 6.0e+21 8.0e+21 1.0e+22 1.2e+22 1.4e+22 1.6e+22 1.8e+22 B) 0.0 2.0e+21 4.0e+21 6.0e+21 8.0e+21 1.0e+22 1.2e+22 1.4e+22 1.6e+22 1.8e+22 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40P h o t o g e n e r a t e d C a r r i e r s ( p a i r s / c m3)X C o o r d i n a t e ( m )Y C o o r d i n a t e ( m ) 0.0 2.0e+21 4.0e+21 6.0e+21 8.0e+21 1.0e+22 1.2e+22 1.4e+22 1.6e+22 1.8e+22 Figure 2-61. Photogenerated carr ier distribution for a simulated unit cell with (A) LD = 40 nm and (B) LD = 40 and the P3HT absorption coefficient increased by 50%. 119
PAGE 120
X Coordinate ( m) 0.000.010.020.030.040.050.06 Y Coordinate ( m) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 -4e+20 -3e+20 -2e+20 -1e+20 0 1e+20 2e+20 3e+20 4e+20 Figure 2-62. Difference in photogenerated carriers (in pairs/cm3) between 150% and 100% P3HT absorption coefficients in simulated cells with 40 nm LD. 120
PAGE 121
121 Voltage (V) 0 00 10 20 30 40 5 Current Density (mA/cm2) -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 Standard Absorption Coefficient + 50% Absorption Coefficient Figure 2-63. J-V curves for simulated cells with varying absorption coefficient in P3HT. Table 2-19. Generated carrier s and performance measures for simulated solar cells. Cell Total Carriers (pairs/cm3) JSC (mA/cm2) VOC (V) FF (%) 40 nm 3.14 x 1025 1.47 0.45 0.78 0.52 Full Cell 3.23 x 10251.27 0.45 0.85 0.48 40 nm, 150% 3.42 x 1025 1.24 0.45 0.79 0.44
PAGE 122
CHAPTER 3 ORGANIC AND HYBRID SOLAR CELL PROCESS DEVELOPMENT Introduction The experimental details for process devel opment of organic and hybrid solar cells are presented in this chapter. In the first section, pretreatment of the indium tin oxide anode is studied. This work was done in collaboration wi th the research group of Dr. Chinho Park at Yeungnam University in South Korea, particular ly Jiyoun Seol. This work was previously presented at the 2006 World Conference on Phot ovoltaic Energy Conversion sponsored by IEEE and was published in their Conference Record [80]. The second section focuses on the development of bilayer photovoltaic cells using ab sorbing polymers. The next section describes efforts to characterize solvents appropriate for use in hybrid bulk hete rojunction films. The fourth section details work ch aracterizing hybrid films and the fabrication of photovoltaic cells from these films. The final section introduces Particle Induced Nanos tructuring, a process for hybrid film deposition intended to control the distribution of nanoc rystals in the polymer matrix. ITO Anode Treatment Organic solar cells incorporate transparent conducting substrates as an anode, with indiumtin-oxide (ITO) coated glass most often used. ITO films offer several positive characteristics as substrates for optical devices including a high luminous tr ansparency, good electrical conductivity, and good infrared reflectivity. For these reasons, ITO is widely adopted as transparent anodes in light-emitting diodes, liquid crystal displays, and solar cells [4, 81-82]. ITO coated glass substrates are commercia lly produced by sputter deposition followed by processes to improve surface roughness and micr ostructure. As-received substrate surfaces, however, have to be further proc essed prior to application to cu rrent flowing devices such as OLEDs and solar cells because a sputter-de posited surface microstructure and chemical 122
PAGE 123
composition can be degraded during extended de vice operation. Surface treatment of an ITO surface by several techniques can alter the chemical and physical properties of the surface such as work function, surface roughness, and oxidation property, and thus it could affect the device performance. In this study, nitrogen and oxygen plasma treatment along with electron bombardment of commercially available ITO-coated glass substrat es were revisited as approaches to improving the efficiency and stability of organic solar cells. The effect of these surface treatments on the surface morphology and chemical composition was ch aracterized, and organic solar cells with the structure ITO/PEDOT:PSS(50 nm)/CuPc(25 nm)/C60(15 nm)/Al(100 nm) were fabricated and the device performance measured. The substrate used in this study was commercia lly available ITO-coated glass with an ITO film thickness of ~ 1800 and sheet resistance of ~7 /sq. The as-received substrate was chemically cleaned by sequential ultrasonificatio n in trichloroethylene (TCE), acetone, and methanol, followed by nitrogen blow-drying. The cl eaned substrate was then exposed to either a nitrogen or oxygen plasma or an electron beam. Plasma treatment was ca rried out in a barreltype plasma chamber for 10 min at a power input in the range 50 to 300 W and pressure in the range 50 mTorr to 1 Torr with N2 gas flow. Electron beam irradiation was performed for 15 sec in a nitrogen environment with beam power varied from 0.5 to 2 kGy (kJ/kg). After treatment a PEDOT:PSS layer was spin-coated and dried in a vacuum oven. Organic films (CuPc and C60) and aluminum were deposited at room temperature in a thermal evaporator with a base pressure of 2.0x10-6 Torr. The chemical composition and surface morphology of the ITO surface were measured using XPS, AFM, and a video cont act angle system (VCAS). Power conversion 123
PAGE 124
efficiencies were measured under illumination fr om a solar simulator set to produce an AM 1.5 100 mW/cm2 spectrum. The measured contact angle of a water drople t on an ITO substrate changed significantly after surface treatment. The c ontact angle for untreated ITO s ubstrates was ~ 65, while films subjected to electron beam treatment showed a reduced contact angle of ~ 50. After exposure to the N2 plasma, the contact angle decreased dramatica lly to a value < 10. In both cases, the reduction is due to two effects. First, the e xposure to reactive radicals removes contamination from the film surface that may have remained after wet chemical cleaning. Secondly, these treatments increase the activity of the film surf ace by incorporating nitrogen radicals into the film. This effect is significan tly stronger in the case of N2 plasma than in the case of electron beam treatment because the plasma treatment su pplied a greater flux of highly reactive nitrogen radicals to the film surface. The change in surface polarity in th e case of electron beam treatment was not as significant as that in the ca se of plasma treatment, and as electron beam energy is significantly increased (larger than 2 kG y), the ITO film started to change its color to light gray, which degraded the luminous transparency of the substrate. AFM measurements were performed to quant ify the surface roughness of the ITO films. As-received ITO films showed a RMS surface roughness of 1.1 nm. This value was reduced to 0.8 nm by electron beam treatment and < 0.6 nm by N2 plasma treatment at optimized conditions. Sputtered ITO film generally contai ns irregular surface features, even though the subsequent polishing and annealing of the film improves its roughness. The surface treatment procedures used in this study are ex pected to attack the higher surf ace features first, resulting in a decrease in surface roughness [83, 84]. The effe ct, however, was not very significant, because the surface roughness of the as-receive d substrate was already low enough. 124
PAGE 125
XPS spectra from treated films showed a chan ge in the chemical composition of the film surface. Films treated with N2 plasma and electron beam bombardment showed an increase in nitrogen content accompanied by a decrease in oxygen composition, confirming that near-surface oxygen is replaced by nitrogen in films subjected to these treatments as shown in Table 3-1. In the case of N2 plasma exposure, there was a decrease in the In/Sn ratio of the films as well, which can cause a slight decrease in the work f unction of the films [85]. It is believed this change is a result of indium reacting with nitroge n radicals to form InN, which is subsequently removed from the film. The chemical shift of In-Sn-O bonding was also investigated by XPS, but no noticeable change in chemical shift was observed under th e treatment conditions investigated in this study. The incorporation of nitrogen into the n ear surface region of films treated in N2 plasma was confirmed by glow discharge spectroscopy (GDS). Films exposed to the plasma at a constant power but increasing pressure showed an increase in nitrogen levels in the film up to 700 mTorr. At higher pressures the nitrogen level decreased slightly, due to a decrease in nitrogen radical activity in the plasma. Organic solar cell devices were fabricated fr om the surface-treated IT O films to determine the impact on device performance as shown in Figure 3-1 and Table 3-2. Electron beam treatment produced no change in device effici ency; however, there was a decrease in VOC and an increase in JSC in these devices. O2 plasma treatment gave a slight increase in JSC, but a significant drop in VOC led to a reduction in overall device efficiency. N2 plasma treatment resulted in cells with efficiency nearly double th at of the untreated ITO cells. All treated ITO devices showed an improvement in the shape of th e I-V curve compared to that of untreated ITO devices. The changes in performance can be attri buted to several effects due to the treatment, 125
PAGE 126
including surface roughness, chemical stability, a nd reduced work function. Electron beam and N2 plasma treatment replace oxygen with nitrogen, re ducing the electron affin ity of the ITO film and resulting in a lowering of the work function that improves the charge collection efficiency. The effect of several ITO surface treatments on the performance of organic solar cells was determined. It was shown that exposure to a N2 plasma was more beneficial than either the oxygen plasma or e-beam treatment. The N2 plasma was most successful in improving the surface characteristics, as evidenced by the extent of lowering the contact angle and decreasing the surface roughness (AFM). This treatment incor porates nitrogen into the near surface region and produces a slight change in the In/Sn rati o, which reduces the ITO work function. These changes optimize the energy band diagram and im prove charge collection at the ITO anode. Bi-layer Organic Solar Cell Fabrication A process for fabricating bi-layer orga nic solar cells with the cell structure ITO/PEDOT:PSS/P3HT/C60/Al was developed. Figure 3-2 s hows the energy band diagram for a cell with this structure [45, 86]. This materi al system has received much attention for applications in bulk heterojunction solar cells due after the discovery of ultrafast charge transfer at interfaces between conjugated polymers and C60 molecules. In these bi-layer organic cells, P3HT serves as the primary absorber layer in the cells, with a bandgap of approximately 1.7 eV and a very str ong absorption coefficient. Excitons generated in the polymer are separated at the C60 interface, with electrons dropping to the lower energy level of the C60 and holes returning to the P3HT. Electrons are transported to the backside aluminum contact, while holes are transported through the P3HT layer and the hole transport layer (HTL) of PEDOT:PSS to the tran sparent ITO frontside contact. 126
PAGE 127
Cell Fabrication Procedure Bi-layer organic solar cells were fabricated on ITO-coated glass substrates. The substrates were patterned, cleaned, a nd treated in-house under N2 plasma. A hole transport layer of PEDOT:PSS and the ab sorber layer of P3HT were spin-coated onto th e substrate and dried under vacuum. The electron transport layer of C60 and the aluminum back contact were deposited by evaporation. For cells to be transported, gla ss encapsulation was perf ormed to extend the cell lifetime. The encapsulation process is described in greater detail below. A graphic synopsis of the fabrication process is shown in Figure 3-3. Substrate Preparation The cells were fabricated on ITO-coated gla ss substrates obtained from Samsung-Corning. The ITO films had a resistance of < 7 /sq. and a thickness of approximately 0.18 m. Asreceived ITO substrates were cut into 2.5 x 2.5 cm squares and patterned by HCl vapor etching. A 2 mm wide strip of elec trical tape was fixed to the ITO substrate to cover a strip that would become the ITO anode. The substrates were suspended from the inside lid of a glass dish with a small amount of HCl in the bottom of the dish to generate vapor. Vapor etching occurred over 25 min, after which the substrates were thoroughly rinsed with de-ionized water and blown dry with nitrogen. They were then chemically cleaned by successive 10 min sonication steps in trichloroethylene, acetone, and methanol, fo llowed by blow-drying under nitrogen. Cleaned substrates were subjected to N2 plasma for 10-min in a barrel-type plasma chamber wrapped in induction coils. The chamber offered a continuous flow of nitrogen at a constant flow rate during the treatment, and the conditions used for treatment were optimized to a supplied power of 50 W and a chamber pressure of 200 mTorr. 127
PAGE 128
Spin-Coating A solution of PEDOT:PSS in water obtaine d from Bayer was filtered with a 0.45 m syringe filter and spin-coated onto th e treated substrate for 30 sec. The film was then dried in a vacuum oven at 90 C for 30 min. A calibration cu rve was generated for dried film thickness vs. spin-coating speed and is shown in Figure 3-4. Th e curve was fit to power law as has been noted in the literature (87), with the curve fitting Equation 3-1. t = (e14.7)* -0.99 (3-1) The absorber layer of P3HT was then spin-coated from a solution of 5 mg/ml in chlorobenzene. The films were spin-coated for 30 sec and then dried under vacuum. The film thickness versus spin-coating speed was fit to Eq uation 3-2, and the results are shown in Figure 3-5. t = (e12.5)* .93 (3-2) Both spin-coating steps occurred in a clean room environment. Film thicknesses for these measurements were performed with a profilomete r. After the active layer was dried, selected areas of the substrates were wiped clean of th e polymer films to provide clean surfaces for external contacts. Evaporation The cells were placed in a glove box under nitrog en and loaded into a thermal evaporator. C60 was evaporated through a shadow mask at a ra te of approximately 2 /s under a pressure of 10-6 Torr. Finally, a thick layer of aluminum wa s rapidly evaporated through a second shadow mask to form the back contact of the cells. Encapsulation Encapsulation was performed to extend the lifetime of the fabricated cells by shielding them from moisture in the atmosphere. The two components of the encapsulation procedure are 128
PAGE 129
the encapsulant and the dessicant. The encapsulant consisted of a small gla ss slab to serve as the backing and a rubber spacer to prevent contact be tween the substrate and encapsulant glass. The assembly of substrate / spacer / en capsulant was sealed with epoxy. Prior to fixing the encapsulant onto the substrat e, a dessicant was adde d to protect the cells from moisture exposure. The dessicant used wa s barium oxide. A small pouch was created from part of a piece of weighing paper. Inside th e nitrogen glove box, the pouch was filled with BaO powder and sealed with double-sided tape. It was then affixed to the inside of the encapsulant glass. Several tiny holes were punctured in the pouch to allow moisture to reach the dessicant, and the final assembly was sealed to the substrate with epoxy. Film Drying Due to the potential impact of residual solvent in these films, care must be taken to ensure drying is complete after each film deposition step. Residual organic solvent from the active layer film serves as an insulator to cripple electrica l performance, while excess water remaining from the HTL film can oxidize and degrade the active layer polymer. To confirm the effectiveness of the drying step, FTIR spectra were compared betw een the dry films, the solutions used for spincoating, and spectra obtained from Sigma-Aldrich for the pure solvent. Figure 3-6 shows the spectra for a P3HT film spin-coated from chlorobenzene, as well as spectra for the solution and pure solvent. The large peak at approximately 3050 cm-1 in the chlorobenzene spectra is clearly visible in the solution spectra, but is noticeably missing in the P3HT film spectra. Also, several sharp, narrow peaks between 1600 and 500 cm-1 match in the solution and solvent spectra, but are missing from the film spectra. In Figure 3-7, the broad O-H stretching peak from approximately 3700 to 3000 cm-1 is an obvious feature in the FTIR spectrum for pure water. This wide peak is also obvious in the soluti on spectrum but is missing from the PEDOT:PSS film spectrum. Similarly, the str ong peak at approximately 1640 cm-1 in the water spectrum is present 129
PAGE 130
in the solution spectrum but missing from the film spectra. These FTIR spectra provide confirmation that film drying is complete under the conditions specified, and in that respect, these films are suitable for use in organic solar cells. Both solvents analyzed, chlorobenzene and water, produce strong peaks that are easily distin guished from the polymer spectrum, making this an effective technique for identifying residual solvent in the films after deposition. PEDOT:PSS The PEDOT:PSS layer in bi-layer solar cells wa s deposited with a thic kness in the range of 80 to 100 nm. One potential issue in this charge transport layer is that it is susceptible to pinholes, which can lead to shor ting of the devices. Pinholes can be generated during the drying process as the solvent evaporates and rises through the drying pol ymer. To see the if pinholes were problematic in this deposition process, cells were fabricated using a single-layer and double-layers of PEDOT:PSS. The double-layer devi ces should eliminate the presence of layerspanning pinholes by providing two separate films so that any pinholes would only span half of the final film. Experiments were performed by fabricating bi -layer solar cells using singleor doublelayers of PEDOT:PSS. A consistent final film thickness of 80 nm was used for the PEDOT:PSS layer: one 80 nm layer for the single-layer devi ces, and two 40 nm layers for the double-layer devices. All other cleaning, preparation, deposition, and encapsulation steps were held constant. After fabrication, the ce ll performance was characterized under 100 mW/cm2 light from a solar simulator. The resulting J-V curves are show n for single-layer devices in Figure 3-8 and for double-layer devices in Figure 3-9. From the results, it is clear that the performance of cells using the double-layer PEDOT:PSS film suffer dramatically. Of the fo ur cells using the double-la yer structure, two fail to show any diode characteristics, while the ot her two show only a minimal photovoltaic effect. 130
PAGE 131
While the performance of the cells using th e single-layer PEDOT:PSS were not phenomenal, they all showed a significant photocurre nt and diode characteristics. Cells fabricated with a single 80 nm layer of PEDOT:PSS showed efficiencies as high as 0.168% (cell 8-2), with all cells showing efficiencies of at l east 0.047%. In contrast, cells fabricated with the double-layer structure failed to show a measurable efficiency, although estimates put them in the range of 0.015% for the be st cell (cell 9-1). Open circuit voltage values for the single-layer cells were in the range of 0.15 V, while valu es for the double-layer cells are less than 0.05 V. Short circ uit current density values ranged from 1.5 to 2.5 mA/cm2 for the single-layer cells and from 0.7 to 1.4 mA/cm2 for the double-layer cells. The data show that pinholes are not a c oncern for the PEDOT:PSS films under these deposition and drying conditions. In fact, there is a different effect causing the double-layer films to perform more poorly than the single-layer films. The to tal thickness of the layers was held constant to keep series re sistance constant. However, the series resistance is impacted by the film resistivity in ad dition to the path length. It seems that the double-layer films showed a higher resistance to current flow due to the lower cell performance. For positive biases, the single-layer films showed current densities 7 10 % higher than the double-la yer films. This can be attributed to two possible cau ses interfacial resistance and f ilm resistance. Because of the double-layer structure, there is an extra film interface which could cause an increase in the overall resistance. Additionall y, the inherent film re sistivity could be increased due to the deposition conditions. The films were deposit ed via spin-coating, w ith the 40 nm films deposited at a much higher rpm than the 80 nm fi lms. This causes faster solvent evaporation and can inhibit the ability of the polymers to self -align in a configurati on that could minimize 131
PAGE 132
resistance. This effect has been observed in the past, particularly in acti ve layer films where extremely slow solvent evaporation has resulted in strong increases in the films conductivity. Another effect observed in PEDOT:PSS deposition is that of different particle filtering speeds that produce different solution properties. In polymer electronics processing, PEDOT:PSS is typically filtered prior to deposition to remove particles in the solution and provide a smoother film upon spin-coating. In our labs, the PEDOT:PSS solution was filtered by hand using a 0.45 m disc filter attached to a 100 ml syri nge. It was found that there were two different methods of filtering depending on the person doing the work, but because the work was done by hand rather than with a mechanical system. In one method, the solution was filtered very slowly, with the worker applying just e nough pressure to force the solution through the membrane in a dropwise fashion. The filter was pe riodically replaced as the filtration became more difficult due to clogging. This slow filtrat ion method resulted in a relatively low viscosity solution. In the other method, the solution was fo rced through the filter in one steady motion. This process was much faster, with several ml of the solution passed through in a matter of seconds, and resulted in a more viscous solution. The difference was noticed while developing calibration curves for film thickness depending on sp in-coating speed. The slow filtration results in a much thinner film than the fast filtration st ep because a lower percentage of the polymer is forced across the membrane at the lower pressure difference. Calibration curves using the two solutions are shown in Figure 3-10. The data set labeled 116 represents the slow filtration method, while the other three data sets show the faster filtration method. For cell fabrication, the fast filtration method was used to provide the ability to deposit approximately 100 nm thick films while still using a reasonably fast spin-coating speed to pr oduce smooth, uniform films. 132
PAGE 133
P3HT Prior to cell fabrication, P3HT films were analyzed under AFM to determine the surface quality of the films. Measurements were made under different solution concentrations and spincoating speeds. The results, s hown in Figure 3-11, show high-qua lity films with low surface roughness, indicating their suitability for bi-layer cell construction. All films spin-cast from the 5 mg/ml solution showed an RMS surface roughness of around 1 nm. For the films cast from the 10 mg/ml solution, the RMS surface roughness was between 2.5 nm and 3.6 nm, with the higher roughness occurring at the slowest spin speed. The data from the images shown in Figure 3-11 is tabulated in Table 3-3. Bi-layer Cell Fabrication Following these film and process characteriza tion investigations, bi-layer organic solar cells were fabricated with th e device structure ITO/PEDOT:PSS/P3HT/C60/Al. The bi-layer device structure is commonly used in OLED devi ces [88] and has been explored in molecular organic photovoltaics [89], but le ss work has been performed rega rding bi-layer devices using polymer active layers [90]. The devices were fabr icated on ITO-coated glass substrates prepared as described previously. A PEDOT:PSS film was deposited by spin-coating for 30 sec at 2500 rpm and drying for 30 min. The active layer of P3HT was deposited by spin-coating from a 5 mg/ml solution in 1,2-dichlorobenzene for 30 s ec at 3500 rpm and dried under vacuum. The samples were loaded into an ev aporator where a 150 film of C60 was deposited, followed by a 800 Al electrode. Bi-layer cell J-V measurements were perfor med at Busan National University in Busan, South Korea using a Keithley I-V measurement system under illumination from a solar simulator. Current measurements were convert ed to current density by dividing by the active cell area (0.04 cm2). Illuminated measurements were performed under 100 mW/cm2 133
PAGE 134
illumination. Power conversion efficiency ( ) was calculated using Equation 3-3, with Vmax, Imax, and Jmax representing the voltage, current, and cu rrent density at the maximum power point. The fill factor (FF) is calculated using Equation 3-4 with VOC and JSC being the open circuit voltage and short circu it current density. %100 /100 %1002 maxmax maxmax cmmW JV P IVin (3-3) SCOCJV JV FFmaxmax (3-4) The rectification ratio (RR) of a diode is the ratio of forward to reverse current at some applied bias. Higher rectificati on ratios show stronger diode characteristics in th e I-V curve for devices. Rectification ratios displayed in this section were calculated from dark current measurements at 0.5 V of forward a nd reverse bias unless otherwise noted. Bi-layer solar cells were fabricated using the procedure detailed pr eviously, but with no treatment performed on the ITO electrode. J-V cu rves for these cells are shown in Figure 3-12. The cells fabricated in this set all show ed a measurable photocurrent, demonstrating functioning solar cell behavior. The best-performing cell in the se t, Set I 3, showed a power conversion efficiency of 0.04%. Despite having the lowest VOC of all cells in the set, this champion cell showed a short-circu it current density of 1.55 mA/cm2, which was significantly higher than any other cell in the se t. Performance was low, but measu rable, in all cells. With the exception of cell I 3 with a VOC of 0.11 V, all cells showed a VOC of almost exactly 0.15 V. The fill factor for the cells ranged from 0.15 for ce ll I 4 to 0.24 for cell I-1. Cells I-1 and I-4 showed a JSC of approximately 0.5 mA/cm2. The JSC for cell I-2 was 1.03 mA/cm2. Another set of bi-layer cells were fabricated with the same cell structure, but applying plasma treatment to the ITO substrate. The perf ormance of these cells was poor, with the lack of 134
PAGE 135
performance attributed to possible poor encapsulation resulting in cell decay before measurement, or non-optimized plasma treatment inhibiting the cell performance. These cells showed a minimal photovoltaic effect, with open-circuit voltages below 0.1 V and short circuit current densities no high er than 1.10 mA/cm2. Although fill factors were in the range of 0.200.25, efficiencies were no higher than 0.02% due to the low photovolta ge and current in the cells. J-V curves for the cells are shown in Figure 3-13 and the cell perfor mance is detailed in Table 34. Solvent Comparisons Various solvents were compared to dete rmine their applicab ility for hybrid bulk heterojunction films. The solvents were all commonly available chemicals, so no high-cost specialty materials were used th at would add substantial cost to the fabrication process. The goal was to identify a solvent or a family of solvents that woul d provide the highestquality hybrid films for cell fabricat ion. The solvents used for th e tests are shown in Table 24. Only those which showed an approp riate level of solubility were u sed in further testing. Polymer solubility was qualitatively asse ssed by mixing a small amount of the polymer with a few milliliters of the chosen solvent in to a small vial and subjecting th e mixture to ultrasonication for at least one hour. Solvents with solubility labeled as No showed solid flak es of the polymer in the clear solvent after the mixing. Those labeled as Poor result ed in a color change of the liquid to indicate some degree of dissolution, bu t still contained significant solid particles of polymer. The label of Yes indicates that the po lymer was fully dissolved to give a red colored solution that is characteristic of P3HT. The solvents normal bo iling temperature and polarity values are listed in Table 3-5 as these parameters strongly impact the f ilm-forming properties of a solution. The values of polarity in the chart were taken from a solvent miscibility and polarity chart from Phenomenex [91] where higher values correspond to more polar solvents, with water 135
PAGE 136
having a value of 9.0. In the table, DMF is di methyl formamide, MED is methyl ethyl ketone, THF is tetrahydrofuran, and TCE is trichloroethylene. Hybrid films deposited from the solv ents that showed solubility for P3HT were analyzed with profilometry, AFM, and imaged with optical microscopy and SEM. Film roughness measurements taken from profilometry are shown in Figure 3-14. The results are plotted against both boiling temperature and polar ity, although no trend is evident in either case. These measurements were taken over 5 line scans of 1.0 to 1.5 mm on the surface of two separate substrates. AFM surface scans were taken for films deposited using 6 of the select ed solvents, and rms surface roughness was measured for 5 x 5 m and 1 x 1 m surface areas. The results for each test are shown in Figures 3-15 and 3-16, with separate graphs shown to plot rms surface roughness vs. solvent boiling temperature and solvent polarity. On the large-scale profilometer measurement s, chlorobenzene was the best performing solvent, showing a mean surface roughness of 64.9 nm on a line scan. The worst performers were THF and o-xylene, with mean rms values of 136 and 198 nm, respectively. For the 5 x 5 m AFM measurements, chloroform showed the lowest roughness, with the rms roughness being measured at 16.2 nm. THF displayed the highest roughness, with a value of 36.8 nm. For the small-area 1 x 1 m AFM measurement, Toluene was the supe rior solvent with an average rms value of 2.93 nm. Again, THF showed the hi ghest value, at 15.0 nm. The results are summarized in Table 3-6. SEM and optical microscope images of film su rfaces were taken to vi sually assess the film quality in conjunction with the surface roughness data. Figure 3-17 shows optical images of 136
PAGE 137
selected films that were represen tative of the samples taken for each solvent. Figure 3-18 shows SEM images from films, also representative of the data taken fo r each film solvent. In the optical microscope images, dark s pots on the image represen t surface features on the films. Some of the larger dark spots on the imag e, such as in the top left corner and near the right hand side of the 100x THF image, appear fuzzy because they are beyond the depth of field of the image, and different areas of this spot co uld be visualized by re-foc using the lens. In the 10x magnification images, it is noted that all solvents produce at least a few of these large black features. Chlorobenzene, o -dichlorobenzene, and toluene show these spots in smaller sizes and regularities than the other materials. Benzene shows large wispy collections of the dark spots that were not seen from other solvents, which seem to suggest precipitation or clustering occurring on a much larger scale than the other solvents show. These feat ures are also visible in the 100x image for benzene, and again, it does not appear for other solvents. The 100x image for THF shows a high frequency of very large features. Dichlorobenzene, on the other hand, shows a very low frequency and size of features in both the 10x and 100x images. Chlorobenzene shows a few large features, but the image overall shows a high-quality film with a low frequency of features. The SEM images shown in Figure 3-18 displa y surface images of the hybrid films at 5,000x and 15,000x magnification. These magnification levels are relatively low for SEM images, but attempts to focus the electron beam to produce images in the 30,000x to 50,000x range resulted in beam damage of the sample. In fact, an example of this is visible in the 15,000x image for o -dichlorobenzene. The dark box near the center of the image, just to the left of the bright surface feature, represents an area where a tighte r focus was attempted and the sample was burned. In these imag es, surface features of the films are clearly visible as bright 137
PAGE 138
spots in the secondary electron imag e. Images for chlorobenzene and o -dichlorobenzene show high-quality films with relatively few surface features. The images for TCE also shows a minimal amount of surface features, particularly in the 15,000x image. Chloroform displays a moderate amount of features, but less so than THF, which simila rly to the optical images shows the poorest film quality. Attempts to identify the composition of th ese surface features failed to yield results. Backscattered electron images of the films showed images of the features similar to the secondary electron images, but with no additional bright spots to demonstrate possible clustering of the higher density nanocrystals. Compositiona l scans were performed using energy-dispersive x-ray analysis (EDX), and these scans detect ed cadmium and selenium at a constant concentration in both the features and in smooth areas of the film surface. From this study, it was determined that chlo roform, chlorobenzene, and o-dichlorobenzene are good candidates for hybrid bulk heterojunction film deposition. These solvents all produce films that were among the low est in surface roughness for all of the measurement techniques used. Additionally, the film quality could be visually confirmed from optical microscopy and SEM surface images. Hybrid Bulk Heterojunction Cell Fabrication Bulk heterojunction pho tovoltaic cells were fabricated using P3HT as the absorbing semiconductor and nano-CdSe as the electron transpor ter. The cell design was the modeled after the bi-layer organic cell design described previously. The bulk heterojunction ce ll structure was ITO/PEDOT:PSS/P3HT:CdSe/Al, which is very similar to the bi-layer structure with the exception of CdSe replacing C60 as the electron acceptor, and that acceptor is now blended into the active layer film ra ther than deposited on top. Performance measurements for these cells were performed in-house rather than remotely, so the encapsulation process was not performed. 138
PAGE 139
Measurements were performed with a Keithley I-V-L measurement system under illumination from a JEOL solar simulator. Nanocrystal Synthesis and Surfactant CdSe nanocrystals were synthesized using the solution-phase growth mechanism demonstrated by Peng et al. [92]. In this met hod cadmium oxide is adde d to a flask containing trin -octylphosphine oxide (TOPO) and hexyl-phosphonic acid. The solution is then heated before the addition of selenium powder dissolved in liquid trin -octylphosphine. The reaction progresses and is halted by removal from the hea ting source. The nanocrystals, coated in a surfactant layer of TOPO, are pr ecipitated from the solution by the addition of methanol and centrifugation. The TOPO-coate d nanocrystals can further be modified by dissolution in pyridine, precipitation using hexane, and centrifugation to isolate the crystals. This ligand exchange process replaces the TO PO molecules with pyridine mo lecules on the surface. After ~5 steps of this process, the TO PO coating is fully replaced a nd the pyridine-coated nanocrystals can be used for processing. CdSe nanocrystals u sed in this dissertation were synthesized by Md. Azizul Hasnain and Trong Nguyen Tam Nguye n at Yeungnam Universi ty unless otherwise noted. Solutions were generated using CdSe nanocryst als with both types of surfactant coatings, and films deposited from these solutions were tested to determine the appropriate deposition parameters. The solutions consisted of a 60:40 mixture of CdSe nanocrystals and P3HT polymer dissolved in chloroform with a variable volume fraction of pyrid ine added to enhance the CdSe solubility. The surface roughness of spin-coated films is shown in Figure 3-19, plotted against the pyridine content of the solution. Note th e scale difference in the two graphs. These measurements were performed by Md. Azizul Hasnain at Yeungnam University and are displayed to clearly represent the effect of nanocrystal surfactant on film quality for hybrid films. 139
PAGE 140
For TOPO-coated nanocrystals, the surface rough ness of the films decreased as the pyridine content increased, up to 50%. This is due to the low solubility of TOPO-coated nanocrystals in chloroform. Pyridine concentrations of 45-50% we re required to generate films with less than 10 nm rms roughness. After surface exchange with pyr idine is performed, the pyridine-terminated nanocrystals become significantly more soluble in chloroform. In this case, pyridine concentrations of less than 10% yield rms su rface roughness values of less than 10 nm. Hybrid Films To fabricate bulk heterojunc tion solar cells with organic polymers and inorganic nanocrystals, great care must be taken to ensu re proper mixing between the two phases. The exciton diffusion length of most semiconducting polym ers is in the range 5 to 20 nm. Because of this limitation, the organic and inorganic phases in the hybrid active layer mu st be well-mixed so that excitons generated in the organic phase can reach an inorganic phase that is within one diffusion length. In this investigation, the properties of hybrid films is studied through optical microscopy, electron microscopy, atomic force microscopy, and surface profilometry. Solutions are generated by dissolving blends of P3HT polymer and CdSe nanopowde r into solvent mixtures of chloroform and pyridine. Hybrid films are sp in-cast from these solutions onto ITO-coated glass substrates that were subjected to N2 plasma treatment as described previously, and the films were dried under vacuum. TOPO-coated CdSe Initially, hybrid solutions were created with P3HT and TOPO-coated CdSe nanocrystals with a radius of approximately 5 nm. The solven t used in these solutions was a 50-50 mixture of chloroform and pyridine, based on the results show n in Figure 3-19. Three hybrid solutions were prepared with composition shown in Table 3-7. 140
PAGE 141
Films were spin-cast from these solutions, a nd their visual quality was studied using optical microscopy. The resulting images are s hown in Figure 3-20. The images were taken at 100x magnification, and the films imaged were spin-coated at 3000 rpm from the solutions described in Table 3-7. In these images, darker regions correspond to features on the surface of the films. From the images, it is observed that films from the higher-concentrated solutions show significantly more numerous and larger features than the f ilm from the weaklyconcentrated solution. The w eakly-concentrated Solution 3 show s only small and well-dispersed dark regions. Solutions 1 and 2 show a darker overall image, as is to be expected as these higher-concentrated solutions produce thicker film s. However, they al so show large surface features, some of which appear out of focus due to the depth of field of the microscope. These observed differences are due to the low solubility of both the polymer and nanocrystals in these composite solutions when TOPO-coated nanocrystal s are used. The polymer is poorly soluble in pyridine, and the TOPO-coated CdSe nanocrystals are poorly soluble in chloroform, so the only way to control the film morphology in this solvent sy stem is to reduce the to tal solute lo ad in the solution. The 20.4 mg/ml and 19 mg/ml solutions produced a significantly rougher surface than that of the 5 mg/ml solution, seen in the images in Figure 3-20. In an attempt to decrease the roughness of these films, a second spin-coating step was performed using pure chloroform. The hypothesis was that the solvent would selectively attack the mo re pronounced surface features, similar to the effect of nitrogen plasma smoothi ng the surface of ITO. The resulting film did show a decrease in roughness, simila r to that of the 5 mg/ml solu tion. An optical microscope image of the film is shown in Figure 3-21. The surface roughness decreas ed, but the film was nearly completely etched during this process. Even under spin speeds up to 8000 rpm, designed 141
PAGE 142
to significantly reduce the contac t time between the chloroform and the film, the original film was nearly completely removed. P3HT solubility in chloroform and pyridine To further investigate the effect of the so lvent composition on film qualities, studies were performed to compare the quality of pure P3HT films cast from pure chloroform, a 1:1 chloroform:pyridine mixture, and pure pyridine. Films were de posited from 5 mg/ml solutions of P3HT in each of the three solvent types a nd then examined under optical and secondary electron microscopes and the roughness measured with a surface profilometer. Under an optical microscope at 100x magnificat ion, clear differences are observed in film quality depending on the type of solvent used, as shown in Figure 3-22. The chloroform solution produces a film that appears smooth with a number of small surface features. The film deposited from the mixed solvent shows a rougher base f ilm with larger and more numerous surface features. The pyridine solution results in a smooth base film, but with sev eral very large surface features. SEM analysis of the chloroform and mixed so lvent films show results similar to those obtained under optical microscopy. Figure 3-23 displays a co mparison of these films at 2,000 and 10,000 x magnification. For the films deposited from chloroform solvent, there is a significant reduction in the number of surface features of the films. In the 2,000x image, there is one large area feature that is visible at the botto m of the image, but this was not observed to be common in the film. On the ot her hand, the films deposited fr om the mixed solvent show a distribution of surface features wi th varying shape and sizes rangi ng from a few microns to tens of microns. This offers further evidence of ph ase separation occurring in the films deposited from the mixed solvent. Because these films are pure P3HT rather than hybrid films, the observed features must be regions of P3HT that formed a non-uniform surface. This could occur 142
PAGE 143
through non-uniform precipitation of the polymer dur ing deposition. In the chloroform solvent, due to the high polymer solubility the smooth film is a result of a uniform deposition and film drying process. In the mixed solvent, however, the solubility is poorer and the P3HT will precipitate more quickly as solvent evaporates during the spin-coating and drying processes, resulting in the polymer freezing in its current state rather than being allowed to relax to a preferred alignment that r esults in a smoother film. Surface profiles of the films m easured through profilometry revealed films with significant increases in surface roughness as the amount of pyridine in the solvent mixture increased, as shown in Figure 3-24. When the film thickness was m easured using this technique, it was found that the film thickness decrea sed as the pyridine content of the solvent increased, due to decreasing solubility of the P3HT polymer. The profiles shown in Figure 3-24 are the result of a line scan across the surface of a sample that had the film stripped away on the left-hand side. The tall, wide peak at approximately 1 mm is the edge where the film was wiped clean and the film surface was dist orted. The polymer film is shown on the right-hand side, which is where the mean and median film thickness was measured. On the profiles, the solid teal line represents the median film thickness, the red dash-dot line repr esents the mean film thickness, and the blue dashed lines represent one standa rd deviation from the mean. Note that not all lines appear in all graphs due to the scaling. These scans show an increase in the mean film thickness as the pyridine content of the solvent increases, resulting from an increase in surface rough ess. They show, however, a decrease in median film thickness, due to the base film thickness being smaller as the polymer is poorly dispersed by the pyridine solvent. These properties are summarized in Table 3-8, and represent averages for measurements across different regions of each sample. It is interesting to 143
PAGE 144
note that for the pyridine solvent, the error range for the mean film thickness and the standard deviation of film thickness is larger than the ac tual measurement. This is due to the extremely nonuniform film which shows extremely larg e features over a very thin base film. Cell Fabrication Replacing TOPO-coated nanocrystals with pyrid ine-coated nanocrystals is the key to reducing the amount of pyridine needed in the solvent mixture. Because of this, pyridine-coated nanocrystals were used for attempts at hybrid bulk heterojunction ce ll fabrication. The processing steps for these cells were similar to those used for th e organic bi-layer solar cells described earlier. ITO-coated glass substrates were etched with HCl, then cleaned under ultrasonication in TCE, acetone, and methanol The clean surface was exposed to a N2 plasma under specific conditions, followed by deposition of PEDOT:PSS, then the hybrid solution, followed by evaporation of the Al electrode. Cell performance was tested in the dark and under simulated solar illumination. Cells deposited from chloroform solutions Bulk heterojunction cells were fabricated from pure chloroform solu tion and a 2% pyridine in chloroform solution. Nitrogen plasma treatment was pe rformed at 50 W and 200 mTorr for 10 minutes after etching and cleaning of the substrates. A thin film of PEDOT:PSS was deposited via spin-coating and dried under vacuum. The hybrid solution concentr ation was 5 mg/ml, consisting of 60% CdSe by wei ght. The aluminum electrode was between 125 and 150 nm in thickness. The resulting dark and light J-V curv es for these cells are shown in Figure 3-25. The active cell performance is low for both cells, particularly the cell deposited from pure chloroform solvent. The first cell, deposited fr om the chloroform-pyridine mixed solvent, shows JSC = 8.99 x 10-3 mA/cm2 with a maximum efficiency = 7 x 10-4 %. The second cell, in pure chloroform solution, shows JSC = 2.62 x 10-5 mA/cm2 and maximum efficiency = 1.3 x 10-6 %. 144
PAGE 145
Although these films show some photovoltaic effect it is extremely low, presumably due to the very thin nature of the films. From visual insp ection, the films are faint pink in color and highly transparent. These thin films fail to absorb a significant number of photons to yield highperformance solar cells. To enhance film thickness, a new hybrid so lution was created with the total solute concentration doubled to 10 mg/ml while maintaining the nanocrysta l weight percentage at 60%. The solvent was 2% pyridine in chloroform. The resulting J-V curves are shown in Figure 3-26. This cell showed higher values of JSC than the cell deposited from the 5 mg/ml solution. The measured cell performance characteristics were JSC = 3.39 x 10-2 mA/cm2, VOC = 0.256 V, FF = 0.275, and efficiency = 2.4 x 10-3 %. This is a 2.5x increase in efficiency compared to the 5 mg/ml hybrid solution. Although performance is s till low, the film produces photocurrent that is half an order of magnitude higher. Thicker-film cells in chloroform solution By further increasing the concentration of the active layer solution, light absorption in the cells can be further enhanced. In these cells the weight ratio of the hybrid film was altered from 60% CdSe to 50% CdSe. Because the CdSe serv es primarily as an elect ron transporter while P3HT is the absorber, the cell pe rformance can be improved by decreasing the CdSe volume in the hybrid films as long as electron transport path ways exist to allow current flow to the electrode. A 25 mg/ml hybrid solu tion with equal weights of P3HT and pyridine-coated CdSe was dissolved in a solvent of ch loroform with 2% pyridine. To determine the effects of the nanocrystals in the solution, a so lution consisting of 12 mg/ml P3HT in the 2% pyridine mixed solvent was also generated. This solution has essen tially the same P3HT concentration in solution (12.5 mg/ml in the hybr id solution, 12 mg/ml in the pure polymer solution), so the presence of nanocrystals in the hybri d solution differentiates the two. 145
PAGE 146
Surface profiles of the films generated with a profilometer are shown in Figure 3-27. For these scans, the film was wiped cl ean at the left-hand side of the image to allow measurement of the film thickness. Note that the scale for both prof iles is the same. The P3HT film shows a thickness of 95.7 17.8 nm with an rms roughne ss of 15 nm. The hybrid solution produces a slightly thicker film at 130 16.8 nm, but with an rms roughness of 239 nm. The inclusion of nanocrystals increases the surface roughness of the film by more than an order of magnitude. However, the nanocrystals have only a minimal effect on the film thickness. Although the hybrid solution had a total solute concentra tion that was twi ce as high as the P3HT solution, the final film thickness only increased by 35%. This demonstrates that the polymer is a much stronger factor in film thickn ess than the nanocrystals. Dark and illuminated J-V curves for cells fa bricated from the 25 mg/ml hybrid solution and the 12 mg/ml P3HT solution are shown in Figure 3-28. The hybrid cell showed performance characteristics of JSC = 4.99 x 10-2 mA/cm2, VOC = 0.701 V, FF = 0.232, and efficiency = 0.0175%. These measurements show a short-circuit current density 150% higher than that of the 10 mg/ml hybrid cell shown in Figure 3-26. As expected, J-V curves for the pure P3HT cell showed poorer performance than that of the hybrid cell. Perf ormance characteristics for the P3HT cell showed JSC = 2.03 x 10-2 mA/cm2, VOC = 0.523 V, FF = 0.310, and efficiency = 1.25 x 10-3 %, which are similar to that of the 10 mg/m l hybrid cell. Despite the poor measurements for the P3HT cell, the dark J-V curve shows a strong rectification ra tio, signifying a strong diode. J-V measurements in the dark for each of th ese cells showed a str ong rectification ratio, something that was not observed previously for thinner cells. This shows that the thicker films limit reverse leakage current that was easily driven through the thin active films in the previous cells. 146
PAGE 147
Cell lifetime measurements were also performed fo r these two cells. The cells were periodically measured under dark and illuminated conditions under exposure to room air. The cell performance decreased quickly with exposur e time, as seen in Figure 3-29. The first measurements for each cell occu rred after approximately 40 minutes after removal from a nitrogen glove box. The short-circuit current density for the hybrid cell decays approximately 70% in 12 min between the firs t and second measurements and by more than 95% in the 30 min between the first and third sets of measurements. The short-ci rcuit current density for the P3HT cell decays more slowly by approximately 40% in 11 min between the first and second sets of measurements, 50% in 26 min betw een the first and third sets, and 80% in approximately 3.5 hr between the first and final measurements. The da ta was fit to an exponential decay function of the form shown in Equation 35 with J in units of mA/cm2 and t in units of min. This fit was performed using Sigmaplot, and the resulting eq uations are displayed ove r the graph in Figure 329. Rsq values for the curves were 0.939 fo r the hybrid cell and 0.950 for the P3HT cell, showing a good fit for the data range. )exp(0tbaJJ (3-5) The hybrid cell shows a half-life of 7.42 min, while the P3HT cell shows a half-life of 16.05 min. This short half-life was confirmed for other hybrid cells fabricated in the same data set. It is interesting to note the difference in the initial and final values extrapolated from the exponential decay curves. The hybrid cell curve is extrapolated to an initial value of JSC = 1.53 mA/cm2, while the P3HT cell curve shows an in itial value of 8.09 x 10-2 mA/cm2. The P3HT cell, however, shows a JSC approximately twice as high as that of the hybrid cell as time approaches infinity: 5.80 x 10-3 mA/cm2 vs. 2.89 x 10-3 mA/cm2. These extrapolated values cannot be taken as absolute truths due to the uncertain nature of curve-fitting, but the 147
PAGE 148
experimental data in Figure 3-29 confirm that the short-circuit current density for the hybrid cell is initially higher than that of the P3HT cell and drops to a lower value after approximately 60 min of exposure time. The cause of the accelerate d decay for the hybrid cells as compared to the pure P3HT cells is unknown. The exposure times shown in Figure 3-29 were measured from the time the cells were removed from a glove box after the back contact de position, but this may not be an appropriate measurement of the exposure time, as exposure occurred throughout the fabrication process. After weighing out appropriate amounts of the polymer and nanocrystals, both of which were stored in a glove box under nitrogen, these material s were removed from the glove box in a vial where the solvent mixture was added under exposure to room air. Solvents were not stored in the glove box to prevent contamin ation of that environment. Th e solutions were sealed in their vials to limit further exposure dur ing mixing, but the vial contained room air rather than nitrogen at this point. Spin-coating occurred in a cleanro om environment to limit particle contamination, so the air was filtered and humidity-controlled to some degree, but not inert. After the films were dried under vacuum, they were again opened to the cleanroom atmosphere where the films were wiped clean in the electrode contact areas. At th is point, the cells en tered the glove box for electrode deposition. From the time that the solid P3HT was removed from the nitrogen environment to the time that the J-V curves of fi nished cells were measured, the total amount of air exposure could vary from 75 min to over 2 hr during the fabrication process. A large part of this exposure occurs while the P3HT is in solution, as the mixing tim e was typically at least 1 hr. A second set of cells fabricated from these so lutions yielded improved performance for the hybrid cell. For these cells, car e was taken to minimize the am ount of environmental exposure of the cells and films prior to testing. Approximately 15 min elapsed from when the finished 148
PAGE 149
cells were removed from the glove box and when J-V testing was performed. The best hybrid cell in this set showed performance measures of JSC = 0.205 mA/cm2, VOC = 0.705 V, FF = 0.288, and efficiency = 0.0416 %. The top performing P3HT cell resulted in JSC = 5.26 x 10-2 mA/cm2, VOC = 0.237 V, FF = 0.242, and efficiency = 3 x 10-3 %. The J-V curves for these cells are shown in Figure 3-30. The hybrid cell shown in Figure 3-30 displays a 300% improvement in short-circuit current density and a 24% improvement in fill factor when compared to the hybrid cell from Figure 91. The open-circuit voltage remained nearly constant, resulting in a nearly 140% improvement in power conversion efficiency. The P3HT cell shown in Figure 3-30 shows a diode with a lower rectification ratio as compared to the one shown in Figure 3-28. This cell showed a nearly 160% increase in short-circuit current density, but a 55% reduction in th e open-circuit voltage resulted in an approximately 20% reduction in fill factor. Despite these changes, the maximum efficiency improved by 140%. These cells saw approximately 15 min of air ex posure between removal from the glove box and J-V measurement. Using the short-circuit current density vs exposure time curves shown in Figure 3-29, the predicted exposure times for th ese cells would have been 21 min for the hybrid and 10 min for the polymer. This is within abou t 5 min of the actual exposure time, which is a reasonable prediction considering this curve doe s not take into account exposure time accrued during the fabrication process. Chlorobenzene solvent Based on comparisons of hybrid films deposited from various solvents, chlorobenzene was found to be a solvent that granted good mor phology with low surface roughness. To test the performance of hybrid bulk heterojunction cell s deposited from chlorobenzne, a 30 mg/ml solution was generated with a 1:1 mixture of P3HT and pyridine-coated CdSe dissolved in a 149
PAGE 150
mixture of 98% chlorobenzene with 2% pyridine. The best cell generated from this solution showed maximum performance of JSC = 0.138 mA/cm2, VOC = 0.391 V, FF = 0.292, and efficiency = 0.0158%. Dark and illuminated J-V curves for this cell are shown in Figure 3-31. This cell showed a lower short-circuit current cu rrent and open-circuit voltage than the similar cell deposited from a 25 mg/ml hybrid solutio n in a chloroform:pyridine solution. Commercial CdSe nanocrystals To test the quality of the synthesized CdSe nanocrystals used for cell fabrication, commercial CdSe nanopowder was acquired from Meliorum Technologies, Inc. [93] and used for the fabrication of solar cells. The particl es were 5 nm in diameter, just as the particles synthesized in-house. The hybrid solution using these particles was 25 mg/ml using a 1:1 by weight mixture of the commercial CdSe and P3HT in a solvent of chloroform with 2% pyridine by volume. The best cell fabricated from this solution showed the following performance characteristics: JSC = 9.44 x 10-2 mA/cm2, VOC = 0.329 V, FF = 0.247, and efficiency = 7.65 x 103 %. Dark and illuminated J-V curves for this cell are shown in Figure 3-32. These curves are inferior to those fabricated using in-house CdSe nanocrystals. The dark J-V curve shows a rectification ratio of only 0.98 at 1 V. Although the short-circu it current is lags only the bestperforming hybrid cell, the open-circ uit voltage was less than 0.4 V and led to the low efficiency. The terminating group on these nanocrystals is u nknown, as the company refused to give up this information. No attempts were made to alter the crystals; all cells we re fabricated with these particles as they were received. Hybrid cell performance summary Selected J-V curves for hybrid bulk heterojunc tion solar cells are s hown in Figure 3-33. These curves are all shown in previous figures, but are presented here on one graph for comparison. A compilation of the J-V data for hybrid cells shown in this section is shown in 150
PAGE 151
Table 3-9. The cell fabricated from a 25 mg/ml solution of equal weights P3HT and CdSe dissolved in 2% pyridine in chloroform with minimal air exposure showed the highest performance by simultaneously demonstrating the highest short-circuit cu rrent density and opencircuit voltage of all cells measured. The top-performing bi-layer and hybrid cell J-V cu rves are plotted together in Figure 3-34. The current flow through the bi-lay er cell is an order of magnitude higher than that of the hybrid cell, demonstrating the need for improved morpho logy control in the hybrid active layers. The bi-layer cell shows a low VOC compared to that of the hybrid cell. The series and shunt resistances of these cells were estimated from the J-V curves using Equation 3-6 and Equation 3-7, respectively [62]. This calculation is good for shunt resistance, but series resistance is more accura tely calculated as the applied vo ltage approaches infinity. For the bi-layer cell, J-V data was not sufficiently collected to make this calculation. From these equations, resistances for the bi-l ayer cell were calculated as Rs = 1.59 x 103 and Rsh = 2.25 x 103 For the hybrid cell, resist ances were calculated as Rs = 4.5 x 104 and Rsh = 8.33 x 104 The series resistance for th e hybrid cell calculated at the maximum measured voltage point was 1.5 x 10-2 For another bi-layer cell with a more extensive set of J-V data, the series resistance is calculated as 1.8 x 10-2 0 I sdI dV R (3-6) 0 V shdI dV R (3-7) These resistance calculations s how shunt resistances approximat ely 5 orders of magnitude higher than the series resistan ce, which should result in high-quality solar cells. Further improvements in cell design to maximize absorpti on and charge separation in the hybrid cells 151
PAGE 152
should result in further reduction of the series resistance and drastic improvements in current flow. Particle Induced Nanostructuring A key challenge in hybrid bulk heterojunction solar cells is control of the nanocrystal distribution throughout the act ive layer. The film must be well -mixed to allow efficient exciton dissociation, but also provide pe rcolation pathways to provide charge collection pathways. A new concept developed to resolve this issue is particle induced nanostructuring (PIN). The PIN concept involves depositing multiple ultra-thin layers to control nanocrystal distribution throughout the thickn ess of the film. In this study, ITO-coated glass substrates we re chemically cleaned and treated with nitrogen plasma before film deposition. Mu ltiple layers were deposited from a weakly concentrated 5 mg/ml solution of 60 wt. % CdSe and 40 wt. % P3HT in chloroform with 2% pyridine by volume. Deposition occurred at a sp in-coating speed of 5000 rpm to produce very thin films. The thicknesses of the films were measured with a profilo meter after removing a portion of the film to create a step. Film thickness measurements on samples with 1 to 6 film layers showed an interesting phenomenon. As expected, the film thickness grew with the addition of multiple layers. The film thickness increased, however, only after two deposition steps were performed, as shown in Figure 3-35. On the first deposition, a film of approximately 10 nm was deposited on the substrate. After the second, the film thickness remained constant. On the third, another film of approximately 10 nm was deposited, followed by another spin-coating resulting in no film growth. After the next deposition step, th e film grows by another 10 nm, with the 6th resulting in a minimal amount of growth. 152
PAGE 153
The dotted line drawn through the data in Figure 3-35 is the best fit line for film thickness vs. number of layers, which follows Equation 3-8. With this trend, to reach a hybrid film thickness of 100 nm, 21 layers would be required. In the fabricati on setup used for these experiments where film deposition equipment is open to atmosphere, this would result in a prohibitive amount of atmosphe ric exposure time for the P3HT. For deposition in a nitrogen or argon environment like a glove box, how ever, this process is feasible. )(#*903.4 Layers ess FilmThickn (3-8) The film surface roughness was measured with AFM, and the surface was visualized with SEM and optical microscopes. Figure 3-36 displays images of the results of these measurements at 1, 3, 5, and 7 layers. The films analyzed by SEM and AFM were different from the films analyzed with profilometry due to a difference in the sample size required for these techniques. All SEM images are shown at the same scale (at 30,000x magnification), and all AFM images show a 3 x 3 m scan area with a 100 nm scale on the z-axis. Optical microscope images of the films are s hown in Figure 3-37. These images are taken at 100x magnification and are representative of the multiple images taken of these surfaces. The rms surface roughness was measured from the AFM images for 3 x 3 m and 1 x 1 m areas on the surface. The results are shown in Figure 3-38. Additionally, lines are fit to determine the trends for roughness as more layers are deposited. The trendlines follow Equation 3-9 for the large area measurement and Equation 3-10 for the small area measurement. If these trendlines are ex trapolated to the required 21 layers for a 100 nm active layer film, the rms surface roughness is predicted to be 39.9 nm for a 5 x 5 m area and 15.0 nm for a 1 x 1 m area. From previous m easurements of surface roughness by profilometry and AFM, it was found that profilometer roughness measurements were 153
PAGE 154
approximately 3.4x higher than 5 x 5 m AFM measurements and approximately 14x higher than 1 x 1 m AFM measurements. By this extrapolation, the rms surface roughness of a 21 layer PIN film will be in the range 136 210 nm. Th e hybrid film surface profile shown in Figure 327 displayed an rms roughness of 239 nm when measured by profilometry. This comparison relies on a large amount of extrapolation, but imp lies that this technique is capable of producing films with surface roughness as low or lower than similarly thick films deposited from concentrated solutions in a single deposition step. (3-9) 65.14)(#20.1 Layers RRMS (3-10) 60.9)(#26.0 Layers RRMSCells were not fabricated using this tec hnique, as lifetime measurements on other cells indicated that the required air exposure time would cr ipple the cells before their performance could be evaluated. Once the cell fabrication process line is housed under an inert environment, this limitation no longer applies and cells consisti ng of multiple PIN layers can be compared to single-layer cells to quantitatively determ ine the applicability of this technique. 154
PAGE 155
Table 3-1. Chemical composition of ITO films Condition Atomic Percent Atomic Ratio O Sn In N O/In In/Sn No Treatment 56.75 2.65 40.6 1.397 15.32 N2 Plasmaa 55.49 2.67 40.49 1.35 1.37 15.16 e-beam b 56.67 2.38 38.76 2.19 1.462 16.28 a After N2 plasma treatment at 50 W, 1 Torr, 10 min b After e-beam treatment at 2kGy Voltage (V) -0.6-0.4-0.20.00.20.40.60.81.0 Current Density (mA/cm2) -4 -2 0 2 4 6 8 10 12 No treatment O2 Treatment Electron Beam N2 Treatment Figure 3-1. J-V curves for organic so lar cells on treated ITO substrates Table 3-2. Performance of organic solar cells on treated ITO substrates Treatment VOC (V) JSC (mA/cm2) (%) None 0.52 1.61 0.36 N2 Plasma 0.46 2.73 0.62 E-Beam 0.38 2.08 0.36 O2 Plasma 0.32 1.82 0.27 155
PAGE 156
Figure 3-2. Energy band diagram for bi-layer organic solar cells. All energy levels are listed in units of eV. Figure 3-3. Steps for bi-layer sola r cell fabrication. A) ITO-coat ed glass substrate. B) Patterned ITO anode. C) After PEDOT:PSS spin-coating. D) After P3HT spin-coating. E) After polishing. F) After C60 evaporation. G) After Al eva poration. H) Encapsulated solar cell. 156
PAGE 157
Spin Speed (rpm) 10002000300040005000 Film Thickness (A) 0 1000 2000 3000 4000 Figure 3-4. PEDOT:PSS film thickness vs. spin-coater speed. Spin Speed (rpm) 1000 2000 3000 4000 Film Thickness (A) 0 100 200 300 400 500 600 Figure 3-5. P3HT film thickness vs. spin-coating speed. 157
PAGE 158
Chlorobenzene P3HT P3HT in Chlorobenzene Figure 3-6. Upper Spectrum: FTIR spectrum for chlorobenzene from Sigma Aldrich. Lower Spectrum: P3HT film (purple) spin-coated from P3HT in chlorobenzene solution (red). 158
PAGE 159
Water PEDOT:PSS PEDOT:PSS in Wate r Figure 3-7. Upper spectrum: FT IR spectrum of water from Sigm a Aldrich. Lower spectrum: PEDOT:PSS film (gray) spin-coated from solution of PEDOT:PSS in water (blue). 159
PAGE 160
A) Voltage (V) -0.2-0.10.0 0.1 0.2 0.3 Current Density (mA/cm2) -10 0 10 20 30 8-2 Dark 8-1 Illuminated 8-2 Illuminated 8-3 Illuminated 8-4 Illuminated B) Voltage (V) -1.0 -0.5 0.0 0.5 1.0 Current Density (mA/cm2) 0.0001 0.001 0.01 0.1 1 10 100 1000 8-2 Dark 8-1 Illuminated 8-2 Illuminated 8-3 Illuminated 8-4 Illuminated Figure 3-8. Linear (A) and b ase 10 log-scale (B) J-V curves for bi-layer organic solar cells fabricated with a single 80 nm thick layer of PEDOT:PSS. 160
PAGE 161
A) Voltage (V) -0.2-0.10.0 0.1 0.2 0.3 Current Density (mA/cm2) -10 0 10 20 30 9-1 Dark 9-1 Illuminated 9-2 Illuminated 9-3 Illuminated 9-4 Illuminated B) Voltage (V) -1.0 -0.5 0.0 0.5 1.0 Current Density (mA/cm2) 0.001 0.01 0.1 1 10 100 1000 9-1 Dark 9-1 Illuminated 9-2 Illuminated 9-3 Illuminated 9-4 Illuminated Figure 3-9. Linear (a) and base 10 log-scale (b ) J-V curves for bi-layer organic solar cells fabricated with two 40 nm thick layers of PEDOT:PSS. 161
PAGE 162
Spin Speed (rpm) 10002000300040005000 Film Thickness () 0 500 1000 1500 2000 2500 3000 051116 051122 051123 051130 Figure 3-10. Calibration curves for slowa nd fast-filtered PEDOT :PSS. Slow-filtered PEDOT:PSS is shown in the data set. Fast-filtered PEDOT:PSS is shown in data sets , and A B D E C F F C E DB A Figure 3-11. AFM images of P3HT films. Images A C were spin-cast from 5 mg/ml P3HT in chlorobenzene solutions, while images D F were spin-cast from 10 mg/ml P3HT in chlorobenzene solutions. Images A and D were spin-cast at 2000 rpm, images B and E at 3000 rpm, and images C and F at 4000 rpm. For all images, the scale bar for the film height axis is 50 nm, and the scan area is 1 m x 1 m. 162
PAGE 163
Table 3-3. RMS surface roughness of P3HT films shown in Figure 3-11. Label Solution Spin Speed RMS Roughness A 5 mg/ml 2000 rpm 0.89 nm B 5 mg/ml 3000 rpm 1.27 nm C 5 mg/ml 4000 rpm 0.94 nm D 10 mg/ml 2000 rpm 3.66 nm E 10 mg/ml 3000 rpm 2.59 nm F 10 mg/ml 4000 rpm 2.54 nm Voltage (V) -0.2-0.10.0 0.1 0.2 0.3 Current density (mA/cm2) -4 -2 0 2 4 6 Set I dark Set I 1 Set I 2 Set I 3 Set I 4 Figure 3-12. J-V curves for bi-layer solar cells fabricated on untreated ITO substrates. 163
PAGE 164
A) Voltage (V) -0.2-0.10.0 0.1 0.2 0.3 Current Density (mA/cm2) -4 -2 0 2 4 6 8 10 II-1 dark II-2 dark II-3 dark B) Voltage (V) -0.4 -0.2 0.0 0.2 0.4 Current Density (mA/cm2) 0.001 0.01 0.1 1 10 100 II-1 dark II-2 dark II-3 dark Figure 3-13. J-V curves in the da rk (A and B) and under 100 mW/cm2 illumination (C and D). 164
PAGE 165
C) Voltage (V) -0.2-0.10.0 0.1 0.2 0.3 Current Density (mA/cm2) -4 -2 0 2 4 6 8 10 II-1 dark II-1 II-2 II-3 II-4 D) Voltage (V) -0.4 -0.2 0.0 0.2 0.4 Current Density (mA/cm2) 0.001 0.01 0.1 1 10 100 II-1 dark II-1 II-2 II-3 II-4 Figure 3-13 continued. J-V curves under 100 mW/cm2 illumination (C and D). 165
PAGE 166
Table 3-4. J-V data for bi-layer solar cells. Sample RR JSC (mA/cm2) VOC (V) FF (%) I-1 2.4 0.55 0.16 0.23 0.02 I-2 1.03 0.15 0.19 0.03 I-3 1.55 0.11 0.23 0.04 I-4 0.41 0.16 0.15 0.01 II-1 8.53 1.10 0.08 0.24 0.02 II-2 3.96 0.68 0.07 0.22 0.01 II-3 1.81 0.70 0.06 0.19 0.01 II-4 0.88 0.07 0.21 0.01 Table 3-5. Solvents considered for hybr id bulk heterojunction film deposition. Solvent B.T. (C) Polarity P3HT Solubility Acetone 56.2 5.1 No 2-butanol 79.6 4.0 No DMF 153 6.4 No Methanol 64.6 5.1 No 2-propanol 82.4 3.9 No MEK 80.0 4.7 Poor Pyridine 115.3 5.3 Poor Benzene 80.1 2.7 Yes Chlorobenzene 131.7 2.7 Yes Chloroform 61.2 4.1 Yes o-dichlorobenzene 180 2.7 Yes THF 66.0 4.0 Yes Toluene 110.6 2.4 Yes TCE 87.2 1.0 Yes o-xylene 144.4 2.5 Yes 166
PAGE 167
A) Solvent Boiling Temperature oC 406080100120140160180200220 RMS Roughness (nm) 0 100 200 300 400 THF Chlorobenzene o-dichlorobenzene Chloroform Toluene o-xylene TCE Benzene B) Solvent Polarity 0.51.01.52.02.53.03.54.04.5 RMS Roughness (nm) 0 100 200 300 400 THF Chlorobenzene o-dichlorobenzene Chloroform Toluene o-xylene TCE Benzene Figure 3-14. Surface roughness measurements by profilometry for hybrid films deposited from various solvents. 167
PAGE 168
A) Solvent Boiling Temperature (oC) 406080100120140160180200 RMS Roughness (nm) 0 10 20 30 40 50 60 THF Chlorobenzene o-dichlorobenzene Chloroform Toluene o-xylene B) Solvent Polarity 0.51.01.52.02.53.03.54.04.5 RMS Roughness (nm) 0 10 20 30 40 50 60 THF Chlorobenzene o-dichlorobenzene Chloroform Toluene o-xylene Figure 3-15. RMS surface roughness for 5 x 5 m surface area samples measured with AFM. 168
PAGE 169
A) Solvent Boiling Temperature (oC) 406080100120140160180200 RMS Roughness (nm) 0 2 4 6 8 10 12 14 16 18 20 THF Chlorobenzene o-dichlorobenzene Chloroform Toluene o-xylene B) Solvent Polarity 0.51.01.52.02.53.03.54.04.5 RMS Roughness (nm) 0 2 4 6 8 10 12 14 16 18 20 THF Chlorobenzene o-dichlorobenzene Chloroform Toluene o-xylene Figure 3-16. RMS surface roughness for 1 x 1 m surface area samples measured with AFM. 169
PAGE 170
Table 3-6. Mean rms surface roughness in nm fo r hybrid films deposited from selected solvents Solvent Profilometer AFM (5 x 5 m) AFM (1 x 1 m) THF 136 36.8 15.0 Chlorobenzene 64.9 25.0 8.42 o-DCB 81.2 21.1 9.58 Chloroform 71.8 16.2 3.22 Toluene 66.2 27.4 2.93 Xylene 198 16.8 7.76 TCE 72.0 Not measured Not measured Benzene 73.9 Not measured Not measured THF Chlorobenzene 10 x 100 x 100 x 10 x Figure 3-17. Optical microscope images of selec ted films. The first column displays images taken at 10x magnification, while the sec ond column displays images taken at 100x magnification. 170
PAGE 171
o-dichlorobenzene Chloroform Toluene 10 x 100 x 100 x 10 x 100 x 10 x Figure 3-17 continued. Optical microscope imag es of selected films. 171
PAGE 172
o-xylene Benzene TCE 100 x 10 x 100 x 10 x 100 x Figure 3-17 continued. Optical microscope imag es of selected films. 10 x 172
PAGE 173
THF Chlorobenzene o-dichlorobenzene Figure 3-18. SEM images of selected hybrid films. The first column image was taken at 5,000x magnification and the second column imag e was taken at 15,000 x magnification. 173
PAGE 174
Chloroform TCE Figure 3-18 continued. SEM imag es of selected hybrid films. 174
PAGE 175
A) B) Figure 3-19. Film surface roughn ess vs. pyridine concentration in the chloroform solvent for TOPO-coated CdSe nanocrystals (A) and pyr idine-coated CdSe nanocrystals (B). 175
PAGE 176
Table 3-7. Hybrid solutions of P3HT and TOPO-coated CdSe nanocry stals in a mixed solvent of chloroform and pyridine. Solution P3HT Wt. CdSe Wt. Chloroform Vol. Py ridine Vol. Solution Concentration 1 26.8 mg 35.5 mg 1.5 ml 1.5 ml 20.4 mg/ml 2 24.8 mg 23.2 mg 1.5 ml 1.5 ml 19 mg/ml 3 15.0 mg 25.0 mg 1.5 ml 1.5 ml 5 mg/ml A) B) C) Figure 3-20. Optical microscope images of hybrid films deposited from 20.4 mg/ml (A), 19 mg/ml (B) and 5 mg/ml (C) solutions. Images were taken at 100x magnification and the films were deposited at 3000 rpm and dried under vacuum at 120 C. 176
PAGE 177
Figure 3-21. Optical microscope image of 19 mg/ml hybrid film deposited at 3000 rpm and subjected to a pure solvent spin-coating step at 8000 rpm. The image was taken at 100x magnification. A) B) C) Figure 3-22. Optical microscope images of P3HT films deposited from 5 mg/ml solutions in chloroform (A), 1:1 chloroform:pyridine (B ), and pyridine (C). All images are taken at 100x magnification. 177
PAGE 178
2,000 x 10,000 x Chloroform Solvent Mixed Solvent Figure 3-23 continued. SEM images of P3HT films deposited from pure chloroform and an equal mixture of chloroform and pyridine solv ents. Images are shown at 2,000x and 10,000x for each film. 178
PAGE 179
A) Lateral Distance (mm) 0.0 0.5 1.0 1.5 2.0 2.5 Film Thickness (nm) 0 20 40 60 80 100 ITO P3HT Film Figure 3-24. Surface profiles of P3HT films deposited from chloroform (A), 1:1 chloroform:pyridine (B), and pyridine (C) solvents. In the graphs, the bare substrate appears on the left hand side with the film on the right hand side beyond the wide peak. The solid teal line represents the me dian film thickness, the red dash-dot line represents the mean film thic kness, and the blue dashed li nes represent one standard deviation from the mean. 179
PAGE 180
B) Lateral Distance mm) 0.0 0.5 1.0 1.5 2.0 2.5 Film Thickness (nm) 0 20 40 60 80 100 C) Lateral Distance (mm) 0.0 0.5 1.0 1.5 2.0 2.5 Film Thickness (nm) 0 20 40 60 80 100 ITO P3HT Film Figure 3-24 continued. Surface profiles of P3HT films deposited from chloroform (A), 1:1 chloroform:pyridine (B), and pyridine (C) solvents. 180
PAGE 181
Table 3-8. Film properties for P3HT films deposited from various solvents. Solvent Mean Thickness (nm) Median Thickn ess (nm) Standard Deviation (nm) Chloroform 21.2 8.18 23.1 7.95 9.58 5.41 Mixed 56.2 34.3 16.2 4.42 227 Pyridine 78.5 85.4 10.9 8.61 338 A) Voltage (V) -1.5-1.0-0.50.00.51.01.5 Current Density (mA/cm2) -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 Dark Illuminated -0.2-0.10.00.10.20.30.40.5 -0.04 -0.02 0.00 0.02 0.04 Figure 3-25. Dark and illuminated J-V curves for cells generated from 5 mg/ml composite solutions in (A) chloroform mixed with 2% pyridine and (B) pure chloroform. Note the scale difference of the graphs. 181
PAGE 182
B) Voltage (V) -1.5-1.0-0.50.00.51.01.5 Current Density (mA/cm2) -1.2e-3 -1.0e-3 -8.0e-4 -6.0e-4 -4.0e-4 -2.0e-4 0.0 2.0e-4 4.0e-4 Dark Illuminated -0.2-0.10.00.10.20.30.40.5 -1e-4 -5e-5 0 5e-5 1e-4 Figure 3-25 continued. Dark and illuminated J-V curves for cells generated from 5 mg/ml composite solution in pure chloroform. Voltage (V) -1.5-1.0-0.50.00.51.01.5 Current Density (mA/cm2) -0.4 -0.2 0.0 0.2 0.4 Figure 3-26. Dark and illuminated J-V curves for 10 mg/ml hybrid solutio n deposited with lowspeed spin-coating. 182
PAGE 183
A) Lateral Distance (mm) 0.0 0.5 1.0 1.5 2.0 2.5 Film Thickness (nm) -100 0 100 200 300 400 500 P3HT Film ITO P3HT FilmB) Lateral Distance (mm) 0.0 0.5 1.0 1.5 2.0 2.5 Film Thickness (nm) -100 0 100 200 300 400 500 Hybrid Film ITO Hybrid FilmFigure 3-27. Surface profiles of f ilm deposited from (A) 12 mg/ml P3HT and (B) 25 mg/ml hybrid solutions in 2% pyridine in chloroform. 183
PAGE 184
Voltage (V) -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Current Density (mA/cm2) -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 Hybrid Dark Hybrid Illuminated P3HT Dark P3HT Illuminated -0.8 -0.6 -0.4 -0.2 0.0 0.2 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 Figure 3-28. Dark and illuminated J-V curves for hybrid bulk heterojunction solar cell deposited from 25 mg/ml solution and P3HT polymer cell deposited from 12 mg/ml solution. The insert shows a zoom-in on the active area of the cells. 184
PAGE 185
Exposure Time (min) 0 50100150200250300 Short Circuit Current Density (mA/cm2) 0.00 0.01 0.02 0.03 0.04 0.05 0.06 Hybrid Cell P3HT Cell Decay Hybrid Decay P3HT Hybrid Cell J = 2.894E-3 + 1.525 exp(-9.369E-2 t) P3HT Cell J = 5.804E-3 + 7.506E-2 exp(-4.820E-2 t) Figure 3-29. Short-circuit current de cay for hybrid (gold circles) and P3HT (green squares). 185
PAGE 186
Voltage (V) -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Current Density (mA/cm2) -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 Hybrid Dark Hybrid Illuminated P3HT Dark P3HT Illuminated -0.8 -0.6 -0.4 -0.2 0.0 0.2 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 0.25 Figure 3-30. Dark and illuminated J-V cu rves for hybrid bulk heterojunction and pure P3HT solar cells with limited ai r exposure during processing. 186
PAGE 187
Voltage (V) -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Current Density (mA/cm2) -30 -20 -10 0 Dark Illuminated -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 -0.4 -0.2 0.0 0.2 0.4 Figure 3-31. Dark and illuminated J-V cu rves for hybrid solar cell fabricated from chlorobenzene with 2% pyridine solution. Red curves represent illuminated J-V and black curves represent dark J-V. 187
PAGE 188
Voltage (V) -1.5-1.0-0.50.00.51.01.5 Current Density (mA/cm2) -1.0 -0.5 0.0 0.5 1.0 1.5 -0.20.00.20.40.60.81.0 -0.4 -0.2 0.0 0.2 0.4 Figure 3-32. Dark and illuminated J-V curves fo r a hybrid solar cell fabr icated with commercial CdSe nanopowder. Green curves represen t illuminated J-V and black curves represent dark J-V. 188
PAGE 189
189 Voltage (V) -1.0 -0.5 0.0 0.5 1.0 Current Density (mA/cm2) -0.4 -0.2 0.0 0.2 0.4 25 mg/ml in Chloroform-Pyridine 25 mg/ml in Chloroform-Pyridine short exposure time 25 mg/ml in Chloroform-Pyridine commercial CdSe 30 mg/ml in Chlorobenzene-Pyridine Figure 3-33. Illuminated J-V curves for hybrid bulk heterojunction sola r cells with various fabrication conditions. Voltage (V) -0.20.00.20.40.60.81.0 Current Density (mA/cm2) -3 -2 -1 0 1 2 Bi-layer Cell Hybrid Cell Figure 3-34. J-V curves for the best bi-layer and hybrid cells shown in this dissertation.
PAGE 190
Table 3-9. Hybrid solar cell fabricat ion information and performance data. Figure Solution Concentration CdSe wt. % in P3HT % Pyridine in Chloroform JSC (1) VOC (V) FF (2) Notes 3-25 5 mg/ml 60% 2% 8.99 0.338 0.230 0.7 3-25 5 mg/ml 60% 0% 0.026 0.194 0.252 0.001 Pure Chloroform Solvent 3-26 10 mg/ml 60% 2% 33.9 0.256 0.275 2.4 3-28 25 mg/ml 50% 2% 49.9 0.701 0.232 17.5 3-28 12 mg/ml 0% 2% 20.3 0.523 0.310 12.5 Pure P3HT film 3-30 25 mg/ml 50% 2% 205 0.705 0.288 41.6 Minimized air exposure 3-30 12 mg/ml 0% 2% 52.6 0.237 0.242 3.0 Pure P3HT film Minimized air exposure 3-31 30 mg/ml 50% 2% ( 3 ) 138 0.391 0.292 15.8 Pyridine in Chlorobenzene Solvent 3-32 25 mg/ml 50% 2% 94.4 0.329 0.247 7.65 Commercial CdSe (1) JSC displayed in units of A/cm2 (mA/cm2 x 10-3) (2) Efficiency displayed in units of % x 10-3 (3) This solvent is 2% pyridine in chlorobenzene 190
PAGE 191
Number of Layers 0123456 7 Film Thickness (nm) 0 10 20 30 40 Figure 3-35. Film thickness fo r multi-layer hybrid films. # SEM Image AFM Image 1 Figure 3-36. SEM and AFM surface im ages of multi-layer hybrid films. 191
PAGE 192
# SEM Image AFM Image 3 5 7 Figure 3-36 continued. SEM and AFM surface images of mu lti-layer hybrid films. 192
PAGE 193
a) b) c) d) e) f) Figure 3-37. Optical microscope images for multilayer hybrid films. The number of film layers were a) 1, b) 2, c) 3, d) 4, e) 5, and f) 6. 193
PAGE 194
194 Number of Layers 01234567 RMS Surface Roughness (nm) 5 10 15 20 25 30 35 Figure 3-38. RMS surface roughness for multi-la yer hybrid films. Red squares represent measurements over a 3 x 3 m area and blue circles represent measurements on a 1 x 1 m area.
PAGE 195
CHAPTER 4 CONCLUSIONS AND FUTURE WORK Conclusions This dissertation presented the results of exploratory research on the processing and performance of hybrid PV. It provided encour agement for a more complete study of hybrid photovoltaic devices. With collaboration and assi stance from several teams and individuals, the groundwork was laid for future st udies to continue this project and achieve high-performance hybrid photovoltaic devices. The innovations of this study include the firs t simulations of hybrid photovoltaics using existing semiconductor modeling software, devel opment of anode surface treatment processes, solvent selection for hybrid films, and hybr id bulk heterojuncti on photovoltaic process development, including an interesting multiple spjn coating process sequence to better disperse the inorganic phase. Hybrid Photovoltaic Simulation Simulations of an ordered hete rojunction photovoltaic cell provided interes ting results. It was found that reported values of certain parameters repo rted in the literatur e from organic fieldeffect transistor fabrication were poor estim ates for organic photovol taic simulation. The hole mobility values given in the liter ature [71] proved to be higher than the simulations estimated. This value of 0.01 cm2/V-s for the hole mobility produced J-V curves with fill factors around 0.85, which is considerab ly larger than published values which typically range from 0.4 to 0.6 [27, 33-34, 44, 75]. By reducing the mobility to 1 x 10-4 cm2/V-s the fill factor was reduced to 0.78, which is closer to the published range. The short-circuit current density of the real cell could not be matched by the simulations. All attempts using the two-step model gave values lower than the observed ones, including 195
PAGE 196
simulations that allowed the full P3HT region to generate photocur rent and a 50% increase in the absorption coefficient over a 40 nm exciton diffu sion length. The highest short-circuit current density achieved with this simu lation technique was 1.47 mA/cm2 from a cell featuring a 40 nm LD region, but this value is approximately 0.8 mA/cm2 lower than the real ce ll. It is believed that the two-step simulation process creates a strong attractive for ce between charges when very high levels of charges are generated along a na rrow line, resulting in increased annihilation between the free carriers. Anode Surface Treatment Treatment of the ITO substrate with a N2 plasma was optimized to grant a smooth, stable surface, which served as the basis of cell fabrication. This work was done in close collaboration with Jiyoun Seol at Yeungnam Univ ersity. The chemically cleaned substrate was subjected to nitrogen plasma at 50 W and 200 mTorr for 30 min. This process resulted in a smoother, more hydrophilic surface which aided deposition of furthe r layers, and an increase in nitrogen content which lowered the film work function and eased the pathway for hole current. This treatment procedure was used for all hybrid film and hybrid photovoltaic studies that followed to ensure a stable surface and allow direct comparison of the results. Solvent Selection The solvents TCE, THF, benzene, toluene, o-xylene, chlorobenzene, o-dichlorobenzene, and chloroform were compared for their app licability to hybrid film deposition. Although no correlations were found between film propert ies and solvent polarity or solvent boiling temperature, certain solvents were identified as strong candidates for these films. These included chloroform, and chlorobenzene, which were eac h used in further studies. These solvents produced films with low surface roughness at all measurement scales, from microns to 196
PAGE 197
millimeters. These solvents provided simultaneous dissolution both the organic and inorganic phases which limited early precipitati on of the solutes, resulting in a more uniform film surface. Hybrid Bulk Heterojunction Photovoltaic Development Hybrid photovoltaic cells were fa bricated with varying film pr operties. It was found that the best cells featured thick absorber films spin-coated from solutions of at least 25 mg/ml with a 50:50 weight ratio between P3HT and the nano-CdSe. Cells consisting of pure P3HT sandwiched between the PEDOT:PSS and Al layers showed maximum performance approximately an order of magnitude lower than the best hybrid cells. Cells fabricated from commercial CdSe na nopowder performed at less than 20% of the best cells using in-house CdSe nanopowder coated with pyridine, althoug h a direct comparison between the two is difficult because the surf ace passivation material remains unknown for the commercial nanopowder. Hybrid cells fabricated with chlorobenzene solution rath er than chloroform saw a 60% reduction is efficiency despite an increase in solution concentration from 25 to 30 mg/ml. Despite JSC and fill factor values that were nearly th e same as that of the chloroform cell, the chlorobenzene cell showed a VOC of 0.3 V lower than the chloroform cell. The best cell was fabricated from a 25 mg/ml solution of 50% CdSe in P3HT dissolved in 2% pyridine in chloroform. The process was ta ilored to minimize the amount of time the cell was exposed to atmosphere during fabrication and between fabrication and test ing. It showed a maximum efficiency of 4.16 x 10-2 %, JSC of 0.205 mA/cm2, VOC of 0.705 V, and a fill factor of 0.288. 197
PAGE 198
Future Work Although the field of organic photovoltaics is rapidly grow ing and advancing, several research directions are suggested by this work for its continua tion. This section describes some potential research directions stemming fr om the work presented in this dissertation. Organic Photovoltaic Simulations The models presented in this dissertation u sed the powerful modeling software package Medici. Although the software h as powerful optical and electric al simulation abilities, this research seems to have pushed its limits by dema nding nano-scale material specifications and low levels of free carriers, carrier mobilities, and current flow in the devices. Simulation attempts were frequently cut short due to conve rgence issues in the soft ware under the specified conditions. Additionally, a key component of th e modeling work focused on the correct way to simulate the effect of excitons in Medici which are not explicit in the package. In light of these difficulties, a new software package designed to simulate organic electronic materials would be a great boost to this work. A product such as Fluxim [94] would be able to more accurately simulate the effect of excitons in these hybrid devices. Another direction that should be pursued is the variation of the geometry and materials of the simulated cells. While ZnO is a well-researc hed material for the growth of aligned and ordered nanowires, these stru ctures are now being grown for other materials with stronger absorption spectra such as CdSe and InP [66, 95]. These materials could see more use in nanowire hybrid cells in the fu ture, driving the need for eff ective simulations to study device properties. The most common organic cell design is cu rrently the bulk heterojunction cell using semiconductor nanoparticles or soluble C60 derivatives. This design presents a challenge for simulations because little work has been done to characterize the pa rticle distribut ions in these 198
PAGE 199
films. Because the physical dime nsions of the model are vital for accurate simulation, this provides both a limitation and an opportunity. On the one hand, this lack of information limits the results that can be generated through simulati ons. On the other hand, if an accurate, robust model can be developed for a hybrid material system, it could be used to back-calculate unknown physical dimensions of the system. Development of Hybrid Photovoltaic Cells A central theme of this dissertation is the morphology control of hybrid films for photovoltaic applications. This study only focused on nearly sphe rical nanocrystals, but future studies should expand the field to include nanoparticles with dimensionality. Shaped nanoparticles, nanowires, tetrapods and more exotic branched st ructures are bei ng grown using techniques similar to the one used for spheri cal nanocrystals in th is study [35-37]. These dimensional crystals offer the promise of direct ed charge transport without the multiple electron jumping processes required for sm all spherical nanocrystals. Alternative semiconductor nanoparticles should be considered as well. Although CdSe is one of the easiest particles to be synthesized, other compound semiconductors such as CdS, CdTe, PbSe, and CIS can be grown on the nano scal e. Some semiconductors such as PbSe have demonstrated multi-carrier generation on s hort timescales that offer the possibility of constructing photovoltaic devices with quantum e fficiencies greater than 1 if these charges can be harvested [96-97]. Regioregular P3HT, as used in this study, is curren tly the most promising candidate for polymer in electronic devices such as organic thin-film transist ors and solar cells [33-34]. However, this polymer shows some limitations fo r solar applications. The absorption spectrum shows a cut-off above 650 nm, limiting absorption of near-IR photons which are plentiful in the solar spectrum. Although the hole mobility of P3HT is high compared to many conductive 199
PAGE 200
200 polymers, it is several orders of magnitude lower than most inorganic semiconductors and limits charge collection in devices. Additiona lly, it degrades quickly under water and oxygen atmospheres, particularly in the presence of ra diation. As new polymer s are developed with a broader absorption spectrum, improved carrier m obility, improved environmental resistance, and the ability to deposit high-quality films with strong adhesion and low surface roughness they will quickly find relevance in or ganic photovoltaics research. Regardless of the targeted fu ture research direction, some processing equipment changes should be considered. The recent installation of a photolithographic patterning system will allow for the fabrication of multiple cells on single substrates, and this will greatly improve the speed of research and also the quality of the devices fabricated. As long as P3HT is used for the active layer film, all processing equipment should be set up under an inert atmosphere of argon or nitrogen in a glove box. Because of the sensitiv ity of this polymer to water and oxygen, highquality device fabrication at this scale requires that it be protec ted from exposure at all phases of the process. This is not true fo r the HTL layer of PEDOT:PSS, which is in fact deposited from a water solution and shows no negative effects of short-term atmospheric exposure. All processing steps beyond the deposition of the HTL should be contained in this glove box, including mixing for the hybrid solution, depos ition and drying of the hybrid films, possible inclusion of exciton blocking layers back electrode deposition, and cell characterization. This is a difficult challenge due to the size of some pro cessing equipment, but this is the design used by groups fabricating world-record organic photovo ltaic cells. The processing equipment line currently in use already involv es an evaporation chamber opening directly into the glove box, which is the largest piece of equipment used in the process. In order to improve these cells to the highest performance level, this environmental protect ion is a step that absolutely must be taken.
PAGE 201
LIST OF REFERENCES 1. International Energy Agency, Trends in Photovoltaic Applicat ions, Report IEA-PVPS T1-16:2007, August 2007. 2. U.S. Department of Energy, Energy Efficiency and Renewable Energy, http://www1.eere.energy.gov/solar/deployment.html (2006) July 2008. 3. Solarbuzz, LLC., www.solarbuzz.com/moduleprices.htm (2008) July 2008. 4. C.W. Tang, Appl. Phys. Lett. 48 (1986) 183. 5. Md.K.H. Bhuiyan, T. Mieno, Th in Solid Films 441 (2003) 187. 6. N.S. Sariciftci, L. Smilowitz, A.J. Heeg er, F. Wudl, Synth. Met. 59 (1993) 333. 7. B. ORegan, M. Gratze l, Nature 335 (1991) 737. 8. M.K. Nazeeruddin, P. Pechy, T. Renouard, S. M. Zakeeruddin, R. Humphry-Baker, P. Comte, P. Liska, L. Cevey, E. Costa, V. Shklover, L. Spiccia, G.B. Deacon, C.A. Bignozzi, M. Grtzel, J. Am Chem. Soc. 123 (2001) 1613. 9. K. Hara, M. Kurashigo, Y. Dan-oh, C. Kasasa, Y. Ohga, A. Shinpo, S. Suga, K. Sayama, H. Arakawa, New J. Chem. 27 (2003) 783. 10. Q.B. Meng, K. Takahashi, X.T. Zhang, I. Su tanto, T.N. Rao, O. Sato, A. Fujishima, H. Watanabe, T. Nakamori, Uragami Langmuir 19 (2003) 3572. 11. L.J. Rothberg, M. Yan, F. Papadimitrakopoulos M.E. Galvin, E.W. Kwock, T.M. Miller, Synth. Met. 80 (1996) 41. 12. J.J.M. Halls, K. Pichler, R.H. Friend, S.C. Moratti, A.B. Holmes, Appl. Phys. Lett. 68 (1996) 3120. 13. L.A. Pettersson, L.S. Roman, O. I nganas, J. Appl. Phys. 86 (1999) 487. 14. J.J.M. Halls, C.A. Walsh, N.C. Greenham, E.A. Marseglia, R.H. Friend, S.C. Moratti, A.B. Holmes, Nature 376 (1995) 498. 15. T.J. Savenije, J.M. Warman, A. Goo ssens, Chem. Phys. Lett. 287 (1998) 148. 16. M. Theander, A. Yartsev, D. Zigmantas, V. Sundstrom, W. Mammo, M.R. Andersson, O. Inganas, Phys. Rev., B 61 (2000) 12957. 17. H.L. Wong, L.S.M. Lam, K. W. Cheng, K.Y.K. Man, W. K.Chan, C.Y. Kwong, A.B. Djurisic, App. Phys. Lett. 84 (2004) 2557. 18. G.D. Sharma, R. Kumar, S.K. Sharma, M.S. Roy, S. En. Mater. & Solar Cells 90 (2006) 933. 201
PAGE 202
19. P. Peumans, S. Uchida, S.R. Forrest, Nature 425 (2003) 158. 20. R.P. Raffaelle, B.J. Landi, J.D. Harris, S.G. Bailey, A.F. Hepp, Mat. Sci. & Eng., B 116 (2005) 233. 21. A.D. Pasquier, H.E. Unalan, A. Kanwal, S. Miller, M. Chhowalla, App. Phys. Lett. 87 (2005) 203511. 22. N.S. Sariciftci, L. Smilowitz, A.J. Heeger, F. Wudl, Science 258 (1992) 1474. 23. N.S. Sariciftci, L. Smilowitz, A.J. Heeg er, F. Wudl, Synth. Met. 59 (1993) 333. 24. G. Yu, K. Pakbaz, A.J. Heeger, Appl. Phys. Lett. 64 (1994) 3422. 25. J.C. Hummelen, B.W. Knight, F. LePeq, F. Wudl, J. Yao, C.L. Wilkins, J. Org. Chem. 60 (1995) 532. 26. G. Yu, J. Gao, J.C. Hummelen, F. W udl, A.J. Heeger, Science 270 (1995) 1789. 27. S.E. Shaheen, C.J. Brabec, N.S. Sariciftci, F. Padinger, T. Fromherz, J.C. Hummelen, Appl. Phys. Lett. 78 (2001) 841. 28. N. Camaioni, G. Ridolfi, G. Casalbore-Miceli, G. Possamai, L. Garlaschelli, M. Maggini, Solar Energy Materials & Solar Cells 76 (2003) 107. 29. M. Al-Ibrahim, H.K. Roth, M. Schroedner, A. Konkin, U. Zhokhavets, G. Gobsch, P. Scharff, S. Sensfuss, Orga nic Electronics 6 (2005) 65. 30. W. Geens, T. Aernouts, J. Poortmans, G. Hadziioannou, Thin Solid Films, 403 (2002) 438. 31. D. Gebeyehu, F. Padinger, C.J. Brabec, T. Fr omherz, J.C. Hummelen, N.S. Sariciftci, Int. J. of Photoenergy 1 (1999) 1. 32. A.J. Mozer, N.S. Sariciftci, C.R. Chimie 9 (2006) 568. 33. F. Padinger, R.S. Rittberger, N.S. Sari ciftci, Adv. Funct. Mat. 13 (2003) 85. 34. G. Li, V. Shrotriya, J. Huang, Y. Yao, T. Moriarty, K. Emery, Y. Yang, Nat. Mat. 4 (2005) 864. 35. M. Chen, Yi Xie, J. Lu, Y. Xiong, S. Zhang, Y. Qian, X. Liu, J. Mat. Chem. 12 (2002) 748. 36. L. Manna, D.J. Milliron, A. Me isel, E.C. Scher, A.P. Alivis atos, Nat. Mat. 2 (2003) 382. 37. K.A. Dick, K. Deppert, M.W. Larsson, T. Ma rtensson, W. Seifert, L.R. Wallenberg, L. Samuelson, Nat. Mat. 3 (2004) 380. 202
PAGE 203
38. W.J.E. Beek, L.H. Slooff, M.M. Wienk, J.M. Kroon, R.A.J. Janssen, Adv. Funct. Mater. 15 (2005) 1703. 39. P.A.C. Quist, L.H. Slooff, H. Donker, J.M. Kroon, T.J. Savenije, L.D.A. Siebbeles, Superlattices and Microstr uctures 38 (2005) 308. 40. A. Watt, H. Rubinsztei n-Dunlop, P. Meredith, Mat. Lett. 59 (2005) 3033. 41. X. Peng, J. Wickham, A.P. Alivisat os, J. Am. Chem. Soc. 120 (1998) 5343. 42. Z.A. Peng, X. Peng, J. Am Chem. Soc. 123 (2001) 183. 43. M. Kuno, J.K. Lee, B.O. Dabbousi, F.V. Mikulec, M.G. Bawendi, J. Chem. Phys. 106 (1997) 9869. 44. W.U. Huynh, J.J. Dittmer, W.C. Libby, G.L. Whiting, A.P. Alivisatos, Adv. Funct. Mater. 13 (2003) 73. 45. H.J. Choi, J.K. Yang, H.H. Park, Thin Solid Films 494 (2006) 207. 46. K.M. Coakley, M.D. McGehee, App. Phys. Lett. 83 (2003) 3380. 47. K.M. Coakley, Y. Lui, M.D. McGeehee, K.M. Frindell, G.D. Stucky, Adv. Funct. Mater. 13 (2003) 301. 48. J.J. Wu, S.C. Liu, Adv. Mater., Vol. 14 (2002) 215. 49. W. Lee, M.C. Jeong, J.M. Myoung, Acta Mater. 52 (2004) 3949. 50. C. Geng, Y. Jiang, Y. Yao, X. Meng, J.A. Za pien, C.S. Lee, Y. Lifshitz, S.T. Lee, Adv. Funct. Mater. 14 (2004) 589. 51. S.C. Lyu, Y. Zhang, C.J. Lee, H. Ruh, H.J. Lee, Chem Mater. 15 (2003) 3294. 52. L. Vayssieres, K. Keis, A. Hagfeldt, S. Lindquist, Chem. Mater. 13 (2001) 4395. 53. Y. Tak, K. Yong, C. Park, J. Crystal Growth 285 (2005) 549. 54. M. Wang, E.J. Kim, S.H. Hahn, C. Park, Physica Status Solidi (a) 203 (2006) 2418. 55. R. Corkish, D.S.P. Chan, M.A. Green, J. Appl. Phys. 79 (1996) 195. 56. Y. Zhang, A. Mascarenhas, S. Deb, J. Appl. Phys. 84 (1998) 3966. 57. M. Burgelman, B. Minnaert, Thin Solid Films 511 (2006) 214. 58. B. Kannan, K. Castelino, A. Ma jumdar, Nano Lett. 3 (2003) 1729. 203
PAGE 204
59. W. Geens, T. Martens, J. Poor tmans, T. Aernouts, J. Manca, L. Lutsen, P. Heremans, S. Borghs, R. Mertens, D. Vanderzan de, Thin Solid Films 451 (2004) 498. 60. B. Mazhari, S. En. Mater. & Solar Cells 90 (2006) 1021. 61. A. Takshi, A. Dimopoulos, J.D. Madden, Solid-State Electronics 52 (2008) 107. 62. W.U. Huynh, J.J. Dittmer, N. Teclemariam, D.J. Milliron, A.P. Alvisatos, K.W.J. Barnham, Phys. Rev., B 67 (2003) 115326. 63. R.D. Bettignies, J. Leroy, M. Firon, C. Se ntein, Syn. Met. 156 (2006) 510. 64. V. Lemaur, M. Steel, D. Belj onne, J.L. Bredas, J. Cornil, J. Am. Chem. Soc. 127 (2005) 6077. 65. Medici Users Manual, Avant Corporati on, Beaverton, OR, Release 2001.4, (2000). 66. W.K. Kim, Ph.D. Dissertation, University of Florida, Gainesville, FL, (2006). 67. E.J. Meijer, C. Detcheverry, P.J. Baesjou, E.V. Veenendaal, D.M. de Leeuw, T.M. Klapwijk, J. Appl. Phys. 93 (2003) 4831. 68. H. Sirringhaus, N. Tessler, R.H. Friend, Science 280 (1998) 1741. 69. C. Tanase, E.J. Meijer, P.W.M. Blom, D. M. de Leeuw, Org. Elec. 4 (2003) 33. 70. T.A. Chen, X. Wu, R.D. Reike, J. Am. Chem. Soc. 117 (1995) 233. 71. A.J. Casico, J.E. Lyon, M.M. Beerbom, R. Sc hlaf, Y. Zhu, S.A. Jenekhe, App. Phys. Lett. 88 (2006) 62103. 72. A.J.C. Wen, K.L. Chen, M.H. Yang, W.T. Hsiao, L.G. Chao, M.S. Leu, Surf. Coat. Tech. 198 (2005) 362. 73. Melles Griot Inc., http://mellesgriot.com/p roducts/optics/oc_5_1.htm (2002) July 2008. 74. Kim, Y., S.A. Choulis, J. Nelson, D.D.C. Bradle y, S. Cook, and J.R. Durrant, J. Mat. Sci. 40 (2005) 1371. 75. D.C. Olson, J. Piris, R.T. Collins, S.E. Shaheen, D.S. Ginley, Thin Solid Films 496 (2006) 26. 76. GetData Graph Digitizer, http://getdata-graph-digitizer.com (2008) July 2008. 77. K.R. Amundson, B.J. Sapjeta, A.J. Lovinger, Z. Bao, Thin Solid Films 414 (2002) 143. 78. S. Li, Semiconductor Physical Electronics, 2nd Ed., Springer, 2006. 204
PAGE 205
205 79. C.J. Brabec, A. Cravino, D. Meissner, N.S. Sariciftci, T. Fromherz, M.T. Rispens, L Sanchez, J.C. Hummelen, Adv. Funct. Mater. 11 (2001) 374. 80. J. Seol, M.L. Monroe, T.J. Anderson, M.A. Hasnain, C.H. Park, Conference Record of the 2006 IEEE 4th World Conference on Photovoltaic Energy Conversion 1 (2006) 236. 81. C.W. Tang, S.A. VanSlyke, C.H. Chem, J. Appl. Phys. 65 (1989) 3610. 82. Kobayashi, H. Hori, and Y. Tanaka, in Handbook of Liquid Crystal Research, edited by P.J. Collings and J.S. Patel (Oxford Univ ersity Press, New York, 1997), Chap. 10, 415. 83. H.S. Kim, J.H. Lee, C.H. Park, J. Kor. Phys. Soc. 41 (2002) 395. 84. J.S. Kim, R.H. Friend, F. Cacial li, J. Appl. Phys. 86 (1999) 2774. 85. K. Sugiyama, H. Ishii, Y. Ouchi, J. Appl. Phys. 87 (2000) 295. 86. W.W. Flack, D.S. Soong, A.T. Bell, D. W. Hess, J. Appl. Phys. 56 (1984) 1199. 87. U. Mitschke, P. Bauerle, J. Mat. Chem. 10 (2000) 1471. 88. P. Peumans, A. Yakimov, S.R. Fo rrest, J. App. Phys. 93 (2003) 3693. 89. E. Kymakis, E. Kourdoumas, I. Franghiadak is, S. En. Mater. & Solar Cells 90 (2006) 1705. 90. Phenomenex, Solvent Miscibility Table, http://designerdrugs.com/pte/12.162.180.114/dcd/ pdf/solvent.miscibility.pdf July 2008. 91. Z.A. Peng, X. Peng, J. Am Chem. Soc. 123 (2001) 183. 92. Meliorum Technologies, Inc., www.meliorum.com (2008) July 2008. 93. Fluxim, www.fluxim.com (2008) July 2008. 94. H. Burkhard, H.W. Dinges, E. K uphal, J. App. Phys. 53 (1982) 655. 95. R.J. Ellingson, M.C. Beard, J.C. Johnson, P. Yu, O.I. Micic, A.J. Nozik, A. Shabaev, A.L. Enfros, Nano Lett. 5 (2005) 865. 96. R.D. Schaller, V.I. Klimov, Phys. Rev. Lett. 92 (2004) 186601.
PAGE 206
BIOGRAPHICAL SKETCH Matthew Lowell Monroe was born in 1978, in Ma rietta, Georgia, to Ronald L. and Debbie B. Monroe. He earned a Bachelor of Science degree from the Ch emical Engineering Department at the Georgia Institute of Tec hnology in Atlanta in 2002. He jo ined the Chemical Engineering Department at the University of Florida in 2002 and joined Dr. Andersons research group in 2003. He earned a Doctor of Philosophy in chemical engineering in 2008.
|
