Embed Size (px)
Citation preview
Materials interface engineering in perovskite photovoltaics
by
Jixian XU
A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy
The Edward S. Rogers Sr. Department of Electrical & Computer Engineering
University of Toronto
© Copyright by Jixian XU 2017
ii
Materials interface engineering in perovskite photovoltaics
Jixian XU
Doctor of Philosophy
The Edward S. Rogers Sr.
Department of Electrical & Computer Engineering »
University of Toronto
2017
Abstract
Solar photovoltaics (PV) offer a sustainable solution to the daunting challenge of meeting
the global energy demand. Perovskite solar cells, whose high efficiencies are attainable via
low-cost and high-throughput solution processing, are an emerging technology that has
captivated the PV research community. Further advances in efficiency are limited by the
abundant interfaces that make up these polycrystalline devices. Important issues in
perovskite device operation, such as instability and hysteresis, arise from perovskites’ ionic
nature, and need to be addressed for this technology to fulfill its potential.
In this thesis, I explore interfaces within perovskite devices: grain boundaries, and
electron- and hole-extraction junctions. With the aid of density functional theory (DFT)
simulations and nano-probe characterization, I provide insight into the origins of defect
formation and hysteresis. By leveraging these findings, I demonstrate control of film
growth conditions and interface materials chemistry to create new device architectures
with improved performance. The DFT-based analysis of defect formation energies
identifies the key defects (Pb atom substituted by I, known as antisites) and indicates that
films grown under iodine-rich conditions are prone to forming deep electronic traps. This
iii
finding motivated my exploration of a new precursor (anhydrate lead acetate) for device-
quality films.
I then report the first perovskite-PCBM hybrid solid. Here, I find that PCBM, when it
infiltrates throughout the grain boundaries and electron-extraction interfaces, suppresses
hysteresis in devices. Materials characterization and DFT simulations reveal the PCBM-
perovskite interaction: the PCBM passivates the key defects during the perovskite self-
assembly. Using conductive AFM, I reveal the memristive properties of perovskite films
and identify the major origin of hysteresis as ion, especially halide, migration.
I close by developing the first crosslinked hole-extraction top contact with the goal of
obviating degradation of the underlying perovskite. A remote-doping strategy introduces
the needed hole conductivity. The new top contact produces an insoluble and heat-resistant
protecting interlayer that is band-aligned with the perovskite. The resultant family of
devices is hysteresis-free, with fill factors exceeding 80% and resilience to thermal stresses
that exceed 100˚C, conditions under which conventionally-contacted devices fail. This top
contact methodology also paves the way for building multi-junction devices on top of the
perovskite cell. I close this work by offering a roadmap for future improvements in
perovskite photovoltaics.
iv
Acknowledgments
Firstly, I give my sincere thanks to my supervisor, Prof. Ted Sargent, for providing me with the
opportunity to work with fantastic colleagues and for guiding me throughout my doctoral work.
His passion and broad view of research encouraged me to solve problems across multi-
disciplinary fields. His ambitions to address world-class challenges inspired me to realize my
potential.
I also thank all the professors that have spent time on my courses and committees, and
collaborated with me in my research. I would like to acknowledge the support from the Edward
Rogers Sr. Graduate Scholarship and the Hatch Graduate Scholarship.
I feel so lucky to have worked with remarkable people through these years. A big thanks to
Andrei Buin and Oleksandr Voznyy for collaboration on DFT and optoelectronic simulations and
related publications. I give special thanks to Alex Ip, Brandon Sutherland and Grant Walters for
close collaboration in projects and patience with manuscript revision. I would like to thank Dr.
Zhijun Ning for ramping me up when I first started in the group. I also would like to thank Dr.
Sjoerd Hoogland, Larissa Levina, Elenita Palmiano, Damir Kopilovic, and Remi Wolowiec for
their continuous assistance with key aspects of my research through these years. As well, I would
like to thank Wei Li, Riccardo Comin, Xinzheng Lan, and Mingjian Yuan for their insights on
chemistry and physics. Thanks to the unsung heroes that keep the group running, Jeannie and
Stacy. Thanks and best of luck to those I had a chance to discuss and work alongside, Valerio
Adinolfi, Chris Wong, André Labelle, Susanna Thon, Illan Kramer, David Zhitomirsky,
Pongsakorn Kanjanaboos, Jeffrey McDowell, Lisa, Xiwen Gong, Mengxia Liu, Lina Quan, Min
Liu, Bo Zhang, Fengjia Fan, Zhenyu Yang, and Haopeng Dong. Although I won’t be able to
name all of them here, I am grateful for contributions from everyone.
I am grateful to my family, Mom, Dad, brother and sisters for their devotion and wholehearted
support.
Finally, to my beloved wife, Bin Wang, and our kids. I dedicate this thesis to them in their trust
and company.
v
Contributions
1. A. Buin, P. Pietsch, J. Xu, O. Voznyy, A. H. Ip, R. Comin, E. H. Sargent. Materials
Processing Routes to Trap-Free Halide Perovskites. Nano Letters 2014, 14, 6281.
This work is mainly contained in Chapter 3. It focused on identifying the growth conditions
for producing low-trap-density, device-level films for planar PV, from theoretical and
experimental perspectives. A. Buin led the project and performed the DFT simulations. P.
Pietsch and I performed the experiments. I developed the material processing protocol of
using a new precursor to achieve the device-quality films and conducted the PL
measurements. I collaborated with Andrei to define the key questions and strategies in the
DFT studies.
2. J. Xu, A. Buin, A. H. Ip, W. Li, O. Voznyy, R. Comin, M. Yuan, S. Jeon, Z. Ning, J. J.
McDowell, P. Kanjanaboos, J.-P. Sun, X. Lan, L. N. Quan, D. H. Kim, I. G. Hill, P.
Maksymovych, E. H. Sargent. Perovskite-fullerene hybrid materials suppress hysteresis in
planar diodes. Nat Commun 2015, 6, DOI 10.1038/ncomms8081.
This work is mainly contained in Chapter 4. It focused on the materials chemistry of the
grain boundaries and electron-extraction interfaces to reveal the origins of hysteresis in
perovskite devices. I conceived the idea and conducted the experimental design, analysis, and
manuscript preparation, in collaboration with co-authors. Density functional theory
calculations were performed by A. Buin. Peter Maksymovych and Jon-Paul Sun carried out
cAFM studies.
3. J. Xu, O. Voznyy, R. Comin, X. Gong, G. Walters, M. Liu, P. Kanjanaboos, X. Lan, E. H.
Sargent. Crosslinked Remote-Doped Hole-Extracting Contacts Enhance Stability under
Accelerated Lifetime Testing in Perovskite Solar Cells. Advanced Materials 2016, 28, 2807.
This work is mainly contained in Chapter 5. It focused on the new design of a hole-
extraction interface to improve device stability and performance simultaneously. I initiated
the idea and conducted the experimental design, analysis, and manuscript preparation in
collaboration with co-authors. O. Voznyy, R. Comin, and X. Gong assisted the theoretical
studies and optoelectronic simulations.
vi
Table of Contents
Table of Contents
Acknowledgments.......................................................................................................................... iv
Contributions....................................................................................................................................v
Table of Contents ........................................................................................................................... vi
List of Tables ................................................................................................................................. ix
List of Figures ..................................................................................................................................x
List of Common Acronyms and Symbols ..................................................................................... xii
Chapter 1 ..........................................................................................................................................1
Introduction .................................................................................................................................1
1.1 Solar photovoltaics: a carbon-neutral energy capture strategy ............................................1
1.2 From wafer-based to solution-processed photovoltaics .......................................................2
1.3 Theoretical and practical limits in photovoltaic efficiency ..................................................4
1.4 Materials interface engineering for solution-processed photovoltaics ................................6
1.4.1 Charge-separating interfaces ....................................................................................7
1.4.2 Interfaces within the absorbers: grain boundaries ...................................................8
1.5 Roadmap of thesis: engineering interfaces in perovskite photovoltaics ..............................9
Chapter 2 ........................................................................................................................................11
Perovskite photovoltaics ...........................................................................................................11
2.1 Crystal structure and charge transport ...............................................................................11
2.2 Device architectures ...........................................................................................................13
2.3 Challenges in perovskite photovoltaics .............................................................................14
2.3.1 Stability ..................................................................................................................14
2.3.2 Hysteresis ...............................................................................................................15
2.4 Research goals and methodology.......................................................................................15
vii
Chapter 3 ........................................................................................................................................17
Materials processing for low-trap-density perovskite solids ....................................................17
3.1 Introduction ........................................................................................................................17
3.2 Perspective from DFT on perovskite growth .....................................................................17
3.3 Growth conditions control for low-trap perovskite film ....................................................22
3.3.1 Diffusion length .....................................................................................................23
3.3.2 Crystalline morphology .........................................................................................25
3.4 Conclusions ........................................................................................................................27
Chapter 4 ........................................................................................................................................28
Fullerene-perovskite interaction at electron-extraction interfaces ............................................28
4.1 Introduction ........................................................................................................................28
4.2 Improvement of hysteresis and photovoltaic performance ................................................29
4.3 Mechanistic studies of perovskite-PCBM interaction .......................................................31
4.3.1 Material characterization .......................................................................................31
4.3.2 DFT simulations.....................................................................................................32
4.4 Perovskite-PCBM mixture phase distribution ...................................................................33
4.5 Charge dynamics and hysteresis characterization ..............................................................35
4.6 Discussion: Ionic motion and hysteresis in perovskites ....................................................39
Chapter 5 ........................................................................................................................................40
Crosslinked hole-extraction interface improves hysteresis and stability ..................................40
5.1 Introduction ........................................................................................................................40
5.2 Crosslinked interface on perovskite top surface ................................................................41
5.3 Remote doping for hole-extraction conductivity ...............................................................44
5.4 Efficient PV with reduced hysteresis .................................................................................45
5.5 Mechanistic study of remote-doped hole-extraction ..........................................................48
5.5.1 Material characterization .......................................................................................48
viii
5.5.2 Optoelectronic simulations ....................................................................................48
5.6 Improved stability under external stress ............................................................................50
5.7 Conclusions ........................................................................................................................54
Chapter 6 ........................................................................................................................................55
Conclusions ...............................................................................................................................55
6.1 Summary and Impact .........................................................................................................55
6.2 Outlook for perovskite solids and PV ................................................................................56
6.2.1 Maximum efficiency ..............................................................................................56
6.2.2 Long-term stability.................................................................................................57
References ......................................................................................................................................59
Appendices .....................................................................................................................................65
ix
List of Tables
Table 3-1. Computed defect formation energies under (a) I-poor (Pb-rich) (b) I-rich (Pb-poor)
conditions. ..................................................................................................................................... 21
Table 4-1. Statistics of steady-state performance with different PCBM distribution and thickness
....................................................................................................................................................... 30
x
List of Figures
Figure 1-1. Estimated potential of various renewable energy sources.. ......................................... 1
Figure 1-2. The Sun’s power spectrum reaching the Earth and the limits of solar conversion. ..... 4
Figure 1-3. Photogenerated charge carrier transport in a single-junction solar cell.. ..................... 5
Figure 1-4. The current–voltage characteristic and definition of important terms in photovoltaics
......................................................................................................................................................... 6
Figure 1-5. Architectures and interfaces in solution-processed solar cells. .................................... 8
Figure 1-6. Outline of thesis: material interfaces engineering in perovskite PV .......................... 10
Figure 2-1. Crystal structure of perovskites with the generic chemical formula ABX3.. ............. 12
Figure 2-2. Perovskite solar cell architectures evolution. ............................................................. 13
Figure 2-3. Instability and hysteresis in perovskite solar devices. ............................................... 14
Figure 2-4. The framework of full interfaces in the perovskite solar cells. ................................. 16
Figure 3-1. Surface states of the Pb halide perovskites. ............................................................... 18
Figure 3-2. Tetragonal perovskite, its formation from experimentally employed precursors, and
its defect energy levels. ................................................................................................................. 19
Figure 3-3. Formation energies and volume densities of key defects in tetragonal lead perovskites
....................................................................................................................................................... 22
Figure 3-4. Physical configuration of PbI neutral antisites in the tetragonal phase ...................... 23
Figure 3-5. Experimental investigation of transport in novel iodide-poor perovskite films. ....... 25
Figure 3-6. Non-continuous perovskite films grown using PbI2 precursor .................................. 26
Figure 3-7. Dense and smooth perovskite planar films grown using anhydrate lead acetate. ...... 26
xi
Figure 4-1. Steady-state photovoltaic performance of an ultra-thin perovskite-PCBM hybrid film
....................................................................................................................................................... 29
Figure 4-2. Solution-processed planar device structures in this study .......................................... 31
Figure 4-3. Perovskite-PCBM hybrid process and in situ passivation mechanism.. .................... 32
Figure 4-4. 3D phase separation and homogeneous PCBM distribution in hybrid solid ............. 34
Figure 4-5. PCBM phase separation at perovskite grain boundaries ............................................ 35
Figure 4-6. cAFM study of hysteresis-ion relationship for control films and hybrid films.......... 36
Figure 4-7. Long-term steady-state dark current measurement of planar devices. ....................... 37
Figure 4-8. Effect of PCBM on charge carrier dynamics ............................................................. 38
Figure 5-1. Hole extraction contact employing material crosslinking and interface doping ........ 43
Figure 5-2. Thermally crosslinked VNPB is insoluble and enables the layer-by-layer deposition
....................................................................................................................................................... 44
Figure 5-3. Ultraviolet photoelectron spectroscopy (UPS) studies of VNPB layer ..................... 45
Figure 5-4. Improved photovoltaic performance with interface doping ....................................... 46
Figure 5-5. Electrical simulation of devices using interface doping ............................................ 49
Figure 5-6. Evolution of performance, morphology and material under external stress .............. 51
Figure 5-7. Assessment of device evolution under the external solvent attack ............................ 53
Figure 5-8. Evolution of material and morphology under the external solvent attack. ................ 53
Figure 6-1. Fraction of Shockley-Queisser detailed-balance limit for voltage and current
achieved by record cells. ............................................................................................................... 57
xii
List of Common Acronyms and Symbols
PV – photovoltaic
PCE – power conversion efficiency
VOC – open circuit voltage
JSC – short circuit current density
FF – fill factor
MPP – maximum power output point
EQE – external quantum efficiency
DFT – density functional theory
Ef – Fermi level
Eg – semiconductor’s bandgap
WF – work function
VBM – valence band maximum
CBM – conduction band minimum
ETL – electron transport layer
HTL – hole transport layer
PCBM – phenyl-C61-butyric acid methyl ester
spiro-MeOTAD – 2,2’,7,7’-Tetrakis(N,N-di-p-methoxyphenylamine)-9,9’-spirobifluorene
VNPB – crosslinked N4,N4' -Di(naphthalen-1-yl)-N4,N4' -bis(4-vinylphenyl)biphenyl-4,4'-
diamine
DMF – N,N-Dimethylformamide
cAFM – conductive atomic force microscope
XPS – X-ray photoelectron spectroscopy
UPS – ultraviolet photoelectron spectroscopy
SIMS – secondary ion mass spectrometry
1
Chapter 1
Introduction
1.1 Solar photovoltaics: a carbon-neutral energy capture strategy
Identifying clean energy sources to meet the world’s growing demand is one of society’s
foremost challenges. Projected worldwide population expansion and economic growth will more
than double the current global energy consumption rate to ~30 TW by 2050, even with
aggressive conservation. The threats of global warming and climate change due to excessive
greenhouse gas (such as CO2) emission impose a second requirement on new energy resources.
To balance the increased energy consumption rate and CO2 emission rate, 15 TW will need to be
derived entirely from zero-carbon-intensity (C-neutral) renewable energy sources by 20501.
Figure 1-1. Estimated potential of various renewable energy sources. For a fair comparison, the potential of each
power source (in units of W) is converted to “equivalent chemical fuel” power (in units of Wc). The estimated
efficiency of this conversion is included in the final comparison. In this regard, if the produced energy is already in
the form of a chemical fuel, such as in solar fuels, then there is no conversion. For energy in an electrical form, such
as in solar electricity, the conversion factor from the electrical to chemical form is 75 %. The conversion factor from
energies in a mechanical form, such as wind power, the conversion factor to a chemical form has a lumped efficiency
Ocean Thermal Gradient
Ocean Tidal
Ocean Salinity Gradient
Ocean SurfaceCurrent
2001 Supply
GeothermalHydropower
Ocean wave
Wind
SolarFuel
Solar Thermal
Solar Electricity
Sources with >15TW extractable and technical
potential
10-6 10-3 1 103 10610-6
10-3
1
103
106
Extractable Potential (TWc)
Tech
nic
al P
ote
nti
al (
TWc)
2
of 25%. This conversion factor includes a 33 % efficiency associated with the conversion to an electrical form followed
by another 75 % efficiency associated with the conversion to the chemical form. Energy generated by heat sources,
such as from ocean thermal power, a Carnot efficiency ηc (second law of thermodynamics) is first used for a conversion
to a mechanical form, such as turbine rotation, followed by another 25 % efficiency associated with a conversion from
the mechanical form to chemical fuel, yielding a ηc·25 % lumped efficiency. Figure reproduced from ref. 1, copyright
2006, U.S. Department of Energy.
Solar energy can be directly captured from sunlight and converted into different forms such as
electricity, chemical fuels, and thermal energy. It dwarfs other renewable energy source
including hydro, ocean, wind, and geothermal. It has the greatest potential to meet the daunting
challenge of 15 TW C-neutral power (Figure 1-1). In fact, sunlight powers geological processes
(atmospheric motion, water and vapor transport, etc.) and therefore is largely responsible for
various secondary renewable energies (wind, ocean wave, and hydropower). In a theoretical
scenario, the total solar energy striking the Earth’s surface (~89 300 TW after the atmospheric
loss) within mere hours could meet the world’s energy needs for an entire year. A net 10%
efficient solar energy farm covering ~0.2% of Earth land would provide the global 15 TW C-
neutral power. Indeed, solar energy is the source with the technical potential safely exceeding 15
TW.
Solar electricity produced from solar photovoltaics is far from fulfilling its huge potential, and
the market is in a rapid growth phase today. The worldwide accumulated capacity of solar
photovoltaics was ~177 GW by 2014, with a total power output equal only to ~1% of the
worldwide electricity demand2.
Advances in science and technology provide ways to accelerate further cost-to-efficiency
reductions in solar photovoltaics. The use of photovoltaic power as an energy source also
requires breakthroughs in cost-effective and scalable energy storage and transport to match time-
and-space varying energy supply and demand3.
1.2 From wafer-based to solution-processed photovoltaics
Crystalline silicon (c-Si) solar cells dominate today’s PV production, having advanced greatly in
the past half-century. The best performance (over 25% in top lab-level devices) is based on the
highest-quality crystal wafers4,5. Because of the cost and stringent processes of making and
handling wafers, the production cost would be substantially reduced if devices could be obtained
3
using films grown on glass or other inexpensive substrates. Silicon’s bandgap (1.1 eV) is indirect
and does not match the solar spectrum optimally, thus requiring a large thickness (~200 μm) of
material to completely absorb light with photon energies above the bandgap. Another wafer-
based solar cell, GaAs, can keep the absorber thickness relatively small (~2 μm) because of its
direct-bandgap character and its bandgap being close to the optimum (1.42 eV) and its high
absorption coefficient. GaAs solar cells hold the highest efficiency record (~28.8%) of single
junction PV cells4,6. The materials cost and energy-intensive processing (epitaxial growth using
chemical vapor deposition), however, so far restrict application to niche markets, such as space
technology.
Direct-bandgap, thin film devices represent second-generation photovoltaics. They benefit from
reduced materials usage and less stringent deposition processes. Because they use high
absorptivity direct-bandgap materials (polycrystalline CdTe, CuInxGa(1-x)Se2 or so-called CIGS,
amorphous Si), the thickness of the absorber can be reduced below one micrometer, ~100×
thinner than c-Si solar cells. The films can be deposited on glass, or flexible and lightweight
substrates, increasing possible applications (such as mobile and wearable) compared with rigid
wafer-based cells. Direct deposition is less energy-consumptive and is suitable to produce large-
area cells. Thin film cells are relatively lower in efficiency (~21% in best lab-scale device)
compared to c-Si cells7,8. The scarcity of elemental components (indium and gallium in CIGS
cells; tellurium in CdTe) may also impose limits on cost and scalability.
The development of solution-processable PV materials is a major research frontier in emerging
PVs, such as dye-sensitized solar cells (DSSC, or “Grätzel cell”)9–11, organic and polymer solar
cells (OPV) 12,13, colloidal quantum dot (CQD) photovoltaics14–16, and solution-processed bulk
inorganic PV such as copper zinc tin sulfide (CZTS) 17–19. Though their present efficiencies are
relatively low (~10-12%) and the stability of the absorber is often too short from a commercial
perspective, there is a lot of research invested in these technologies due to their potential as low-
cost, high-efficiency alternatives to present-day commercial modules. The goal of lower module
cost and energy cost are pursued by constructing devices on large, flexible substrates using low
temperature solution-processing. This enables economical and scalable manufacturing
techniques such as spray-painting, ink-jet printing, and roll-to-roll printing. Additionally,
flexible, lightweight modules may lower the installation and maintenance costs. The challenge
now is to continue progress in efficiency and increase stability.
4
1.3 Theoretical and practical limits in photovoltaic efficiency
Achieving high efficiencies in photovoltaics requires an effective use of the broadband solar
spectrum (Figure 1-2)20. In the sun’s spectrum, nearly 50% of solar energy lies in the infrared
range. Natural photosynthesis uses part of the visible band, and its total solar energy conversion
efficiency is ~1%. Solar cells made using semiconductor junctions have the advantage of
broadband harvesting, spanning the visible and infrared, through bandgap engineering. A single-
junction solar cell made using a semiconductor with the optimal bandgap of ~1.34 eV has the
potential to reach a 33.7% solar-to-electricity conversion efficiency. This is known as the
Shockley-Queisser limit21.
Figure 1-2. The Sun’s power spectrum reaching the Earth and the limits of solar conversion. The AM1.5 solar
spectrum is broadband with distinct dips due to molecular absorption in the Earth’s atmosphere. The intensity of this
simulated solar spectrum is 1000 W m-2.20 Photosynthesis uses only the visible band, whereas solar cells can achieve
broadband absorption reaching out into the infrared region. Inset: photons with energy above the bandgap (Eg) are not
fully converted to electrical power due to the thermalization of charge carriers. The separation of quasi-Fermi levels
determines the open-circuit voltage Voc. Voltage loss relative to the bandgap Eg is caused by the spontaneous emission
in the requirement of thermodynamic detailed balance.
The Shockley-Queisser limit originates from a number of factors. Photons with energies below
the bandgap of the semiconductor are not absorbed, whereas photons of energy much above the
bandgap cannot be fully converted to electrical energy due to the thermalization of charge
400 800 1200 1600 2000 2400
1
2
Spectr
al pow
er
(W m
-2 n
m-1)
Wavelength (nm)
VB
CB
Eg qVoc
+
-
UV Visible Infrared
5
carriers (Figure 1-2). Another physical limit is imposed by the thermodynamic detailed balance,
which requires the solar cell to be in equilibrium with its environment. There is spontaneous light
emission in a solar cell when it absorbs light. The corresponding radiative carrier recombination
forms a dark current that reduces the open-circuit voltage (VOC) to be well below the
semiconductor’s bandgap (Eg) (Figure 1-2 inset).
Multi-junction solar cells represent a strategy to overcome the thermalization limit present in
single-junction cells. Absorbers with different bandgaps are stacked sequentially to extract power
from their respective portions of photons in the solar spectrum, which then leads to higher
overall solar conversion efficiency. For example, a triple-junction cell could increase the
efficiency limit from 33.7% to 49%22. Epitaxially-grown triple-junction devices have reached
efficiencies of ~44% in the lab with the aid of concentrated sunlight23,24. Multijunction cells’
costs are high due to the complex structures and the high prices of materials.
In practical solar cells, more sources of loss reduce the efficiency. The achievable VOC is below
the Shockley-Queisser limit due to dark current contributions caused by carrier recombination in
the bulk and at interfaces in a solar cell (Figure 1-3)25,26. The short-circuit current, Jsc (see
definition in Figure 1-4) is also below the theoretical value: not all incident light is absorbed in
the active layer due to optical losses such as reflection and parasitic absorption. Furthermore, not
all generated carriers are collected due to non-idealities such as interfaces and contacts in a cell.
Scientific and technological advances are needed for efficiencies to approach more closely the
Shockley-Queisser limit.
Figure 1-3. Photogenerated charge carrier transport in a single-junction solar cell. Photoelectrons generated in
the p-type material on the left diffuse through the quasi-neutral region and then are swept by the built-in electric field
Efp
Ec
Ev
Efn
Sunlight
Electricity
Depletion regionQuasi-neutral Quasi-neutral
E
Electrode
6
in the depletion region before they diffuse into the electron-collecting contact on the right. The dashed black arrow
represents charge carriers’ recombination pathway facilitated by trap states in the absorber and at the interfaces. The
corresponding dark current causes voltage and current losses in a practical solar cell. A key challenge is increasing
the diffusion length for minority carriers in solar cells. Ec and Ev indicate the band edges of the conduction band and
valence bands, respectively. Efp and Efn indicate the quasi-Fermi levels of holes and electrons respectively.
Figure 1-4. The current–voltage (J-V) characteristic and definition of important terms in photovoltaics.
1.4 Materials interface engineering for solution-processed photovoltaics
Advances toward increasing the efficiency of solution-processed solar cells rely largely on
controlling the abundant interfaces in devices, which, beyond the active layer, are primarily
composed of polycrystalline and amorphous layers25. Even though wafer-based crystalline Si PV
cells are considered to be planar and mature, the recent record performance (25.6%) was
achieved thanks to interface passivation of the crystalline Si (c-Si) surface using an ultrathin
layer of intrinsic hydrogenated amorphous Si (a-Si:H)5. In a classic solution-processed thin film
PV, multiple layers are stacked within a thickness of only a few hundred nanometers (nm), and
thus the conceptual difference between a “layer” and an “interface” is blurred. For clarity, the
interfaces in a device can be classified into two categories:
0
VOC
JSC
MPPJMPP
VMPP
Dark
Illuminated
Voltage
Cu
rren
t
Figures of merit in photovoltaics
Voc Open-circuit voltage. The voltage output by illuminated PV when its contacts are opened, i.e., there is no external load;
JSC The electron current flowing through the illuminated PV when its contacts are shorted;
MPP Maximum power point, where the product of current and voltage reaches its maximum.
FF Fill factor. The ratio of maximum power tothe product of VOC and JSC
PCE Power conversion efficiency. The ratio of power output at MPP to the solar power incident on PV device
EQE External quantum efficiency. The ratio of the number of electrons flowing through the PV per second under short-circuit condition to the number of photons illuminating the device each second
7
a) Charge-separating interfaces: Interfaces between photo-conversion materials, i.e., the
junction between intimately contacted photoelectron donors and acceptors, if applicable;
and interfaces between photo-conversion layers and top and bottom contacts.
b) Interfaces within absorbers, i.e., grain boundaries between the crystalline domains in
the absorber.
Engineering of these interfaces, including the control of morphology, electronic and chemical
interactions, is applied to improve device efficiency through two aspects: a) enhance the
absorption of light trapped in photo-absorbers; and b) reduce the recombination of photo-
generated carriers when they transit through the sequence of interfaces before they reach the
contacts to the external circuit.
1.4.1 Charge-separating interfaces
Solution-processed PV materials are typically polycrystalline and show limited diffusion lengths
(5-500 nm) for the performance-limiting photocarrier – the minority charge carrier. This
thickness is not enough to achieve complete absorption of solar irradiation in a planar absorber.
To break this absorption-extraction compromise, nanostructured or mesostructured interfaces are
employed and have been demonstrated in research on DSSCs (Figure 1-5a). In a DSSC, only a
small fraction (~1%) of sunlight is absorbed if a monolayer of dye molecules is anchored on a
planar electrode. The dye-electrode contact area can be increased more than a thousand times by
using nanostructured electrodes made of high-bandgap metal oxide nanoparticles, such as
titanium oxide (TiO2), tin oxide (SnO2), and zinc oxide (ZnO). The absorption and extraction can
exceed 80% across the entire visible range9–11,27. This concept of large interface areas is also
pursued in organic and polymer solar cells, where donor and acceptor molecules are mixed to
form a bulk heterojunction (BHJ), a 3D interpenetrating network in the photoactive layer (Figure
1-5b), that results in a high quantum yield of exciton dissociation12,13.
Materials chemistry and the electronics of donor-acceptor interfaces are also important. In
DSSCs, for example, back recombination of the separated carriers across the large-area
nanostructured interface becomes more likely and results in low open-circuit voltage. Forming a
dense self-assembled monolayer (SAM) and infilling co-adsorbents at the oxide interface
passivates the exposed surface defects and impedes the backflow of charges efficiently28. This
leads to near-unity net collection of photo-generated carriers.
8
Ultrathin (typically <10 nm) interlayers have been devised to improve the interface between the
photoactive layer and the contacts in organic and polymer PV. The photoactive layer’s surface
potential can be modified via an interface electrical dipole generated by the interaction between
the interlayer’s functional group and the photoactive layer29,30. The optimized energetic
alignment and built-in potential between the active layer and the contacts improves performance
by facilitating forward injection and suppressing back recombination.
Figure 1-5. Architectures and interfaces in solution-processed solar cells. (a) In dye-sensitized solar cells, a
monolayer of dye (orange dots) is anchored onto a 3D electrode such as mesoporous TiO2. (b) In organic solar cells,
the electron donor and acceptor materials are in intimate contact with each other, forming an interpenetrating bulk
heterojunction. (c) In colloidal quantum dot solar cells, the monodispersed nanoparticles (blue dots) are densely
packed to facilitate the charge carrier transport between quantum dots. (d) In inorganic bulk polycrystalline solar cells,
charge carriers diffuse through the grain boundaries in bulk absorbers to be extracted at top and down hetero-contacts.
For each cell, the panel on the left illustrates the device structure; the panel on the right illustrates the material energy
band diagram under open-circuit condition.
1.4.2 Interfaces within the absorbers: grain boundaries
Engineering the interfaces within the absorber is a large contributor to progress in colloidal
quantum dot (CQD) PV, where there exists grain boundaries throughout the film composed of
monodispersed nanocrystals (Figure 1-5c). Long organic ligands retain the colloidal stability of
quantum dots in solution, but isolate quantum dots from electronic communication with each
other. Replacing the long insulating ligands with short ligands (such as organic ligands with
TiO2
TiO2+Dye
Electrolyte
Pt
CQDs
AuTiO2
Au
ZnO
CdS
CIGS
TiO2
PCDTBT
PCBM
Al
a c
b d
9
short anion- and thiol-end functional groups, or halide atomic ligands), a process termed ligand-
exchange, leads to denser packing of quantum dots without dot fusion, thus maintaining quantum
confinement, leading to remarkably improved carrier transport throughout the film. The short
ligand chosen also plays a crucial role in passivating trap states on quantum dot surfaces, thereby
providing a clean bandgap in quantum dot solid films that prolongs the lifetime of charge carriers
to be collected before they can recombine16,31.
Grain boundary passivation in CIGS or CZTS polycrystalline thin films (Figure 1-5d) is also
applied to improve performance32,33. This was implemented by controlling the band offset
between grain boundaries and the bulk. Materials processing such as elemental stoichiometry
tuning and post-annealing are also found to modify the grain structure, grain orientation, and
grain boundaries’ composition, providing a handle to reduce the surface trap density.
1.5 Roadmap of thesis: engineering interfaces in perovskite photovoltaics
Among emerging solar photovoltaics, solution-processed perovskite solar cells show particular
promise to fulfill the simultaneous goals of high efficiency and low cost. As introduced in
Chapter 2, organic-inorganic hybrid perovskite materials can be applied with simple solution-
processing at low temperature. Device efficiencies have increased from 3.8% in 2009 to ~18% in
early 201434,35, and progress shows no signs of abating. However, the materials and devices are
relatively new and far from being well understood. The problems of hysteretic behavior and
operation instability must be overcome in order to further push the efficiency toward its
theoretical limit while maintaining low cost. The photochemical and photophysical pathways of
these new materials need to be revealed. Advances will rely on understanding and managing the
abundant materials interfaces that make up these highly polycrystalline devices – the major focus
of this thesis (Figure 1-6).
An outline of the flow of this thesis is shown in Figure 1-6. The effects of material processing on
perovskite grain crystals and trap state formation are studied theoretically and experimentally
(Chapter 3) and provide guidance on precursor optimization and film processing to form device-
level films. The materials chemistry and resultant interface electronic engineering are then
applied to the grain boundaries, electron-extraction interfaces (Chapter 4) and hole-extraction
interfaces (Chapter 5) of perovskite films to improve the overall device performance
10
progressively. Density functional theory (DFT) and nano-probing technologies were employed
for accurate simulation and direct observation of interface processes at atomic and nanometer
scales. This provided guidance to understand and optimize perovskite interfaces.
Figure 1-6. Outline of thesis: material interfaces engineering in perovskite PV.
Chapter 3Trap state reduction via
material processing
Chapter 4Electron-extraction
interfaces and grain boundaries
Chapter 5Hole-extraction
interfacesand device top surface
Materials interface engineering on planar perovskite solar cells to improve
hysteresis, stability and efficiency
11
Chapter 2
Perovskite photovoltaics
The emerging field of perovskite solar cells has captured the attention of the photovoltaics
research community, with the record efficiency exceeding 18% by 2014. More importantly, this
high efficiency is achievable with materials produced from high-throughput and inexpensive
solution processes. With the potential to achieve even higher efficiency and lower cost,
perovskite solar cells are commercially attractive. However, the origin of the superb performance
is still not fully understood. The performance anomalies, such as hysteretic behavior and
unstable operation, need to be addressed in order to fulfill the theoretical maximum efficiency.
These challenges motivated me to research the abundant material interfaces that make up
perovskite photovoltaic devices.
2.1 Crystal structure and charge transport
The term “perovskite” refers to the class of compounds that have the same type of crystal
structure as CaTiO3. These materials have the general ABX3 stoichiometry, consisting of corner
sharing octahedral BX6 in three dimensions (3D), with cation A occupying the cuboctahedral
cavity in each unit cell (Figure 2-1). The structure provides many degrees of freedom for the
elemental composition as long as charge balance and atomic radius tolerance conditions are
satisfied36,37. The A sites function as “spacers” between the BX6 network and provide a freedom
to diversify the perovskite structure from 3D frameworks to 2D layers, 1D chains, and 0D
frameworks37–40. The perovskite structure and its distortions provide a materials platform with
finely tunable physical properties.
12
Figure 2-1. Crystal structure of perovskites with the generic chemical formula ABX3. Organic or inorganic
cations occupy position A (blue) whereas metal cations and anions occupy the B (grey) and X (purple) positions,
respectively.
Within the family of perovskites, the organo-metal halide perovskites, especially the class of
methylammonium lead halide perovskites (MAPbX3, where MA indicates methylammonium,
CH3NH3; and X the halide, typically I, often with a small fraction of Cl or Br), are attractive as
solar energy harvesters due to efficient ambipolar transport, strong light absorption, ease of
solution-processed manufacturing, and widely tunable optical properties41–44.
Depending on the halide content, the bandgap of mixed halide perovskites (MAPb(I, Br, Cl)3)
can be tuned from ~1.6 eV (pure MAPbI3) to 3.2 eV (MAPbCl3). Smaller bandgaps can be
achieved using appropriate organic cations (e.g., formamidinium, CH(NH2)2) and inorganic
cations (e.g., Sn ). A smaller bandgap, toward 1.3 eV, is desirable because of the higher
Shockley-Queisser efficiency limit for single-junction devices.
The methylammonium lead halide perovskites have long diffusion lengths over 1 μm for both
holes and electrons, as measured in single crystal studies45,46. Although the quality of perovskite
polycrystalline films is highly dependent on the materials processing conditions, it is widely
agreed that a diffusion length much greater than 100 nm can be achieved via appropriate film
deposition47–49. Combined with a high absorption coefficient, the long diffusion length in
perovskites, dwarfing other solution-processed PV materials, enables a break of the “absorption-
extraction compromise” to construct highly efficient planar solar cells.
13
Experimental reports also show that the photo-generated carriers in perovskites are primarily
present as free electrons and holes, rather than as excitons as in other solution-processed PV
materials (e.g., organic solar cells, colloidal quantum dot solar cells). The high dielectric constant
and low exciton binding energy are inherent in halide perovskites, and enable charge separation
at room temperature50,51.
Although not fully understood yet, these outstanding charge transport characteristics in solution-
processed perovskites, causing them to surpass most of their “low-cost” competitors, underpin
the success in PV applications.
2.2 Device architectures
Halide perovskite PVs were introduced in 2009 by Tsutomu Miyasaka34. Initial progress in
perovskite photovoltaics has benefitted from the pioneering works of Nam-Gyu Park, Michael
Graetzel and Henry Snaith on solid-state perovskite-sensitized solar cells in 201252,53.
Mesoporous scaffolds were used to minimize the limits imposed by minority carrier drift and
diffusion via inclusion of the active light absorber within nanometer-sized electron-harvesting
pores (Figure 2-2a). The structures were very similar to those of dye-sensitized solar cells
(DSSC). A worldwide effort to improve device performance with such architectures led to rapid
efficiency improvements in perovskite solar cells (from 3.8% in 2009 to ~18% in early 2014)
34,35,54, rapidly approaching the performance of commercial-grade silicon-based photovoltaic
modules.
Figure 2-2. Perovskite solar cell architectures evolution. (a) Cross-sectional structure of a solid-state perovskite-
sensitized solar cell where the thick mesoscopic metal oxide (TiO2 or Al2O3) scaffold is infiltrated by the perovskite.
(b) Cross-section of a planar heterojunction solar cell lacking the thick metal oxide mesoporous scaffold.
Electron transport contact
Transparent conductive oxide
Hole transport contact
Back contact
Glass substrate (front contact)
Mesoporous scaffold
Perovskite
a
Electron transport contact
Perovskite
Transparent conductive oxide
Hole transport contact
Back contact
Glass substrate (front contact)
b
14
Planar-electrode (as distinct from mesoporous) devices (Figure 2-2b) have also attracted
extensive studies, based on the finding that perovskites act as efficient ambipolar charge-
conductors55,56. Compared with the mesoporous architecture, planar devices are particularly
important in certain realms of application, such as in photodetector arrays (which demand
stringent spatial uniformity pixel-to-pixel), lasers (which require planarity for minimized
scattering while waveguiding), and flexible photovoltaics (which strive to avoid high-
temperature mesoporous oxide processing)57,58.
2.3 Challenges in perovskite photovoltaics
2.3.1 Stability
A major problem in perovskites is the instability of the solar cell, both in materials and devices.
For the perovskite material, the organic cations are vulnerable to moisture and heating cycles,
and easily desorb from the lattice of the perovskite, leaving the inorganic framework of PbI2
(Figure 2-3a)59. It has also been suggested that the lead-halide bond is inherently not photo-
stable. In devices, UV-induced degradation is also linked to the redox photocatalysis effect at the
TiO2 surface and thus is more significant in mesoporous devices60,61. Such detrimental interfacial
interactions have also been shown at the perovskite-hole transfer interface62. The hole transfer
layer’s sensitivity to external variables is also an origin of device instability63,64.
Figure 2-3. Instability and hysteresis in perovskite solar devices. (a) The visible degradation from CH3NH3PbI3
perovskite (dark brown) to PbI2 (yellow) phase in air due to moisture. (b) Hysteresis of perovskite solar cells shown
in their J-V solar response.
Time in air
(a)
Voltage (V)
Cu
rren
t d
ensi
ty (
mA
cm
-2)
0.5 10.0
10
20
0
(b)
15
2.3.2 Hysteresis
Severe hysteresis represents another big challenge for perovskites (Figure 2-3b). This is seen in a
scan-direction- and scan-speed-dependence to photo J-V characteristics65–67. This behavior
makes overestimating the solar-to-electricity efficiency likely. Moreover, the devices with
hysteresis also exhibit current degradation when they are operated under the steady-state
condition. The hysteresis is likely associated with charge recombination at defective interfaces
and grain boundaries48,68–72. The origins of hysteresis in devices are multifold and are far from
well resolved.
2.4 Research goals and methodology
My research vision is geared toward the exploration of novel materials and device designs to
develop hysteresis-free, stable, perovskite solar cells. Studies would include (1) origins of
hysteresis and instability and remedies thereto and (2) ways of making the materials and devices
more stable and hysteresis-free.
Intriguingly, the enhanced diffusion lengths in perovskite single crystals largely result from the
absence of structural and grain boundary defects that are present in perovskite polycrystalline
films73, suggestive of strategies of material processing and interface engineering to improve the
perovskite-film based PV devices.
Bearing this consideration in mind, I start with understanding the trap physics and formation
chemistry of perovskite polycrystalline films during materials processing via DFT simulation in
Chapter 3. Based on this knowledge, I further devise the materials engineering of the full set of
interfaces (grain boundaries, electron-extraction interfaces, and hole-extraction interfaces) in
perovskite devices (Figure 2-4) to improve the hysteresis and stability characteristics in order to
ultimately improve device efficiency (Chapter 4 - 7).
16
Figure 2-4. The framework of full interfaces (green) in the perovskite solar cells. Full interfaces include the grain
boundaries within perovskite polycrystalline films and the electron- and hole-extraction interfaces bridging the
perovskite and the charge-accepting contacts.
Hole-accepting contact (Au/Ag, top reflecting contact)
Electron-accepting contact (ITO/FTO on glass, transparent)
Perovskite grains Grain
boundaries
Sunlight
Electron-extraction interface
Hole-extraction interface
17
Chapter 3
Materials processing for low-trap-density perovskite solids
3.1 Introduction
Forming a low-trap-density film with good morphology is the starting point for making
perovskite PV devices and for further interfaces engineering. Recent reports have consistently
emphasized that the specific choice of growth conditions and chemical precursors is central to
achieving high-quality films. For example, MAPbI3 films made from a mixture of lead-chloride
(PbCl2) and methylammonium-iodide (MAI) exhibit more impressive diffusion lengths (1 µm)
than those made from single-halide approaches. However, the roles and mechanisms are poorly
understood. In this chapter, I use density functional theory (DFT) to explore the electronic levels
and formation energies of trap and defect states as the perovskite MAPbI3 film is grown. This
study provides a framework for understanding the impact of composition on the band-structure
of crystals incorporating vacancies, dopants, and interfaces. More importantly, it answers the
practical questions of identifying the dominant electronic traps formed during perovskite film
growth from solution, and how to choose the growth conditions and chemical precursors to form
PV films with reduced trap states.
3.2 Perspective from DFT on perovskite growth
DFT calculations were performed on tetragonal MA–PbI3 by using a supercell including 1728
atoms, which can accommodate various types of lattice and substitutional sparse defects. In line
with previous reports74–76, the electronic structure is confirmed to be direct at the Γ-point, with a
valence band maximum (VBM) state in an antibonding combination of I 5p (dominant character)
and Pb 6s orbitals (Pb 6s–I 5p, σ*), while the conduction band minimum (CBM) has mainly Pb
6p character.
The first part of the study aims at exploring the impact of the nanocrystalline morphology on the
band structure of MA–PbI3. In particular, one of the remarkable features of these materials is
that, in spite of their polycrystallinity, they have exceedingly sharp bandedges44. Electronic
18
structure calculations on slabs of perovskite crystals were carried out to evaluate the impact of
surfaces on their band structure. The corresponding density of states (Figure 3-1a) shows no
states in the gap either for bulk or surface electronic structures. Indeed the highest-lying valence
band states in asymmetrically terminated stoichiometric slabs (Figure 3-1b) exhibit a high degree
of delocalization, as do the lowest-lying conduction band states (Figure 3-1c). A crucial insight
into the origin of this behavior comes from the computed surface energies: very low values of
∼10 meV Å-2 are obtained. The low surface energy indicates high stability, obviating the need
for a reconstruction of the (001) terminated surface and also avoiding the need to add adsorbates
(e.g., introduce new ligands) in order to remain inert. It was shown recently77 that the surface
energy of terminated (001) PbI2 slabs is indeed small and compares quite well to the value of
surface energy obtained in this study.
Figure 3-1. Surface states of the Pb halide perovskites. (a) The density of states for a 16-monolayer slab (surface)
compared to the case of bulk materials. The inset reveals that no electronic states emerge in the slab compared to the
bulk. The wave functions of surface states at (b) the valence band maximum (VBM) remain highly delocalized in the
case of 16-monolayer slabs, similarly to those at (c) the conduction band minimum (CBM) states. The highest
amplitudes of positive- (negative-) valued wave functions are indicated in yellow (blue). Please note that the CBM
and VBM states are separated due to asymmetric termination.
Looking at defects first required an estimation of the formation energy of the perovskite relative
to its decomposition into pure PbI2 and MAI phases (Figure 3-2a). The calculations show that the
formation energy of the perovskite relative to its decomposition into pure PbI2 and MAI phases
19
has a low value of about −0.1 eV. This indicates that the material and its precursors are
energetically close to phase coexistence of MAI and PbI2 which is consistent with recent
experimental findings54,78 of a residual PbI2 phase remaining even after long annealing times. It
is also consistent with reports60,79–82 that the details of preparation conditions are important to
achieve the best-performing materials.
Figure 3-2. Tetragonal perovskite, its formation from experimentally employed precursors, and its defect
energy levels. (a) The energy of formation of the CH3NH3PbI3 perovskite (room temperature tetragonal structure is
shown) from its bulk precursors CH3NH3I and PbI2, with a kinetic barrier (transition state energy) also depicted.
Calculations discussed in the text report a −0.1 eV difference between the perovskite and its precursors, consistent
with experimental studies that show the presence of a secondary PbI2 phase. (b) Energy levels associated with the
defect states corresponding to neutral and charged vacancies (VPb, VI, VMA), neutral and charged interstitials (Pbi, Ii,
MAi), and neutral and charged states associated with antisites (PbI and IPb).
The electronic levels (Figure 3-2b) and formation energies (Hf) of various classes of defects were
then analyzed. The defects include vacancies (VPb, VMA, VI), interstitials (Pbi, MAi, Ii), and
antisites (PbI, IPb), where in the latter case AB indicates that A is substituted by B. The value
20
of Hf was used to estimate the density of the relevant vacancy species, which will be proportional
to exp(−Hf/kT). One can see from Figure 3-3a and b that the major acceptor defects are VPb–2,
VMA−1 and Ii
–1, while donor defects are VI+1 and Pbi
+2, with an associated charge density
sufficient to induce doping anywhere from p- to n-type, depending on the chemical potentials.
Defects VI+1, VMA
–1, and MAi+1 possess the lowest formation energies over the entire bandgap.
The transition levels of VPb–1 and Ii
–1 are located on top of the VBM, indicating that negative
charge states are stable over the entire bandgap.
The defect formation energies of neutral defects are given in Table 3-1. One can also see that Pbi,
PbI, and IPb represent the “negative-U” defects83,84, indicating that these defects are not stable at
+1, −1, and +2 charge states, respectively. The large structural relaxation in the case of
PbI0 (interplanar I–I–I bridging) is consistent with such “negative-U” behavior of the charge
neutral PbI antisite. Figure 3-3a and b also show defect states with negative formation energies at
various Fermi levels. Usually, this implies that at these chemical potentials and Fermi levels it is
not possible to grow the crystal, and Fermi level pinning happens in such a way that all of the
defect energies are positive85 while satisfying the overall charge-neutrality condition; in other
words, the structure is not stable for such parameters86,87. However, these are values of the single
defect formation energies, and nothing prevents the complexation of single defects whenever the
latter is favorable in energy. Thus, defect clustering expands the range of chemical potentials and
Fermi levels, i.e., growth conditions.
So far, the complex defect formation has been reported in the case of VPbI2 alone88 and shown
recently to have low formation energies89. All vacancies produce either slightly perturbed states
in the bands (which do not capture carriers), or shallow traps and resonances (deep localized
states hybridized with conduction or valence band states90 within the band), implying that
carriers can still relax easily to VBM and CBM. In contrast, certain interstitials and antisites
associated with Pb and I form electronic states deep inside the bandgap. The most important
figure of merit is the density of deep traps in the semiconductor volume, which will determine
the rate of capture of charge carriers and of loss due to recombination. In semiconductors having
good electronic transport parameters, it is indeed the density of recombination centers that
directly controls the diffusion length of charge carriers91.
21
Table 3-1. Computed defect formation energies under (a) I-poor (Pb-rich) (b) I-rich (Pb-poor) conditions.
Further analysis of the tetragonal vs. cubic crystal phase leads to a small number of qualitatively
and quantitatively different conclusions with respect to defect state energies. Overall the
agreement with previously published results79 is good, except that it was previously found that
the PbI neutral antisite defect has a high formation energy in the cubic case. In the tetragonal
phase, the PbI neutral antisite defects possess a much lower formation energy. Physically, this
difference arises because in the tetragonal phase two distinct PbI2 layers are found (Figure 3-4): a
nearly planar PbI2 plane and also a bulged PbI2 plane. The methylammonium cation has been
shown experimentally to exhibit a small reorientational barrier92 and to possess a small energetic
difference between two perovskite structures with different methylammonium cation
orientations, thus affecting the structure of the PbI2 layers.
The formation energy for charged defects, and consequently the volume density of the various
classes of localized electronic states, is a function of the Fermi level in the semiconductor. The
former is determined by whether the crystal is grown from iodine-poor or iodine-rich conditions,
which constitutes the central link between the growth chemistry and the charge transport
performance of these materials. In a crystal grown under I-rich conditions (Figure 3-3a), the
PbI antisite (Pb atom replaced by I) deep trap has a formation energy of less than 0.2 eV for all
choices of Fermi level between the valence and conduction bandedges, i.e., for all nondegenerate
doping conditions. As a prediction, a perovskite grown under I-rich conditions will show a high
density of deep traps that will curtail the diffusion length. In contrast, in a crystal grown under I-
Defect Ef, (eV) (a) Ef, (eV) (b)
2.69 0.47
0.87 2.04
0.85 2.05
1.91 0.62
1.12 2.43
1.05 3.27
1.75 0.57
3.6(4.22) 0.19 (0.9)
1.54 4.88
22
poor conditions (Figure 3-3b), there exists a window of Fermi levels, EF ≥ 0.9 eV (measured
relative to the VBM), where all deep traps have formation energies that exceed 0.38 eV. This
places the equilibrium density of trap states below 1015 cm–3. At this volume density, traps are
spaced a median ∼200 lattice constants in all crystallographic directions, enabling diffusion
lengths above 100 nm. Such low trap densities are also consistent with the impressive open-
circuit voltages seen in the best-reported lead iodide perovskite devices.
Figure 3-3. Formation energies and volume densities of key defects in tetragonal lead perovskites. Defect
formation energies for iodine-poor (a) and iodine-rich (b) growth conditions. Continuous (dashed) lines denote
shallow (deep) traps (see legend). The red region in (a) indicates the range of Fermi energies where trap densities
exceed 1018 cm–3—showing that no Fermi level choice yields a semiconductor substantially free of deep traps in the
case of I-rich growth conditions. Conversely, in the case of I-poor growth (b), trap densities below 1015 cm–3 can be
achieved in the range of Fermi energies (green region). In the case of MA-related defects one has (a) MA-poor and
(b) MA-rich conditions.
3.3 Growth conditions control for low-trap perovskite film
From a chemical perspective, the presence in solution of simple ions (Pb2+ and I–) combined with
that of lead-iodide complex anions such as PbI3–,93 PbI4
2–, and PbI53– produces a motif similar to
the PbI0 neutral antisite defect which corresponds to bridging between three iodine anions in-
plane (Figure 3-4b) and interplane (Figure 3-4c) in the perovskite lattice. The motif is produced
indirectly, by violating local stoichiometry. The I–I–I angle is 172°, and the I···I bond length
between two iodines is 2.98 Å. Knowledge of this geometry allowed me to look for signatures of
this complex, as did the signature wave function pattern (I 5p–I 5p, σ*). The concentration of
PbI3– in solution is higher for PbI2 + MAI than for PbCl2 + MAI. One interesting direction these
23
results suggest is that alternative Pb non-iodine-containing precursors, such as Pb(SCN)2,
Pb(CH3CH2)4, and Pb(C2H3O2)2 (dehydrated lead acetate, Pb(Ac)2), can provide another avenue
to reaching the I-poor conditions required for high-quality perovskites.
Figure 3-4. Physical configuration of PbI neutral antisites in the tetragonal phase. (a) Two different PbI2 planes
in the tetragonal phase. One of them is bulged. Upon structural relaxation of the PbI0 defect, two conformations are
possible: (b) in-plane bridging or (c) interplanar bridging. The bridge configuration has a motif very similar to the
I3– triiodide complex, with an I–I–I angle of 172° and bond length 2.98 Å. This allowed association of the PbI
0 neutral
antisite with macro-ion complexes such as PbI3– and PbI4
2– in solution-phase precursors, and these are expected to be
present in growth under iodide-rich conditions.
3.3.1 Diffusion length
As a way to validate this prediction, and in order to challenge experimentally the proposed
theoretical findings, I investigated the diffusion length in perovskite films formed using a
dehydrated lead acetate precursor (see methods in Appendix 1.1-1.3) in place of the typical
PbI2 or PbCl2. The use of this precursor satisfies the requirement for I-poor conditions while
providing a Cl-free platform in order to rule out any possible effect from residual amounts of
incorporated chlorine. Similarly to the mixed-halide case, I find that acetate does not incorporate
24
into the final product. In order to probe the diffusion lengths for MA–PbI3 grown from Pb(Ac)2, I
excited the perovskite film using a 442 nm laser while measuring the photoluminescence (PL)
signal in a reporter layer of small bandgap quantum dots (see inset of Figure 3-5a). The quantum
dot PL intensity is directly controlled by the carrier diffusion length and by the thickness of the
perovskite film94.
Qualitatively these diffusion length experiments can be understood as follows: for thin
perovskite films with thickness d smaller than the carrier diffusion length LD, the amplitude of
the PL signal correlates positively with LD, because an increase in layer thickness results in a
higher absorption and hence a higher exciton generation rate, and at the same time almost all
excitons can diffuse through the perovskite and reach the reporter layer. For increasing thickness,
the generation rate reaches saturation because the absorption cannot exceed 100%, but fewer
excitons can now diffuse to the reporter layer. The PL profiles for different perovskite film
thicknesses are presented in Figure 3-5a. The overall behavior of PL intensity vs. film thickness
(Figure 3-5b) can be modeled using the following function:
𝑷𝑳(𝒅; 𝒂, 𝑳𝑫) ∝𝑳𝑫𝟐
𝟏−(𝒂𝑳𝑫)𝟐[𝒂𝒆−𝒂𝒅 +
(𝒆−𝒂𝒅
𝑳𝑫) 𝐬𝐢𝐧𝐡(
𝒅
𝑳𝑫)−𝒂
𝐜𝐨𝐬𝐡(𝒅/𝑳𝑫)] (3-1)
Here LD is extracted by fitting the experimental data of Figure 3-5b once the independently
measured absorption coefficient α is known (see inset). The diffusion length of the
Pb(Ac)2 based perovskite films was determined to be ~600 nm, thus considerably larger than
∼200 nm (even lower values of ∼100 nm have been reported in literature47) for films formed
using the PbI2 precursor. This suggests that chlorine itself does not play a central role in large
diffusion lengths, whereas the reduction of iodine content during the film formation process is
likely key to the remarkable charge transport in these materials.
25
Figure 3-5. Experimental investigation of transport in novel iodide-poor perovskite films. (a) Measured
photoluminescence (PL) spectra for various perovskite layer thicknesses as indicated in the legend in the case of
Pb(Ac)2 precursor. The inset shows a schematic of the diffusion length measurement method: the sample is illuminated
at a wavelength (442 nm) that is strongly absorbed in the perovskite layer; photogenerated charge carriers diffuse to
the quantum dot reporter layer where they recombine, providing a spectrally distinct signature of their arrival. (b)
Experimental plot (black and red circles) of the corrected PL peak amplitude vs. perovskite layer thickness both for
the PbI2 and Pb(Ac)2 precursors. A fit based on Equation (3-1) allows estimating the diffusion length to be LD ≈ 600
nm and LD ≈ 200 nm, which is comparable to prior pure and mixed-halide films. Inset: measured absorption coefficient
of the perovskite absorber, where the excitation wavelength is highlighted.
3.3.2 Crystalline morphology
Recent reports have consistently emphasized that in planar devices without a mesoporous metal
oxide scaffold, it is challenging to form continuous and smooth perovskite films via solution
processing41,42,54,78,95. The resultant shunting pathways facilitate carrier recombination and limit
the device performance. Our observations in films grown directly from regular PbI2 protocols are
aligned with these reports. A non-continuous film is usually obtained that exhibits a high
percentage of porosity (Figure 3-6).
26
Figure 3-6. Non-continuous perovskite films grown using PbI2 precursor. (a) Top view of a rough perovskite film
with a high percentage of porosity. (b) and (c) Cross-sectional images show grains in films with different thickness.
The films grown from the new protocol using anhydrate lead acetate (see methods in Appendix
1.1-1.3) exhibited different crystalline behavior. The resultant films were dense and smooth
(Figure 3-7). The film roughness was smaller than 10 nm when the film thickness was bigger
than 200 nm (inset of Figure 3-7a). The pin-hole-free characteristic was maintained even when
the film thickness was reduced down to ~100 nm (Figure 3-7c). Smooth and dense films
spanning a large range of thicknesses are desirable for planar optoelectronics, such as solar cells
and light emitting diodes (LED).
Figure 3-7. Dense and smooth perovskite planar films grown using anhydrate lead acetate. (a) A dense and
smooth perovskite film with thickness > 200 nm and roughness ~8 nm as characterized by AFM (inset). (b) The
zoomed in image of the film morphology. (c) The cross-sectional SEM of a continuous planar film with thickness
~100 nm, deposited on a rough FTO substrate.
27
This work shows that by controlling the perovskite growth conditions, such as via anion
engineering, we can tune the crystalline kinetics, film formation, and therefore device
performance. My work of using anhydrate lead acetate (Pb(C2H3O2)2 ) to resolve the morphology
problem in planar films was supplemented by parallel efforts by my peers, where hydrate lead
acetate (Pb(CH3COO)2·3H2O) was used96.
This study clarified fundamental aspects underlying the impact of growth conditions on the
performance of MA–PbI3 films. It revealed delocalization of the electronic states within the local
nanocrystal surfaces that preserved the integrity of the bulk bandgap. The DFT-based analysis of
defect formation energies provided an explanation for the observation of a larger charge
diffusion length in perovskites prepared using iodide-free precursors78,97–99compared to MAI +
PbI2 growth conditions.
3.4 Conclusions
In summary, iodine-rich growth conditions were found to be prone to forming perovskites with a
high density of deep trap states (recombination centers). The use of mix-halide precursors help to
reduce the formation of key defects (Pb atom substituted by I). I-poor conditions were
constructed using an anhydrate lead acetate (Pb(C2H3O2)2) precursor and showed films with
better diffusion lengths and morphologies than for the single-halide method which is I-rich,
under the same deposition conditions. This is an important aspect when considering the control
over growth conditions for making efficient perovskite devices.
28
Chapter 4
Fullerene-perovskite interaction at electron-extraction interfaces
4.1 Introduction
In Chapter 3, I identified via DFT simulation that I-rich conditions might be a dominant source
of deep trap states during perovskite growth from precursor solutions. I found that using lead
acetate (Pb(Ac)2) as the lead precursor would help construct I-poor conditions and form films
with lower traps and longer diffusion lengths. I engineered films based on these findings and
constructed purely planar perovskite PV devices. I focused on planar devices for two reasons.
First, there is no fundamental limit on perovskites for building planar devices without
mesoporous scaffolds considering the ambipolar transport and large values of diffusion lengths
in the perovskite material. Planar architectures also help minimize surface recombination and
increase the open-circuit voltage. Second, solution-processed planar perovskite devices are
highly desirable in a wide variety of optoelectronic applications such as in photodetector arrays
(which require stringent spatial uniformity pixel-to-pixel), lasers (which require planarity for
low-scatter waveguiding), and flexible photovoltaics (which strive to avoid high-temperature
mesoporous oxide processing)55–58,95,100.
I found, however, perovskite films on planar devices are prone to hysteresis and current
instabilities, which is highly consistent with recent reports in literature (details see Chapter
2.3.2). The performance of planar perovskite devices has, widely agreed to date, suffered from
two potentially interrelated concerns: severe hysteresis41–43; and, relatedly, recombination, likely
associated with defective grain boundaries induced by excess halides48,68–70,101. A dependence of
hysteresis on device architectures has also been observed, where inverted structures have
typically shown less serious hysteresis than regular planar devices, but have a lower open-circuit
voltage71,72.There has been, to date, no consensus as to the origins of these findings.
In this chapter, I pursue a solution-phase in situ passivation strategy with the goal of enabling
simple low-temperature materials processing and efficient passivation throughout the grain
boundaries in the perovskite active layer and electron-extraction interfaces. I also employ density
29
functional theory (DFT) and nano-probing techniques to reveal the underlying sources of
hysteresis.
4.2 Improvement of hysteresis and photovoltaic performance
In the course of device studies of mixed materials made from solutions containing both
perovskites and the electron-acceptor PCBM (a derivative of C60, phenyl-C61-butyric acid
methyl ester), I observed an enhancement in photovoltaic performance (Figure 4-1a-c) and a
reduction in hysteresis (Figure 4-1d-e) relative to control devices based on perovskites alone, and
also compared to separate-layer PCBM-perovskite devices (Table 4-1, see Figure 4-2 for
different device structures). To create the mixed-material films, I dispersed PCBM and lead
acetate (Pb(Ac)2)48 in various ratios and formed films using a one-step deposition
process53,100,102,103 employing MAI as the organohalide precursor. As well as observing reduced
hysteresis (Figure 4-1d-e), I observed in the perovskite-PCBM mixed-material device a
substantial voltage enhancement (~ 0.1 V) (Figure 4-1a) and a higher fill factor compared to the
PCBM-free and bilayer PCBM-perovskite controls (see methods in Appendix 1.1-1.3).
Figure 4-1. Steady-state photovoltaic performance of an ultra-thin perovskite-PCBM hybrid film. (a) The steady
state open circuit voltage, VOC, (b) steady state short circuit current density, JSC, and (c) the steady state power
0.0 0.2 0.4 0.6 0.8 1.0 1.2-10
0
10
20Control
FF(Forward)=0.66
FF(Reverse)=0.42
Me
asu
red
cu
rre
nt
de
nsity (
mA
cm
-2)
Voltage (V)
0.0 0.2 0.4 0.6 0.8 1.0 1.2-10
0
10
20Hybrid
FF (Forward)=0.74
FF (Reverse)=0.70
Me
asu
red
cu
rre
nt
de
nsity (
mA
/cm
2)
Voltage (V)
(d) (e)
Maximum Power Point
0 20 404
6
8
10
12
14
16
Hybrid
Control
Ste
ad
y-s
tate
Eff
icie
ncy,
PC
E(%
)
Time, t(s)
(a) (b) (c)
0 5 10 15
0.9
1.0
1.1
1.2
Hybrid
Control
Op
en
Cir
cu
it V
olta
ge
, V
OC
(V)
Time, t(s)0 2 4 6
12
13
14
15
16
Hybrid
Control
Sh
ort
Cir
cu
it C
urr
en
t, J
SC
(mA
cm
-2)
Time, t(s)
Absorber thickness:150±10 nm
500nm
400 600 800
0
20
40
60
80
100
Measure
d E
QE
(%
)
Wavelength (nm)
Calculated Jsc=15.4mA cm-2
30
conversion efficiency, PCE, of the perovskite-PCBM hybrid film (red) compared with the control perovskite only film
(blue). During steady-state measurement, the integrating time for each point is 0.35 second. (d) The instantaneous J-
V curve of the control device (perovskite film) with high hysteresis. The thicker curve indicates forward scan starting
from open circuit conditions; thin curve is the reverse scan from short circuit conditions. The scanning rate is 0.2 V s-
1. The fill factor (FF) of the forward scan is 66% while the reverse FF is reduced to 42%. The black point indicates
the ‘maximum power output point (MPP)’ is measured from the steady state PCE as shown in (c). The MPP here is
located between two instantaneous J-V curves due to the significant hysteresis and current decay. (e) The J-V scan of
a hybrid device shows very low hysteresis and low current loss, as shown in (b). The FF for forward (reverse) scan is
74% (70%). The steady-state MPP is consistent with the forward J-V curve, which demonstrates the stability of the
hybrid film. The inset of figure (e) shows the external quantum efficiency (EQE) of a hybrid device. The current
density predicted from the EQE is 15.4 mA cm-2, consistent with the steady state current density measured in (b). The
inset figure of (d) shows the thickness of the active layer in both devices is around 150 nm.
Table 4-1. Statistics of steady-state performance with different PCBM distribution and thickness.
Device configuration Steady-state performance Instantaneous
Type Thickness (nm)
VOC (V)
JSC (mA cm
-2)
PCE (%)
FF (%) Forward/Reverse
Control
150±20
0.97±0.02 13.1±0.8 6.7±0.5 62/38±3
Bilayer 1.08±0.02 14.2±0.4 10.6±0.4 71/64±3
Hybrid 1.09±0.02 14.4±0.4 10.9±0.4 72/65±2
Hybrid champion 1.11 14.6 11.9 73/68
Control 300±20
0.98±0.02 14.4±0.8 8.1±0.5 65/40±3
Bilayer 1.06±0.02 16.1±0.4 12.0±0.5 72/56±3
Hybrid 1.07±0.02 17.3±0.4 13.6±0.6 73/66±3
Hybrid champion 1.086 18.0 14.4 75/69
Statistics for each case are based on 20 devices prepared on separate substrates.
31
Figure 4-2. Solution-processed planar device structures in this study. (a) Control device with PCBM-free pure
perovskite (CH3NH3PbI3) as the active layer; TiO2 and Spiro-OMeTAD as the electron transport layer (ETL) and the
hole transport layer (HTL), respectively.(b) Perovskite-PCBM bilayer structure with PCBM cast on TiO2 before
perovskite deposition; and (c) Perovskite-PCBM hybrid device with mixed material as the active layer.
4.3 Mechanistic studies of perovskite-PCBM interaction
4.3.1 Material characterization
I proceeded to seek mechanistic insights regarding the role, or roles, of the PCBM. Specifically, I
asked whether the PCBM could interact with certain chemical species in the mixed-material
solution; and whether studies of incorporation into films using the new process indicated a
homogeneous distribution of PCBM throughout the active layer, compared to segregation into a
bilayer device with PCBM either substantially below or above the perovskite.
Solution-phase spectroscopy provides one means to study the formation of complexes of PCBM
with the various perovskite solution-phase precursors. When PCBM is mixed into the normal
perovskite precursor solution, the bright yellow solution (Figure 4-3a left) turns to dark brown
(Figure 4-3a right). The absorption spectrum of perovskite-PCBM hybrid solution shows a peak
at 1020 nm (Figure 4-3b). This is in contrast with pure PCBM in the same solvent, which is
observed to be transparent in this wavelength region. The 1020 nm spectral feature is associated
in literature reports with the formation of a PCBM halide radical (Figure 4-3b inset)104–106.
FTO/Glasss
Perovskite-PCBMHybrid solid
Spiro-OMeTAD
FTO/Glasss
TiO2
Perovskite
Spiro-OMeTAD
Au
FTO/Glasss
Perovskite
Spiro-OMeTAD
PCBM
Au
TiO2 TiO2
Au
(b) Perovskite-PCBM Bilayer
(a) Control PCBM-free
(c) Perovskite-PCBM Hybrid
32
4.3.2 DFT simulations
This reconfirmation of the strong PCBM-iodide interaction motivated me to explore, using
density functional theory (DFT), what might occur in a solid material. I looked in particular at
reactions PCBM might participate in at the excess-halide-associated defects at grain boundaries
previously reported to be a dominant source of electronic traps in lead methylammonium iodide
perovskites48,68,69,101. I focused specifically on the Pb-I antisite defect, in which iodine occupies
the Pb site and forms a trimer with neighbouring iodine atoms (Figure 4-3c)48. DFT reveals that,
with the introduction of PCBM near such Pb-I antisite defects, the wavefunction of the ground
state (Figure 4-3d) is hybridized between the PCBM and the perovskite surface. The bonding of
PCBM to defective halides is thermodynamically favoured and that this suppresses the formation
of deep traps (Figure 4-3e).
Figure 4-3. Perovskite-PCBM hybrid process and in situ passivation mechanism. (a) Pristine perovskite solution
(left) comprised of Pb(Ac)2 and MAI in dimethylformamide (DMF) solvent is bright yellow; the formulated
perovskite-PCBM hybrid solution (right) is brown; Simple one-step spin-coating is used to convert the hybrid solution
DO
S (
a.u
.)
Ev Ec
Trap state
600 800 1000 1200 1400 1600
0.0
0.5
1.0
No
rma
lize
d a
bso
rba
nce
(a
.u.)
Wavelength (nm)
PCBM in normal solvent
PCBM in Hybrid solution
Ac-e-
I- MA+
Pb2+
(a)
(b)
(c)
(d)
(e)
Grain boundary
Pb-I antistite
Pb I
33
into an hybrid solid film, and the perovskite is in situ passivated by PCBM during self-assembly; (b) UV-Visible
absorption spectroscopy of the hybrid solution shows the interaction between PCBM and perovskite ions. PCBM
radical anion’s absorption peak at 1020 nm is identified in hybrid solution (red); while PCBM in same solvent (black)
has no absorption peak in this wavelength region; Inset of (b) shows details of such interaction: In hybrid solution,
electron-transfer is induced between the perovskite anions (I-) and PCBM and will result in PCBM radical anions and
PCBM-halide radicals. (c) A schematic of in situ passivation of halide-induced deep traps: PCBM adsorbs on the Pb-
I antisite defective grain boundary during perovskite self-assembly. (d) The wavefunction overlap shows the
hybridization between PCBM and the defective surface, enabling the electron/hole transfer for absorbance and
passivation. (e) DFT calculation of the density of states (DOS) shows that deep trap states (black) induced by Pb-I
antisite defects are reduced and become much shallower (red) upon the adsorption of PCBM on defective halides. Ec,
minimum of conduction band; Ev, maximum of valence band.
4.4 Perovskite-PCBM mixture phase distribution
Next I sought to determine whether the PCBM is distributed throughout the entire thickness of
the active layer that had been formed from the mixed perovskite-PCBM solution (Figure 4-4a).
Secondary ion mass spectrometry (SIMS) was used to probe the depth profile of PCBM and
perovskite. Pb and Ti are used as indicators of the perovskite and of the TiO2 substrate,
respectively. Since PCBM does not contain elements to identify it uniquely, I used instead for
this portion of the study a thiophene-containing derivative, [60]ThCBM ([6,6]-(2-Thienyl)-C61-
butyric acid methyl ester), which permits the use of sulfur as the tracer element107. The
[60]ThCBM is present homogeneously throughout the thickness of the hybrid film, with a
uniform concentration as a function of depth (Figure 4-4b). Using X-ray diffraction (XRD), I
found that the perovskite lattice diffraction peaks of the hybrid film are consistent with that of
control films without PCBM (Figure 4-4c). In addition, the average perovskite grain size in
hybrid films, estimated from the XRD peak width, is comparable to that of control films.
Also with the nature of the mixed material in mind, I employed Kelvin probe studies to examine
the work function (WF) of mixed-material films. The work function of the mixed-material films
lies between that of the pure perovskite and pure PCBM. Its value varies monotonically along
this continuum as a function of PCBM fraction incorporated. When very high PCBM fractions
are employed, evidence of phase separation and impacts on film morphology emerge: the PCBM
phase aggregates at perovskite grain boundaries and becomes clearly evident (Figure 4-5).
34
Figure 4-4. 3D phase separation and homogeneous PCBM distribution in hybrid solid. (a) Scheme of planar
perovskite solar cell using perovskite-PCBM hybrid solid as the active absorber; PCBM phase is homogeneously
distributed at grain boundaries throughout the perovskite layer. (b) Secondary Ion Mass Spectrometry (SIMS) depth
profile of perovskite-PCBM hybrid film on TiO2 substrate showing homogeneous distribution of PCBM throughout
the film. The sputter etching begins at the air/film interface. PCBM is tracked by S element using analogous
[60]ThPCBM; perovskite is tracked by Pb element; TiO2 is tracked by Ti atom. (c) XRD patterns of the pristine hybrid
solid film (red) and the control perovskite film without PCBM (black). TiO2 compact layer on FTO is used as substrate.
XRD shows that in the hybrid solid, the perovskite crystal lattice is the same as the control film, and thus PCBM only
exists at the grain boundaries and interfaces throughout the film. (d) The transient photoluminescence of the hybrid
film. Pumping from the top of the film (black) and pumping from the bottom of the film (red) give identical signals,
showing homogeneous PCBM distribution. The hybrid film displays dense grains and full-coverage as observed via
SEM (inset left); the surface is ultra-flat with roughness ~6 nm as characterized by AFM (inset right).
These findings prompted me to posit the following picture of the mixed material. Perovskite
grains are formed with similar size and crystallinity with and without the PCBM (XRD). The
PCBM is distributed uniformly throughout the thickness of the film (SIMS), presumably in
0 100 200 300
0.01
0.1
1
No
rma
lize
d P
L (
a.u
.)
Time (ns)
Pump from Top
Pump from Bottom
(a)
0 100 200 300 400
0
1
No
rma
lize
d S
IMS
Co
un
ts (
a.u
.)
Sputter Time (a.u.)
Pb
S
Ti
TiO2 PCBM
Perovskite
(b)
10 20 30 40 50 60
XR
D in
tensity (
a.u
.)
2theta (deg)
Control:Perovskite
Hybrid:Perovskite:PCBM
(c)
(d)
PCBM
Perovskitegrains
TiO2 (ETL)
Spiro (HTL)
1um 1um
RMS=6.5nm
35
between the grains. The PCBM could bind iodide-rich defects sites on these grain boundaries
(DFT), and/or could simply bind up excess iodide from the solution. From an electronic
standpoint, the incorporation of the PCBM throughout the film influences its work function in
proportion with the PCBM-perovskite ratio (Kelvin Probe).
To seek further indications regarding the extent of electronic interactions between the PCBM and
the perovskite grains in the mixed material, I acquired transient photoluminescence for pure
perovskites, PCBM-perovskite bilayers, and mixed materials, investigating each for the case of
excitation from each side. The mixed-material film shows identical transient PL traces for both
top and bottom illumination schemes (Figure 4-4d). In contrast, perovskite films with PCBM on
one side only exhibit different PL lifetimes when pumped from the different sides. The
invariance of the PL lifetime with top/bottom-side photoexcitation for the mixed-material system
agrees with SIMS, and further indicates that the hybrid film behaves as a homogenous
optoelectronic material throughout its thickness.
Figure 4-5. PCBM phase separation at perovskite grain boundaries. (a) Bulk phase separation is evident when
very high PCBM fractions are employed. (b) Zoomed in view of a region where PCBM phase emerges at grain
boundaries.
4.5 Charge dynamics and hysteresis characterization
A series of conductive AFM studies provide added spatial resolution of the electronic properties
of the films under study. I carried out the cAFM studies under high vacuum and dark conditions
to rule out the effect of light and moisture. By overlapping the grain topography and the
electrical current map of the films (Figure 4-6a and Figure 4-6e), I find that conductivity is
greatest at grain boundaries, both in the pure-perovskite and in the mixed-material films.
30um
PCBM phase
(a) (b)
36
However, the mixed-material films have much higher conductivity near grain boundaries at
positive bias voltages, consistent with the evidence of the electron-transport medium PCBM
accumulating near grain boundaries and providing continuous pathways for electron egress. I
also obtained I-V traces at various spatial positions, and find that control perovskite films exhibit
major hysteresis behavior when scanned in the reverse bias direction (Figure 4-6b, 4c and 4d).
Given the pure perovskites’ slow response on the seconds timescale, the hysteretic I-V curves are
consistent with the proposed hysteresis mechanism of ionic transport in perovskite solids108–111.
To my knowledge, this is the first direct experimental observation of memristive properties
within the perovskite material itself via cAFM111. This observation is generally in agreement
with the very recently reported ionic motion processes in CH3NH3PbI3 perovskite materials71,112.
In contrast, in perovskite-PCBM mixed films, the hysteresis effect is greatly suppressed under all
conditions (Figure 4-6f, 4g and 4h). These observations further substantiate a picture in which
PCBM influences electronic properties when it associates with the perovskite grains at their
grain boundaries.
Figure 4-6. cAFM study of hysteresis-ion relationship for control films and hybrid films. (a) and (e) the gray-
scaled contact-mode AFM (background) with overlaid color-scaled conductive AFM images (positive sample bias
voltage: 1 V). (b-d) I-V hysteresis of control film increases when increase the negative bias and injected current (solid
line: forward sweep, dashed line: reverse sweep). (f-h) I-V hysteresis of hybrid film is suppressed when increasing the
negative bias and injected current. Scanning rate is ~0.5 V s-1 (see Methods).
37
Additional device studies offer further information about the role of PCBM in perovskite device
performance and hysteresis. Planar devices incorporating PCBM – whether at an interface or
throughout in bulk – are consistently superior in performance to control devices without PCBM
(Table 4-1). Incorporating the PCBM into the film becomes even more advantageous to collected
current for thicker active layers, suggesting that the PCBM accepts photocharges and assists in
their extraction to the TiO2. The champion planar devices were obtained using the perovskite-
PCBM mixed material and exhibited steady-state PCE exceeding 14.4%, 1.5 times more efficient
than my PCBM-free perovskite controls.
I also investigated the reverse saturation current density in the various devices and found that the
hybrid films consistently reduced the dark current by 2 orders of magnitude (Figure 4-7).
Rectifying behavior is also maintained much longer in the mixed material compared to
perovskite controls.
Figure 4-7. Long-term steady-state dark current measurement of planar devices. Perovskite-PCBM hybrid
devices (red and black) and control devices (blue and cyan) are tested under reverse bias -0.5 V. Hybrid devices
showed almost 2 orders of magnitude lower dark current and no breakdown during the course of the measurement.
Bias is applied continuously and dark current is sampled every 1 second.
0 5000 10000 15000 20000
1E-9
1E-8
1E-7
1E-6
1E-5
1E-4
Da
rk C
urr
en
t (A
)
Time (S)
Hybrid
Hybrid-2
Control
Control-2
TiO2 Peorvskite Spiro
38
These last observations motivate further evaluation of the role of PCBM in trap passivation at
perovskite grain boundaries. I used transient photovoltage to quantify the prevalence of mid-gap
trap states in each class of materials and devices (details see Methods). I obtained a notably
longer carrier lifetime over a wide range of photovoltages in the mixed material (Figure 4-8a).
This indicates reduced non-geminate recombination for the perovskite-PCBM hybrid films. I
also compare the transient photoluminescence of hybrid films to investigate the impact of PCBM
on carrier extraction. When the PCBM-perovskite hybrid ratio is increased progressively, the PL
exhibits consistently greater quenching, indicating efficient electronic coupling between the
well-dispersed PCBM phase and the perovskite (Figure 4-8b, orange, pink, and red curve).
When the PCBM ratio is extremely high (Figure 4-8b, black curve), the photoluminescence
quenching efficiency began to degrade greatly. Significant phase segregation occurs, with the
appearance of large PCBM domains that lack effective interconnectivity for carrier extraction. I
concluded that comprehensive incorporation of PCBM in the interstitial volumes among grains
in the perovskite system is required (i.e. sufficient PCBM material miscibility in the perovskite
solid is needed) to produce continuous pathways for carrier extraction to enhance
performance113.
Figure 4-8. Effect of PCBM on charge carrier dynamics. (a) Charge carrier lifetime of hybrid device (red) and
control device (blue), determined from transient photovoltage measurement under open-circuit condition. (b)
Transient photoluminescence of hybrid films with increasing PCBM ratio progressively (orange, pink, red, black)
compared with control film on glass (blue), showing the enhanced electron extraction. The quenching efficiency
increases monotonically with increasing hybrid ratio, indicating the increasing PCBM-perovskite interfaces. When
continuously increasing the PCBM hybrid ratio (black), the quenching efficiency begins to reduce abnormally, due to
0.6 0.7 0.8 0.9 1.0 1.1
1E-5
1E-4
1E-3
Hybrid
Control
Re
com
bin
atio
n lifetim
e (
S)
VOC (V)
(a)
0 100 200
1E-3
0.01
0.1
1
No
rmalized P
L (
a.u
.)
Time (ns)
Hybrid (1:200)Hybrid (1:100)Hybrid (1:50)
Hybrid (1:10)Control
(b)
39
the emergence of large domains of agglomerated PCBM, reducing the effective interconnectivity between perovskite
and PCBM.
4.6 Discussion: Ionic motion and hysteresis in perovskites
I close with a discussion of mechanisms likely at work, and one more speculative mechanism, in
the mixed-material films. My data suggests that PCBM, when incorporated at or near perovskite
grain boundaries, makes a significant impact on electronic properties. The transient
photovoltage, combined with the DFT analysis and the spectroscopy showing PCBM radical
formation, suggest that PCBM plays a passivating role at iodide-rich trap sites on the surfaces of
these grains. At the same time, the long timescale of hysteresis in pure perovskite films and its
substantial suppression in the mixed material, combined with the vastly lower reverse dark
current in the mixed material, suggest an additional effect at work in addition to the passivating
role. I propose that ions, such as the iodide anion, can potentially migrate under an applied
electric field, producing an ionic current. This can explain the slow response of hysteresis108,111
and the instability of the dark current when pure-perovskite and bilayer devices are employed.
By tying up iodide-rich surface sites, or simply unincorporated iodide anions, PCBM can reduce
anion migration through defects at grain boundaries108,109. This rearrangement under external,
and also built-in internal, electric fields, could account for solar cell hysteresis. For example,
when the device is poised at the JSC condition, the large built-in field may induce anionic charge
motion that works against this field, leading to a drop in photocurrent in time. A relatively rapid
scan towards VOC will therefore suffer from low photocurrents; whereas, following an extended
pause at VOC, during which anions can diffuse back to equilibrium positions, a rapid scan to JSC
will feature a high current in view of the lack of charge compensating the built-in field.
40
Chapter 5
Crosslinked hole-extraction interface improves hysteresis and stability
5.1 Introduction
So far, I identified that ionic motion may be a major source of hysteresis and instability in
perovskite devices, and this field-induced ionic motion was also found to be highly related with
the defects at interfaces and grain boundaries (Chapter 4). Given the major strides in active layer
engineering and hole-extraction interfaces, there exists now the opportunity to improve hole-
extraction interfaces to enable further progress in hysteresis and stability improvement.
The top surface (the hole-extraction interface in the device) of the perovskite is of particular
interest in the long-term stability of devices, because it is exposed directly to external stresses.
Specifically, since hybrid halide perovskites have a highly ionic character, they can decompose
under external stresses such as moisture, solvents, and heating cycles, especially if not fully
encapsulated41–43. The resultant ionic complexes are then highly reactive with transition metal
oxides114 (such as MoO3) and metal contacts115,116. Materials engineering strategies, such as
adding crosslinking among perovskite grains117, have been shown to enhance the stability of the
active material, including in the presence of moisture.
In perovskite devices employing a top hole-extracting contact, the engineering of the HTL (hole
transport layer) offers an opportunity to add protection to the perovskite that underlies it59,62,118–
120. The HTL should desirably be robust to external stresses, such as high operating temperatures,
and at the same time should efficiently facilitate hole extraction and thus promote overall device
performance. It should also be transparent to produce rear-metal-contact reflections and also to
enable semi-transparent and multi-absorber devices121,122.
The organic HTL widely used in many top-performing perovskite solar cells, spiro-MeOTAD
(2,2’,7,7’-Tetrakis(N,N-di-p-methoxyphenylamine)-9,9’-spirobifluorene), requires the ionic
dopant Li-TFSI (Bis(trifluoromethane)sulfonimide lithium salt) with additive tBP (4-tert-butyl
pyridine)35,53,54,78,103,117,123–126. This additive has been found to evaporate at 85°C59,63,64, limiting
devices’ thermal stability and also curtailing their capacity to withstand subsequent processing
41
steps. Further, this doping mechanism, which involves interactions with oxygen, requires fine
control63,127–129. The additive tBP and ionic dopant Li-TFSI have been found to interact with the
ionic perovskite layer and contribute thereby to the undesired introduction of water into the
active layer, thus contributing to perovskite device degradation59,62,129. Small pinholes in spiro-
MeOTAD layers were recently identified116,120,129 and found to facilitate the migration of iodine-
containing compounds from the perovskite, leading to corrosion of the metal top contact. Finally,
the sensitivity of spiro-MeOTAD to solvents imposes severe constraints on subsequent solution-
phase processing steps atop the device121.
To go beyond reliance on sensitive Li-doped spiro-MeOTAD atop the perovskite, alternatives
such as opaque carbon-based hole-extracting contacts119, hybrid carbon nanotube-polymers59,
and inorganic CuSCN130,131 have been investigated. These have shown improved device stability
and chemical robustness. To date, however, the benefits of these alternatives have come with
costs to performance: they have each quantitatively degraded solar cells’ open-circuit-voltage,
hysteresis, and fill factor. For example, the fill factors shown in these devices59,61,119,131 are
typically appreciably below the benchmark value (~75%) achievable in the state-of-art devices
employing doped spiro-MeOTAD35,54,103,117,123,124,126. These compromises to performance have
been ascribed to poor band level alignment and inefficient egress of charges across the resultant
interface.
In this Chapter, I pursued a new HTL strategy with the goals of protecting the perovskite,
achieving the needed free carrier density and work function without the use of chemical dopant
additives, and ultimately achieving high-performance perovskite solar cells that would exhibit
enhanced stability.
5.2 Crosslinked interface on perovskite top surface
My approach (Figure 5-1a) employed crosslinking of the polymer HTL in contact with the
perovskite in order to render the material insoluble and thermally stable. I would achieve the
needed deep work function and high hole free carrier density via a remote doping strategy (see
detailed methods in Appendix 1.5).
I focused the HTL work on arylamine derivatives, for these feature a HOMO (Highest Occupied
Molecular Orbital) level similar to the ionization potential of the perovskite (5.4~5.5eV). I first
42
explored UV-crosslinkable arylamine derivatives based on cationic ring-opening polymerization
of oxetane groups132–135, for these are known to be substantially inert when they are in intimate
contact with underlying active materials. The crosslinking process employs a cationic
photoinitiator to break the C-O bonds within each oxetane group under UV radiation. It thereby
constructs an insoluble network by forming a crosslinking C-O bond between different oxetane
groups.
Unfortunately, I witnessed much lower photovoltaic performance when compared to devices that
used conventional Spiro-MeOTAD. I proposed that known by-products132,133 produced in situ by
the cationic photo-acid initiator used in the UV-crosslinking process degrades the electronic
quality of the perovskite. I proposed that the organometal halide perovskites are particularly
sensitive and thus not immune to the cationic photoinitiator.
The observed in situ degradation motivated me to devise a perovskite-compatible crosslinking
agent. I focused on non-ionic polymerizable groups and found that by thermally inducing
crosslinking between styrene groups136, I could form crosslinked films using the new arylamine
derivative (N4,N4' -Di(naphthalen-1-yl)-N4,N4' -bis(4-vinylphenyl)biphenyl-4,4'-diamine),
which I term VNPB.
When VNPB was deposited using spin-casting, it allowed me to form crosslinked films that did
not evolve by-products that would degrade the underlying perovskite (Figure 5-1b). The needed
crosslinking proceeds under mild thermal conditions and does not require the use of an initiator:
instead, crosslinking is achieved by an addition reaction through the opening of the double bonds
in styrene groups of adjacent VNPB units. Multiple styrene groups in each VNPB unit enable the
formation of a three-dimensional network with good coverage and strength. Compared to the
corresponding NPB film that lacks crosslinking groups, the VNPB film is insoluble and
thermally stable, factors that enable the stacking of VNPB films via a layer-by-layer solution
process (Figure 5-2). Thermally-induced polymerization of styrene groups has also recently been
used to enhance an organic electrode interlayer that resides under the perovskite. Consistent with
my findings, the crosslinked interlayer showed remarkable resistance to solvent stress
(perovskite-soluble polar solvents such as DMF) and annealing-stress while casting the
perovskite atop137.
43
Figure 5-1. Hole extraction contact employing material crosslinking and interface doping. (a) Two-step scheme
to form the insoluble and thermally-stable hole extraction contact. In the first step, the organic hole transport layer
(HTL) is deposited and then thermally crosslinked; in the second step, an interface doping layer is simply deposited
atop of the HTL and doping is achieved via the interface charge transfer. (b) Details of the thermal crosslinking
process: double bonds (red lines) in styrene groups in the hole transport layer (VNPB) are opened and then crosslinked
via an addition reaction, thereby forming an insoluble, thermally-stable film. (c) Schematic of interface doping:
ground-state electron transfer occurs from the hole transport layer, having low ionization-energy, to the interface with
the high electron-affinity material, in this case transition metal oxide MoO3, thereby enhancing the hole carrier density
throughout the thin HTL. (d) Device structure of the planar perovskite device using a VNPB-MoO3 double-layer as
the top hole extraction contact. (e) The SEM cross-sectional image shows the full device covered by a dense hole
extraction layer based on the VNPB-MoO3 double-layer stack. (f) High resolution TEM further resolves the fine
interface of the VNPB-MoO3 double-layer. VNPB and MoO3 are confirmed to be in a dense and amorphous phase,
forming a smooth interface with the underlying polycrystalline perovskite layer.
Interface electron transfer
a
b c
[ ]
N N
[ ]
n n
N N
… … ……
Thermal crosslink
VNPB
d
FTO on Glass
TiO2/PCBM
Perovskite
VNPBMoO3
Metal (Au)
200nm
e
10nm
Perovskite
MoO3
VNPB
Au
Interface doping layer
(MoO3)
Ener
gy +
-
Hole transport
layer
Ef
Ef
f
44
Figure 5-2. Thermally crosslinked VNPB is insoluble and enables the layer-by-layer deposition. (a) Cross-
sectional SEM image of 2 layers of crosslinked VNPB and (b) 4 layers of crosslinked VNPB film deposited by spin-
casting in a layer-by-layer fashion. It confirms that the final film exhibits a bulk morphology, and there is no detectable
interface between adjacent layers. Noted that for ease of observation, the concentration of VNPB used in this test is
much higher than that used in device hole transport layer.
5.3 Remote doping for hole-extraction conductivity
The VNPB crosslinked layer is intrinsic, and thus incapable of efficient hole extraction from the
perovskite. I sought to introduce free holes, and to do so without chemical doping used in Spiro-
MeOTAD59,62,116,120,129. I pursued an interface remote doping strategy138, wherein I deposited a
deep-workfunction transition metal oxide layer atop the HTL (Figure 5-1a). The dense and
chemically-inert crosslinked VNPB would serve to keep physically separate, and thereby prevent
chemical reactions among, the perovskite and the metal oxide (MoO3) layers114. The free hole
density would be introduced into the otherwise-intrinsic HTL via ground-state electron transfer
to the deep-workfunction metal oxide (MoO3 in this work) at the organic-inorganic interface
(Figure 5-1c)139,140.
High resolution microscopy confirmed intimate contact between the VNPB and MoO3 (Figure
5-1d-f), a precondition for efficient interface electron-transfer. The VNPB-MoO3 double-layer
structure was also confirmed to be in a dense amorphous phase and thus suffered no issues of
lattice mismatch with the polycrystalline perovskite. Ultraviolet photoelectron spectroscopy
studies (UPS) confirmed that the HOMO level of the VNPB layer (-5.4~-5.5 eV) is highly
aligned with perovskite (Figure 5-3), enabling a substantially barrierless hole extraction pathway.
45
Figure 5-3. Ultraviolet photoelectron spectroscopy (UPS) studies of VNPB layer. UPS was carried out using He I
(21.22 eV) photon lines from a discharge lamp. The applied bias is 15 eV. The VNPB film is tested on an Au substrate.
The statistics analysis of multiple samples gives the HOMO level = -5.46 ± 0.08 eV.
5.4 Efficient PV with reduced hysteresis
The robust crosslinked hole transport layer (VNPB), coupled with an inorganic metal oxide layer
(MoO3) in a double-layer fashion, not only provides thermally-stable and solvent-resistant
protection for the perovskite, but also provides a stable and efficient doping process that leads to
a 16.5% solar PCE measured at steady state (Figure 5-4).
In planar devices that employed the new contact strategy (see methods in Appendix 1.5), I
observed highly stable steady-state photovoltaic performance when the devices were operated at
their maximum power point (Figure 5-4c, red curve; see test details in Appendix 2.1). Their
performance was equivalent to the benchmark devices that employed spiro-MeOTAD hole
transport layers (Figure 5-4c, blue curve). External quantum efficiency (EQE) measurements
were carried out and agreed with the measured current densities.
20 15 10 5 0
I. (
a.u
.)
Binding energy (eV)
3 2 1 0 -1
Binding energy (eV)
He I (21.22eV)
46
Figure 5-4. Improved photovoltaic performance with interface doping. (a) The instantaneous J-V curve of control
devices using doped Spiro-MeOTAD (blue) as the hole transport layer, compared with devices using undoped Spiro-
MeOTAD (grey). The undoped Spiro-MeOTAD leads to a sharply reduced fill factor (FF) and performance. Arrows
indicate the voltage scanning direction. The thicker curve is the forward scan starting from open circuit condition
while the thin curve is the reverse scan starting from short circuit condition. The scanning rate is 0.2 V s-1. The inset
of (a) illustrates Spiro-MeOTAD doping by the use of Li salts (blue dot) throughout the film. (b) The J-V curve of
newly-designed devices using the VNPB-MoO3 interface doping hole-extraction contact (red), compared with devices
using VNPB alone (grey). Without the interface doping, the FF and overall performance show a stark decline. In
contrast, the device using interface doping shows an 80% FF with negligible hysteresis. The inset of (b) illustrates the
doping at the interface (red region) of the VNPB-MoO3 double-layer. (c) Steady state power conversion efficiency
(PCE) operated at the maximum power point of devices using interface doped VNPB (red squares), doped Spiro-
MeOTAD (blue circles), undoped VNPB (grey triangles) and undoped Spiro-MeOTAD (grey diamonds). The stable
current output at maximum power point indicates no hysteresis, which is consistent with the observation in the test of
instantaneous JV shown in (b). The decay of steady-state PCE of undoped devices is typically associated with
hysteresis in JV curves. (d) UV-Visible-IR absorption spectroscopy of the VNPB-MoO3 double-layer (red) showing
the signature of the interface charge-transfer-complex (CTC) in the near-infrared absorption region, while MoO3
(black) and VNPB individual layers (grey) exhibit no absorption features in the same wavelength region. Inset of (d)
illustrates that the CTC (red region) resides at the interface when VNPB is covered by the interface doping layer
MoO3. (e) The photoluminescence (PL) quenching effect in a VNPB-MoO3 double-layer (red) versus a VNPB single
0 50 100 150
4
5
6
7
8
10
15
20
Doped-Spiro
Doped-VNPB
Undoped-Spiro
Undoped-VNPB
Time, t(s)0.0 0.2 0.4 0.6 0.8 1.0 1.2
-10
0
10
20
Cu
rre
nt
den
sity (
mA
cm
-2)
Voltage (V)
Li-saltdopant
Spiro
Perovskite
TiO2
Control
Doped
Undoped
0.0 0.2 0.4 0.6 0.8 1.0 1.2
-10
0
10
20
VNPB
MoO3 Interface doping
Perovskite
TiO2
Crosslinked
Doped
Undoped
Steady-state efficiency, PCE(%)c
e
a b
d
400 800 1200 1600 2000
0
10
20
Ab
so
rpta
nce
(%
)
Wavelength (nm)
MoO3
VNPB
VNPB-MoO3
Interface charge-transfer-complex (CTC)
MoO3
VNPB
0 5 10 15 20 25 30
10-2
10-1
100
VNPB
VNPB-MoO3
Tra
nsie
nt P
L c
ou
nt (a
.u.)
Time (ns)
400 500 600
PL
coun
t (a.
u.)
(nm)
47
layer (grey), induced by the interface charge-transfer-complex, is observed in transient and steady-state PL
measurements (inset) of VNPB. The arrow indicates the VNPB PL peak (450 nm) where transient PL was measured.
I observed that the instantaneous JV-curve provides a > 80% fill factor with negligible hysteresis
in planar devices (Figure 5-4b, red curve). From a statistical analysis on a large sampling of
devices that featured the new interfacial remote-doping hole-extraction contact, the average
hysteresis was found to be low (2%), an improvement from my control devices that had used
spiro-MeOTAD (6%). The low hysteresis agrees with the concept that the inert perovskite-
crosslinked interface reduces chemical actions such as those seen with chemically-doped spiro-
MeOTAD59,62,116,120,129. Here I quantify the hysteresis using Equation (5-1):
𝐇𝐲𝐬𝐭𝐞𝐫𝐞𝐬𝐢𝐬 = (𝐀𝐫𝐞𝐚𝒇𝒐𝒓𝒘𝒂𝒓𝒅
𝐀𝐫𝐞𝐚𝒓𝒆𝒗𝒆𝒓𝒔𝒆− 𝟏) × 𝟏𝟎𝟎% (5-1)
where Area𝑓𝑜𝑟𝑤𝑎𝑟𝑑 (Area𝑟𝑒𝑣𝑒𝑟𝑠𝑒) is the integrated area under the forward (reverse) scanning JV
curve. Low hysteresis correlates with stable steady-state output power. Devices with high
hysteresis in JV curves show decay of output current and power when operated at steady-
state65,118,141. To avoid any overestimation of PCE arising from JV hysteresis, I report the solar-
to-electricity efficiency only using the steady-state power-to-power performance of devices
operated at their maximum power point under constant AM1.5 solar illumination.
The high fill factor and low hysteresis indicate efficient charge extraction in the best-designed
remote-doped devices. In contrast, the fill factor was seriously compromised in all undoped
controls (Figure 5-4a and 2b, grey curves). The steady-state performance of undoped controls is
low, decays further during testing, and is accompanied by high hysteresis (Figure 5-4c). The
same trend of device degradation occurs in both classes of undoped devices (spiro-MeOTAD
devices without Li salt mixing; and crosslinked VNPB devices without a MoO3 interface layer).
This reconfirms the crucial role played by the MoO3 layer in introducing across-the-interface
remote doping in the crosslinked VNPB layer.
48
5.5 Mechanistic study of remote-doped hole-extraction
5.5.1 Material characterization
I sought mechanistic insights into the role of the interface doping. Specifically, I investigated the
physical picture of interface doping at the VNPB-MoO3 organic-inorganic heterojunction, and
explored the design criteria for this new double-layer hole extraction structure.
UV-Vis-IR spectroscopy provides one means to study electron transfer at the interface: when
MoO3 is deposited on top of a VNPB film, I observe a near-infrared absorption peak that I
associate with the interface charge-transfer-complex142–145. This is in contrast with either
individual VNPB or MoO3 layers, which on their own are transparent in this wavelength region
(Figure 5-4d). This sub-bandgap absorption feature has previously been associated in literature
reports with the formation of intermediate states induced by charge-transfer-complexes at the
donor-acceptor interface. The attribution of this spectral feature to ground-state charge-transfer-
complexes at the interface is further verified by photoluminescence (PL) quenching effects
observed both in steady-state PL (Figure 2e, inset) and time-resolved PL measurements (Figure
5-4e). The PL tests reveal that the interfacial charge-transfer-complexes behave as quenching
sites for excitons in VNPB films through polaron-exciton quenching146–148. Consistently, in
spiro-MeOTAD, sub-bandgap absorption, PL quenching and time-resolved-PL quenching are
observed only when spiro-MeOTAD is doped using the Li-salt (see doping method in Appendix
1.6). In VNPB-MoO3, charge-transfer doping is accomplished at the interface alone, and
therefore the absolute parasitic absorption is much less than that in bulk doped spiro-MeOTAD.
These findings further agree with the picture of interface doping via ground-state electron-
transfer at the VNPB-MoO3 heterojunction depicted in Figure 5-1c.
5.5.2 Optoelectronic simulations
Next I sought insights into the role of interface doping in devices. I made use of self-consistent
optoelectronic device simulations149 and looked particularly at the organic-inorganic interface of
VNPB-MoO3 in the steady state. The band-bending of VNPB and MoO3 layers (Figure 5-5a and
3b) reconfirms the p-type doping of VNPB via interface electron-transfer. The theoretically-
predicted J-V behavior (Figure 5-5c) of solar devices with and without a MoO3 interface doping
layer are in excellent quantitative agreement with experimental results (Figure 5-4b). The
49
simulations also predict that the interface doping layer will increase photovoltaic performance,
most significantly through the fill factor of devices. When I tuned the HOMO level of the hole
transport layer (Figure 5-5d), I observed linear control over the doping effect: fill factor increases
when the HOMO of the hole transport layer moves toward the HOMO of the perovskite layer.
VNPB satisfies this design rule very well, as confirmed by UPS measurements of the HOMO
level (Figure 5-3). With respect to the choice of work function, there exists a wide performance-
insensitive region (Figure 5-5e and 3f), requiring only that the workfunction of the interface
doping layer be deeper than the HOMO of the HTL (~-5.4 eV). The hole extraction efficiency,
associated with fill factor, declines sharply only when the workfunction of the doping layer is too
shallow to accept electrons. Fortunately, the interface doping material MoO3 resides - even when
one accounts for the spread in reported workfunctions58,139,140,146,147,150- in the performance-
insensitive region (-5.4 ~ -7 eV).
Figure 5-5. Electrical simulation of devices using interface doping. (a) The equilibrium-state energy band diagram
of devices using interface doping hole extraction contacts (VNPB-MoO3). Ec (Ev) indicates the edge of the conduction
(valence) band while Ef and red line denote the Fermi level. (b) Expanded view of the band alignment and band
0.0 0.5 1.0
0
10
20
Interface
doping layer
workfunction =4.6eV
5.2
4.85.0
>5.4eV
Cu
rre
nt D
en
sity (
mA
cm
-2)
Voltage (V)
-5.0 -5.2 -5.4 -5.6
60
80
FF
PCE
Hole transport layer
HOMO level (eV)
FF
(%)
12
16
20
PC
E (%
)
HOMO of VNPB
-4
-3
-2
-1
0
1
2
3
4
5
Ec
Ev
EFE (
eV)
0.0 0.2 0.4 0.6 0.8 1.0 1.20
5
10
15
20
Cu
rre
nt D
en
sity (
mA
cm
-2)
Voltage (V)
VNPB
VNPB-MoO3Ec
Ev
EF
VNPB
MoO3
-5 -6 -7
20
40
60
80
FF
PCE
Interface doping layer
workfunction (eV)
FF
(%)
10
15
20
PC
E (%
)
HOMO of HTL
HOMO of perovskite
a b c
fd e
TiO2
Perovskite
VNPB/MoO3
Au
50
bending at the interface of the VNPB-MoO3 double-layer stack. (c) Performance comparison between devices with
(red) and without (grey) interface doping layers. Without interface doping, the fill factor remarkably decreases,
consistent with experimental observations (Figure 2b). (d) Performance evolution when the HOMO level of the hole
transport layer (HTL) changes. The fill factor increases when the HTL HOMO is well aligned with the HOMO of
perovskite (dash line). The arrow indicates that the HOMO level of crosslinked VNPB (5.46 ± 0.08 eV measured from
UPS, Figure S3) is highly aligned with the HOMO level of perovskite, and therefore is expected to result in the
optimized performance when used as the HTL. (e) Performance dependence on the workfunction of the interface
doping layer. The fill factor is unaltered in a rather extended range (-5.4 ~ 7 eV), as long as the workfunction of the
interface doping layer is deeper than the HOMO of the hole transport layer (dash line). (f) J-V curves corresponding
to the performance evolution shown in (e) showing that the performance drop occurring when the workfunction of the
interface doping layer becomes shallower than the HTL HOMO. The reported MoO3 workfunction range resides in
the optimal performance region.
5.6 Improved stability under external stress
I then proceeded to assess the enhanced stability of the new devices under external stresses such
as heating, moisture, and solvent. Devices with a remote-doped crosslinked top contact (VNPB-
MoO3) were first investigated under thermal stress. Most striking is the devices’ retention of
their superior performance (maintenance of at least 95% of initial performance) and low
hysteresis behavior following fully 1 hour of annealing at ~100˚C (Figure 5-6b). Under the same
stress, control devices with conventionally-doped spiro-MeOTAD lost more than 30% of their
performance irreversibly. This came principally through a severe degradation in fill factor and an
increase in hysteretic behavior (Figure 5-6a). Consistently, the rectification, under dark
conditions, of conventional devices also degraded irreversibly. Similar trends as those for
hysteresis degradation and performance decay in control devices were also observed in long-term
steady-state performance testing. Just as in the undoped-spiro-MeOTAD devices (Figure 5-4a),
the serious degradation in fill factor and the hysteresis indicates the loss of doping efficiency
under such thermal stress. Additional direct evidence came from an optical microscopy study: I
observed a new crystalline pattern in doped spiro-MeOTAD films (Figure 5-6c) following the
thermal stress test. The irreversible morphology degradation is linked to the phase separation of
dopant and spiro-MeOTAD host, a change that coincides with outgassing of the tBP additive at
temperatures that exceed 85˚C59,63,64. In contrast, the morphology of the VNPB-MoO3 double-
layer is thermally stable (Figure 5-6d), attributed to the robustness of both the crosslinked
material and the interface remote-doping mechanism.
51
Figure 5-6. Evolution of performance, morphology and material under external stress. (a) The performance of
devices using Spiro-MeOTAD as the hole-extraction contact tested at room temperature (grey) and after a 110 ˚C
burn-in test (blue) [In the burn-in test, devices are annealed at 110 ˚C for 1 hour in an N2 environment and tested after
cooling down to room temperature]. (b) The performance of devices using VNPB-MoO3 tested at room temperature
(grey) and after 110 ˚C burn-in process (red). (c) Optical microscopy (reflection mode) of a doped Spiro-MeOTAD
film before (left) and after burn-in (right). The annealed film shows chain-like structures, leading to the irreversible
morphology degradation. (d) The morphology of a VNPB-MoO3 film before (left) and after burn-in (right). No visible
morphology evolution can be observed. (e) The evolution of perovskite content in the device active layer, tracked
using the PbI2 peak (•) and perovskite peak (*) in XRD measurements. Devices are tested after storage in air (70%
RH, dark) for 10 days and 30 days. PbI2 peak of the perovskite film in a Spiro-MeOTAD device (left) emerges after
10 days (grey) and dominates the perovskite peak after 30 days (blue), indicating severe decomposition of the
perovskite phase. In contrast, the perovskite layer is well protected by the VNPB-MoO3 film (right) and shows
negligible PbI2 signal even after 30 days (red).
I used X-ray diffraction (XRD) to explore the evolution of the perovskite active layer in devices
placed under stress via the introduction of moisture (70% RH) combined with elevated
0.0 0.2 0.4 0.6 0.8 1.0 1.2
-10
0
10
20
Control
Before
After
Cu
rre
nt d
en
sity (
mA
cm
-2)
Voltage (V)
BeforeAnnealing (110 ˚C, 1 hour)
Co
ntr
ol
Cro
sslin
ked
a b
c
d
0.0 0.2 0.4 0.6 0.8 1.0 1.2
-10
0
10
20
Crosslinked
Before
After
Cu
rre
nt d
en
sity (
mA
cm
-2)
Voltage (V)
10 15 20
Crosslinked
Perovskite
10 15 20
XR
D c
ou
nts
(a
.u.)
2deg
Control
PbI2
**
After 30 days
After 10 days
PerovskiteeAfter
52
temperatures. The degree of perovskite degradation was quantified by the ratio of PbI2 peaks to
perovskite peaks. In control devices with spiro-MeOTAD as the top contact, degradation is
noticeable within 10 days and becomes significant after 30 days (Figure 5-6e, left). After 30 days
the PbI2 peak exceeds the perovskite signal, and the film is visibly yellow.
In contrast, following 30 days of moist heat, the perovskite in devices covered by the crosslinked
layer did not change within measurement uncertainty (Figure 5-6e, right). This finding further
confirms that the crosslinked hole transport layer, coupled with dense inorganic metal oxide in a
double-layer fashion, provides the perovskite with superior physical protection.
I also investigated the potential to adapt the new device structure in the direction of enabling
multi-junction cells. I measured the impact of subsequent solvent exposure (Figure 5-7),
applying a polar solvent (methanol) often used in follow-on layer fabrication. The conventional
spiro-MeOTAD device was much degraded as seen in its bandedge absorption change (Figure
5-8a, left). When the same device is exposed to chlorobenzene, the sandwiched inorganic spiro-
MeOTAD layer dissolves, producing irreversible loss of device structure and morphology
(Figure 5-8b, upper). I concluded that the conventional perovskite materials stack is vulnerable
even when subjected to nominally orthogonal solvents.
In contrast, the device covered with a crosslinked HTL retains the active material and the device
structure when exposed to both polar and nonpolar solvents (Figure 5-8a, right; Figure 5-8b,
downside). The enhanced resistance to heat, moisture, and follow-on solvent-based processing
not only benefit the single cell, but also open avenues for fabricating multi-junction devices atop
the perovskite.
53
Figure 5-7. Assessment of device evolution under the external solvent attack. For the sake of fair comparison and
simulation of multijunction fabrication scenario, the same MoO3 layer and ZnO layer were deposited on top of (a)
Spiro-covered device and (b) crosslinked VNPB covered device, using high vacuum deposition methods. Two devices
(a) and (b), are soaked in both polar solvent (methanol) and nonpolar solvent (chlorobenzene) for 30s to investigate
the material or morphology evolution.
Figure 5-8. Evolution of material and morphology under the external solvent attack. (a) Evolution of perovskite
active layer after devices soaked in polar solvent (Methanol) for 30s. [The device layer configuration and test method
are shown in Figure 5-7.] The device using a Spiro hole transport layer (a, left) shows serious degradation of the
perovskite absorption edge, before (grey) and after (blue) methanol soaking. The device using a crosslinked hole
transport material (a, right) shows identical perovskite absorption before (grey) and after (red) solvent soaking. (b)
Evolution of top surface morphology after devices soaked in nonpolar solvent (chlorobenzene) for 30s. In device using
FTO on Glass
TiO2/PCBM
Perovskite
Crosslinked VNPBMoO3 (thermal evap)
ZnO (Sputter deposit, 50nm)
FTO on Glass
TiO2/PCBM
Perovskite
Doped Spiro-OMeTAD
MoO3 (thermal evap)
ZnO (Sputter deposit, 50nm)
Solvent soaking:Methanol; Chlorobenzene
Solvent soaking:Methanol; Chlorobenzene
a b
Control Crosslinked HETC
500 600 700 800
Spiro
Before
After
Ab
so
rptio
n (
a.u
.)
wavelength (nm)500 600 700 800
Crosslinked
Before
After
Methanol soaking
100µm
Chlorobenzene soaking Before After
Co
ntr
ol
Cro
sslin
ked
a b
54
Spiro (b, upper), the top morphology is flat (b, upper left) before solvent soaking and becomes cracked across the
whole area of device after solvent soaking (b, upper right), due to the dissolution and reconstruction of the underlying
Spiro layer. In the device using the crosslinking VNPB material (b, down), the top surface morphology is kept
essentially the same before (b, down left) and after (b, down right) solvent soaking, due to the underlying insoluble
hole transport material.
5.7 Conclusions
In summary, I demonstrated a new methodology for hole extraction on top of planar perovskite
solar cells. The crosslinked organic hole transport material, coupled with an inorganic metal
oxide, provides an insoluble and inert physical protection layer, combined with high conductance
for hole extraction. This enabled device performance that is stable for longer durations and under
more intense external stresses than in many prior reports.
55
Chapter 6
Conclusions
6.1 Summary and Impact
The research in this thesis was geared to providing new material and device designs to achieve
high efficiency perovskite solar cells. The focus was on resolving the problems – hysteresis and
instability – specific to this emerging solar PV system. By performing first-principles
simulations (DFT) and optoelectronic simulations, I was able to identify engineering routes at
both the materials and device levels. By gaining control over the material chemistry throughout
the interfaces within the photon-absorber and charge-extraction contact, I was then able to create
novel architectures that contributed to improved performance.
The work began with clarifying fundamental aspects underlying the impact of growth conditions
on the performance of perovskite films (Chapter 3). It revealed delocalization of the electronic
states within the local nanocrystal surfaces that preserves the integrity of the bulk bandgap. The
DFT-based analysis of defect formation energies identified the key defects (Pb atom substituted
by I) and indicated that films grown under iodine-rich conditions are prone to a high density of
deep electronic traps (recombination centers). This finding motivated the exploration of a new
precursor (lead acetate) for device-quality films. Insight into defect physics have spawned a
broad range of new perovskite growth processing reported in the wider literature.
The work achieved success in reducing current hysteresis and instability in planar cells through
the materials engineering of electron-extraction interfaces and grain boundaries (Chapter 4). I
reported the first perovskite-PCBM hybrid solid and found that the PCBM throughout the grain
boundaries and electron-extraction interfaces suppresses the hysteresis that has plagued planar
perovskite devices. I then conducted in-depth material characterizations and DFT simulations
that revealed the PCBM-perovskite interaction: the PCBM passivates the key PbI3- antisite
defects during perovskite self-assembly.
Using conductive AFM studies, I revealed the memristive properties of perovskite films and
identified the major cause of hysteresis to be ionic migration under electric fields in films, an
56
initial step along this research front. This proposed mechanism of hysteresis now resonates with
experimental and theoretical results obtained in subsequent research reported in the literature.
I closed with the engineering of the exposed hole-extraction interface on the perovskite top
surface, which is of particular importance for further improving device stability and performance
simultaneously (Chapter 5). I developed the first crosslinked hole-extraction top contact to
obviate in situ degradation of the underlying perovskite. The new crosslinked hole-transport
medium produces an insoluble and heat-resistant materials stack atop the perovskite that is band-
aligned with the perovskite. I also found that hole-extracting contacts, which rely on chemical
doping, were the weak link from a stability perspective in the best-performing perovskite cells. I
therefore employed a new remote-doping strategy to induce the needed work function and free
carrier density. The resultant family of devices is hysteresis-free, with fill factors exceeding 80%
and with excellent resilience to thermal stresses that exceed 100˚C, conditions under which
conventionally-contacted devices fail. The devices are also resistant to stresses produced by
moisture and solvents that cause conventional devices to decompose.
My methodology represents a new strategy for constructing transparent, highly-conductive,
thermally-stable and solvent-resistant top contacts in perovskite devices. This top contact is the
base to build a second cell in a multi-junction configuration, paving the way for broader light
absorption and higher efficiency potential.
The work in this thesis directly resulted in the publication of 3 peer-reviewed journal articles that
have been cited >300 times during the past 2 years.
6.2 Outlook for perovskite solids and PV
6.2.1 Maximum efficiency
During the course of the work reported in this thesis, advances in perovskite PV continued, with
certified efficiencies (not stabilized) above 20% achieved after only 5 years of worldwide
research efforts126,151,152. There, however, is still room for further efficiency improvement toward
the Shockley–Queisser limit (VOC=1.32V; JSC=25mA cm-2; FF=90.5%; PCE=30% for a
semiconductor with bandgap Eg=1.6 eV). The relative current fraction is ~0.88 (Figure 6-1). The
current loss comes primarily from front reflection and parasitic absorption in the hole-extraction
layer and back contact. Compared with c-Si and GaAs solar cells, the voltage and fill factor loss
57
in the presently-best perovskite PV is particularly high, because of the charge carrier
recombination paths in the absorber and at the interfaces, the shunting paths in non-ideal films
and carrier-extraction contacts, and resistive losses due to non-ideal contacts. From this analysis,
there is still a significant opportunity for further progress on material processing and interface
engineering for better light absorption and carrier management26.
Another research front lies in developing smaller bandgap perovskites which are desirable for
achieving a higher Shockley-Queisser efficiency limit. Highly efficient larger bandgap
perovskites are also of interest, valuable as the front cell in a multi-junction cell, such as a
perovskite-Si tandem cell with a potential efficiency that exceeds 30%.
Figure 6-1. Fraction of Shockley-Queisser detailed-balance limit for voltage and current achieved by record
cells. The lines crossing some data points indicate the uncertainty of bandgap of the record cell. ηSQ indicates the
maximum efficiency following the S-Q model. Figure reproduced from ref. 26. Copyright 2016 by the American
Association for the Advancement of Science.
6.2.2 Long-term stability
Despite the exciting efficiencies tested in labs, today’s best perovskite solar cells are known to
degrade within days under standard operating conditions. This is the greatest gap to commercial
cells that are required to work for more than 20 years. The origins of instabilities and hysteresis
are not totally understood yet, and should be kept as a topic of active research, even though UV
photo-reduction, water reaction, and ionic migration from defects have already been documented
perovskite
mc-Si
CdTe
CIGS
InP
GaInP
GaAs
c-Si1.00
0.95
0.90
0.85
0.80
0.75
0.70
0.65
0.6010.90.80.70.5 0.6
Carrier management
Ligh
t m
anag
em
en
t
v x f (FF VOC / FFSQ VSQ)
j(J SC
/ J SQ
)
58
as mechanisms. Developing electron- and hole-extraction materials and contact designs for long-
term operation in perovskite devices is crucial. Finding new perovskite materials with stability
advantages without efficiency compromises will represent a big breakthrough in the future.
Compared with CdTe and GaAs solar cells, toxicity associated with Pb is more serious because
of the higher water solubility of perovskite. Therefore, intense research on perovskite materials
free of toxic elements is also of great worth.
Making big steps forward in perovskite photovoltaics relies on a coordinated international and
interdisciplinary effort. The urgent need for high efficiencies and low costs in photovoltaics
creates a powerful motivation to continue pursuing rapid advances. The further success of
solution-processed perovskite thin film photovoltaics will enable even greater penetration of
renewable electricity into our energy system and daily life in the future, spanning solar farms,
building-integrated systems, and mobile electronics.
59
References
1. Jeff Taso, Nate Lewis & George Crabtree. Solar FAQs. (U.S. Department of Energy, 2006).
2. Renewables 2016 Global Status Report. REN21 (2016). Available at: http://www.ren21.net/status-of-
renewables/global-status-report/.
3. Nathan S. Lewis, George Crabtree, Arthur J. Nozik, Michael R. Wasielewski & Paul Alivisatos. Basic Research
Needs in Solar Energy Utilization. (U.S. Department of Energy, 2005).
4. Green, M. A., Emery, K., Hishikawa, Y., Warta, W. & Dunlop, E. D. Solar cell efficiency tables (Version 45):
Solar cell efficiency tables. Prog. Photovolt. Res. Appl. 23, 1–9 (2015).
5. Panasonic HIT(R) Solar Cell Achieves World’s Highest Energy Conversion Efficiency of 25.6% at Research
Level | Headquarters News | Panasonic Newsroom Global. (2014). Available at:
http://news.panasonic.com/global/press/data/2014/04/en140410-4/en140410-4.html. (Accessed: 3rd November
2016)
6. Kayes, B. M. et al. 27.6% Conversion efficiency, a new record for single-junction solar cells under 1 sun
illumination. in 2011 37th IEEE Photovoltaic Specialists Conference (PVSC) 000004–000008 (2011).
doi:10.1109/PVSC.2011.6185831
7. First Solar Achieves Efficiency, Durability Milestones (NASDAQ:FSLR). Available at:
http://investor.firstsolar.com/releasedetail.cfm?ReleaseID=895118. (Accessed: 3rd November 2016)
8. Jackson, P. et al. Properties of Cu(In,Ga)Se 2 solar cells with new record efficiencies up to 21.7%: Properties of
Cu(In,Ga)Se 2 solar cells with new record efficiencies up to 21.7%. Phys. Status Solidi RRL - Rapid Res. Lett. 9,
28–31 (2015).
9. O’Regan, B. & Grätzel, M. A low-cost, high-efficiency solar cell based on dye-sensitized colloidal TiO2 films.
Nature 353, 737–740 (1991).
10. Mathew, S. et al. Dye-sensitized solar cells with 13% efficiency achieved through the molecular engineering of
porphyrin sensitizers. Nat. Chem. 6, 242–247 (2014).
11. R. Komiya et al. Improvement of the conversion efficiency of a monolithic type dye-sensitized solar cell
module. in (2011).
12. Shaheen, S. E. et al. 2.5% efficient organic plastic solar cells. Appl. Phys. Lett. 78, 841–843 (2001).
13. Yu, G., Gao, J., Hummelen, J. C., Wudl, F. & Heeger, A. J. Polymer Photovoltaic Cells: Enhanced Efficiencies
via a Network of Internal Donor-Acceptor Heterojunctions. Science 270, 1789–1791 (1995).
14. Koleilat, G. I. et al. Efficient, Stable Infrared Photovoltaics Based on Solution-Cast Colloidal Quantum Dots.
ACS Nano 2, 833–840 (2008).
15. Lan, X. et al. 10.6% Certified Colloidal Quantum Dot Solar Cells via Solvent-Polarity-Engineered Halide
Passivation. Nano Lett. 16, 4630–4634 (2016).
16. Sargent, E. H. Colloidal quantum dot solar cells. Nat. Photonics 6, 133–135 (2012).
17. Todorov, T. & Mitzi, D. B. Direct Liquid Coating of Chalcopyrite Light-Absorbing Layers for Photovoltaic
Devices. Eur. J. Inorg. Chem. 2010, 17–28 (2010).
18. Wang, W. et al. Device Characteristics of CZTSSe Thin-Film Solar Cells with 12.6% Efficiency. Adv. Energy
Mater. 4, 1301465 (2014).
19. Haight, R. et al. Band alignment at the Cu2ZnSn(SxSe1−x)4/CdS interface. Appl. Phys. Lett. 98, 253502
(2011).
20. Solar Spectral Irradiance: Air Mass 1.5. Available at: http://rredc.nrel.gov/solar/spectra/am1.5/. (Accessed: 3rd
November 2016)
21. Shockley, W. & Queisser, H. J. Detailed Balance Limit of Efficiency of p-n Junction Solar Cells. J. Appl. Phys.
32, 510 (1961).
22. M. A. Green. Third Generation Photovoltaics. 12, (Springer Berlin Heidelberg, 2006).
23. King, R. R. Multijunction cells: Record breakers. Nat. Photonics 2, 284–286 (2008).
24. Sharp Develops Concentrator Solar Cell with World’s Highest Conversion Efficiency of 44.4% | Press Releases
| Sharp Global. Available at: http://www.sharp-world.com/corporate/news/130614.html. (Accessed: 3rd
November 2016)
25. Graetzel, M., Janssen, R. A. J., Mitzi, D. B. & Sargent, E. H. Materials interface engineering for solution-
processed photovoltaics. Nature 488, 304–312 (2012).
26. Polman, A., Knight, M., Garnett, E. C., Ehrler, B. & Sinke, W. C. Photovoltaic materials: Present efficiencies
and future challenges. Science 352, aad4424 (2016).
60
27. Nazeeruddin, M. K., Baranoff, E. & Grätzel, M. Dye-sensitized solar cells: A brief overview. Sol. Energy 85,
1172–1178 (2011).
28. Yella, A. et al. Porphyrin-Sensitized Solar Cells with Cobalt (II/III)–Based Redox Electrolyte Exceed 12
Percent Efficiency. Science 334, 629–634 (2011).
29. Chen, H.-Y. et al. Polymer solar cells with enhanced open-circuit voltage and efficiency. Nat. Photonics 3,
649–653 (2009).
30. O’Malley, K. M., Li, C.-Z., Yip, H.-L. & Jen, A. K.-Y. Enhanced Open-Circuit Voltage in High Performance
Polymer/Fullerene Bulk-Heterojunction Solar Cells by Cathode Modification with a C60 Surfactant. Adv.
Energy Mater. 2, 82–86 (2012).
31. Ip, A. H. et al. Hybrid passivated colloidal quantum dot solids. Nat. Nanotechnol. 7, 577–582 (2012).
32. Chen, S., Gong, X. G., Walsh, A. & Wei, S.-H. Defect physics of the kesterite thin-film solar cell absorber
Cu2ZnSnS4. Appl. Phys. Lett. 96, 21902 (2010).
33. Cao, Q. et al. Defects in Cu(In,Ga)Se2 Chalcopyrite Semiconductors: A Comparative Study of Material
Properties, Defect States, and Photovoltaic Performance. Adv. Energy Mater. 1, 845–853 (2011).
34. Kojima, A., Teshima, K., Shirai, Y. & Miyasaka, T. Organometal Halide Perovskites as Visible-Light
Sensitizers for Photovoltaic Cells. J. Am. Chem. Soc. 131, 6050–6051 (2009).
35. Jeon, N. J. et al. Solvent engineering for high-performance inorganic–organic hybrid perovskite solar cells. Nat.
Mater. 13, 897–903 (2014).
36. Goldschmidt, V. M. Die Gesetze der Krystallochemie. Naturwissenschaften 14, 477–485 (1926).
37. Peña, M. A. & Fierro, J. L. G. Chemical Structures and Performance of Perovskite Oxides. Chem. Rev. 101,
1981–2018 (2001).
38. Stoumpos, C. C., Malliakas, C. D. & Kanatzidis, M. G. Semiconducting Tin and Lead Iodide Perovskites with
Organic Cations: Phase Transitions, High Mobilities, and Near-Infrared Photoluminescent Properties. Inorg.
Chem. 52, 9019–9038 (2013).
39. Cheng, Z. & Lin, J. Layered organic–inorganic hybrid perovskites: structure, optical properties, film
preparation, patterning and templating engineering. CrystEngComm 12, 2646–2662 (2010).
40. Mitzi, D. B., Wang, S., Feild, C. A., Chess, C. A. & Guloy, A. M. Conducting Layered Organic-inorganic
Halides Containing <110>-Oriented Perovskite Sheets. Science 267, 1473–1476 (1995).
41. McGehee, M. D. Perovskite solar cells: Continuing to soar. Nat. Mater. 13, 845–846 (2014).
42. Grätzel, M. The light and shade of perovskite solar cells. Nat. Mater. 13, 838–842 (2014).
43. Green, M. A., Ho-Baillie, A. & Snaith, H. J. The emergence of perovskite solar cells. Nat. Photonics 8, 506–
514 (2014).
44. De Wolf, S. et al. Organometallic Halide Perovskites: Sharp Optical Absorption Edge and Its Relation to
Photovoltaic Performance. J. Phys. Chem. Lett. 5, 1035–1039 (2014).
45. Dong, Q. et al. Electron-hole diffusion lengths >175 μm in solution grown CH3NH3PbI3 single crystals.
Science aaa5760 (2015). doi:10.1126/science.aaa5760
46. Shi, D. et al. Low trap-state density and long carrier diffusion in organolead trihalide perovskite single crystals.
Science 347, 519–522 (2015).
47. Xing, G. et al. Long-Range Balanced Electron- and Hole-Transport Lengths in Organic-Inorganic
CH3NH3PbI3. Science 342, 344–347 (2013).
48. Buin, A. et al. Materials Processing Routes to Trap-Free Halide Perovskites. Nano Lett. (2014).
doi:10.1021/nl502612m
49. Stranks, S. D. et al. Electron-Hole Diffusion Lengths Exceeding 1 Micrometer in an Organometal Trihalide
Perovskite Absorber. Science 342, 341–344 (2013).
50. D’Innocenzo, V. et al. Excitons versus free charges in organo-lead tri-halide perovskites. Nat. Commun. 5,
(2014).
51. Lin, Q., Armin, A., Nagiri, R. C. R., Burn, P. L. & Meredith, P. Electro-optics of perovskite solar cells. Nat.
Photonics 9, 106–112 (2015).
52. Kim, H.-S. et al. Lead Iodide Perovskite Sensitized All-Solid-State Submicron Thin Film Mesoscopic Solar
Cell with Efficiency Exceeding 9%. Sci. Rep. 2, (2012).
53. Lee, M. M., Teuscher, J., Miyasaka, T., Murakami, T. N. & Snaith, H. J. Efficient Hybrid Solar Cells Based on
Meso-Superstructured Organometal Halide Perovskites. Science 338, 643–647 (2012).
54. Burschka, J. et al. Sequential deposition as a route to high-performance perovskite-sensitized solar cells. Nature
499, 316–319 (2013).
55. Docampo, P., Ball, J. M., Darwich, M., Eperon, G. E. & Snaith, H. J. Efficient organometal trihalide perovskite
planar-heterojunction solar cells on flexible polymer substrates. Nat. Commun. 4, (2013).
61
56. Malinkiewicz, O. et al. Perovskite solar cells employing organic charge-transport layers. Nat. Photonics 8, 128–
132 (2014).
57. Deschler, F. et al. High Photoluminescence Efficiency and Optically Pumped Lasing in Solution-Processed
Mixed Halide Perovskite Semiconductors. J. Phys. Chem. Lett. 5, 1421–1426 (2014).
58. Tan, Z.-K. et al. Bright light-emitting diodes based on organometal halide perovskite. Nat. Nanotechnol. 9,
687–692 (2014).
59. Habisreutinger, S. N. et al. Carbon Nanotube/Polymer Composites as a Highly Stable Hole Collection Layer in
Perovskite Solar Cells. Nano Lett. 14, 5561–5568 (2014).
60. Leijtens, T. et al. Overcoming ultraviolet light instability of sensitized TiO2 with meso-superstructured
organometal tri-halide perovskite solar cells. Nat. Commun. 4, (2013).
61. Ito, S., Tanaka, S., Manabe, K. & Nishino, H. Effects of Surface Blocking Layer of Sb2S3 on Nanocrystalline
TiO2 for CH3NH3PbI3 Perovskite Solar Cells. J. Phys. Chem. C 118, 16995–17000 (2014).
62. Li, W. et al. Montmorillonite as bifunctional buffer layer material for hybrid perovskite solar cells with
protection from corrosion and retarding recombination. J. Mater. Chem. A 2, 13587–13592 (2014).
63. Abate, A. et al. Lithium salts as ‘redox active’ p-type dopants for organic semiconductors and their impact in
solid-state dye-sensitized solar cells. Phys. Chem. Chem. Phys. 15, 2572–2579 (2013).
64. Bailie, C. D., Unger, E. L., Zakeeruddin, S. M., Grätzel, M. & McGehee, M. D. Melt-infiltration of spiro-
OMeTAD and thermal instability of solid-state dye-sensitized solar cells. Phys. Chem. Chem. Phys. PCCP 16,
4864–4870 (2014).
65. Snaith, H. J. et al. Anomalous Hysteresis in Perovskite Solar Cells. J. Phys. Chem. Lett. 5, 1511–1515 (2014).
66. Kim, H.-S. & Park, N.-G. Parameters Affecting I–V Hysteresis of CH3NH3PbI3 Perovskite Solar Cells: Effects
of Perovskite Crystal Size and Mesoporous TiO2 Layer. J. Phys. Chem. Lett. 5, 2927–2934 (2014).
67. Unger, E. L. et al. Hysteresis and transient behavior in current–voltage measurements of hybrid-perovskite
absorber solar cells. Energy Environ. Sci. 7, 3690–3698 (2014).
68. Abate, A. et al. Supramolecular Halogen Bond Passivation of Organic–Inorganic Halide Perovskite Solar Cells.
Nano Lett. 14, 3247–3254 (2014).
69. Agiorgousis, M. L., Sun, Y.-Y., Zeng, H. & Zhang, S. Strong Covalency-Induced Recombination Centers in
Perovskite Solar Cell Material CH 3 NH 3 PbI 3. J. Am. Chem. Soc. 136, 14570–14575 (2014).
70. Suarez, B. et al. Recombination Study of Combined Halides (Cl, Br, I) Perovskite Solar Cells. J. Phys. Chem.
Lett. 5, 1628–1635 (2014).
71. Tress, W. et al. Understanding the rate-dependent J-V hysteresis, slow time component, and aging in
CH3NH3PbI3 perovskite solar cells: the role of a compensated electric field. Energy Environ. Sci. 8, (2015).
72. Shao, Y., Xiao, Z., Bi, C., Yuan, Y. & Huang, J. Origin and elimination of photocurrent hysteresis by fullerene
passivation in CH3NH3PbI3 planar heterojunction solar cells. Nat. Commun. 5, (2014).
73. Quilettes, D. W. de et al. Impact of microstructure on local carrier lifetime in perovskite solar cells. Science
348, 683–686 (2015).
74. Mosconi, E., Amat, A., Nazeeruddin, M. K., Grätzel, M. & De Angelis, F. First-Principles Modeling of Mixed
Halide Organometal Perovskites for Photovoltaic Applications. J. Phys. Chem. C 117, 13902–13913 (2013).
75. Wang, Y. et al. Density functional theory analysis of structural and electronic properties of orthorhombic
perovskite CH 3 NH 3 PbI 3. Phys Chem Chem Phys 16, 1424–1429 (2014).
76. Lindblad, R. et al. Electronic Structure of TiO2/CH3NH3PbI3 Perovskite Solar Cell Interfaces. J. Phys. Chem.
Lett. 5, 648–653 (2014).
77. Haruyama, J., Sodeyama, K., Han, L. & Tateyama, Y. Termination Dependence of Tetragonal CH3NH3PbI3
Surfaces for Perovskite Solar Cells. J. Phys. Chem. Lett. 5, 2903–2909 (2014).
78. Liu, M., Johnston, M. B. & Snaith, H. J. Efficient planar heterojunction perovskite solar cells by vapour
deposition. Nature 501, 395–398 (2013).
79. Yin, W.-J., Shi, T. & Yan, Y. Unusual defect physics in CH3NH3PbI3 perovskite solar cell absorber. Appl.
Phys. Lett. 104, 63903 (2014).
80. Noh, J. H., Im, S. H., Heo, J. H., Mandal, T. N. & Seok, S. I. Chemical Management for Colorful, Efficient, and
Stable Inorganic–Organic Hybrid Nanostructured Solar Cells. Nano Lett. 13, 1764–1769 (2013).
81. Niu, G. et al. Study on the stability of CH3NH3PbI3 films and the effect of post-modification by aluminum
oxide in all-solid-state hybrid solar cells. J. Mater. Chem. A 2, 705–710 (2013).
82. Sun, S. et al. The origin of high efficiency in low-temperature solution-processable bilayer organometal halide
hybrid solar cells. Energy Environ. Sci. 7, 399–407 (2013).
83. Janotti, A. & Van de Walle, C. G. Native point defects in ZnO. Phys. Rev. B 76, 165202 (2007).
84. Walle, C. G. V. de & Neugebauer, J. First-principles calculations for defects and impurities: Applications to III-
nitrides. J. Appl. Phys. 95, 3851–3879 (2004).
62
85. Stampfl, C. & Van de Walle, C. G. Theoretical investigation of native defects, impurities, and complexes in
aluminum nitride. Phys. Rev. B 65, 155212 (2002).
86. Farrell, D. E. & Wolverton, C. First-principles study of point defects under varied chemical potentials in
Li${}_{4}$BN${}_{3}$H${}_{10}$. Phys. Rev. B 85, 174102 (2012).
87. Boukhvalov, D. W. & Solovyev, I. V. Defects of the crystal structure and Jahn-Teller distortion in
${\text{BiMnO}}_{3}$. Phys. Rev. B 82, 245101 (2010).
88. Derenzo, S. E. et al. Experimental and theoretical studies of donor–acceptor scintillation from PbI2. J. Lumin.
134, 28–34 (2013).
89. Kim, J., Lee, S.-H., Lee, J. H. & Hong, K.-H. The Role of Intrinsic Defects in Methylammonium Lead Iodide
Perovskite. J. Phys. Chem. Lett. 5, 1312–1317 (2014).
90. Lany, S. & Zunger, A. Anion vacancies as a source of persistent photoconductivity in II-VI and chalcopyrite
semiconductors. Phys. Rev. B 72, 35215 (2005).
91. Shockley, W. & Read, W. T. Statistics of the Recombinations of Holes and Electrons. Phys. Rev. 87, 835–842
(1952).
92. Knop, Osvald et al. Alkylammonium lead halides. Part 2. CH3NH3PbX3 (X = chlorine, bromine, iodine)
perovskites: cuboctahedral halide cages with isotropic cation reorientation. Can. J. Chem. 68, 412–422 (1990).
93. Zhang, H., Jang, J., Liu, W. & Talapin, D. V. Colloidal Nanocrystals with Inorganic Halide, Pseudohalide, and
Halometallate Ligands. ACS Nano 8, 7359–7369 (2014).
94. Zhitomirsky, D., Voznyy, O., Hoogland, S. & Sargent, E. H. Measuring Charge Carrier Diffusion in Coupled
Colloidal Quantum Dot Solids. ACS Nano 7, 5282–5290 (2013).
95. Liu, D. & Kelly, T. L. Perovskite solar cells with a planar heterojunction structure prepared using room-
temperature solution processing techniques. Nat. Photonics 8, 133–138 (2014).
96. Zhang, W. et al. Ultrasmooth organic–inorganic perovskite thin-film formation and crystallization for efficient
planar heterojunction solar cells. Nat. Commun. 6, 6142 (2015).
97. Wehrenfennig, C., Eperon, G. E., Johnston, M. B., Snaith, H. J. & Herz, L. M. High Charge Carrier Mobilities
and Lifetimes in Organolead Trihalide Perovskites. Adv. Mater. 26, 1584–1589 (2014).
98. You, J. et al. Low-Temperature Solution-Processed Perovskite Solar Cells with High Efficiency and Flexibility.
ACS Nano 8, 1674–1680 (2014).
99. Gonzalez-Pedro, V. et al. General Working Principles of CH3NH3PbX3 Perovskite Solar Cells. Nano Lett. 14,
888–893 (2014).
100. Ball, J. M., Lee, M. M., Hey, A. & Snaith, H. J. Low-temperature processed meso-superstructured to thin-film
perovskite solar cells. Energy Environ. Sci. 6, 1739–1743 (2013).
101. Du, M. H. Efficient carrier transport in halide perovskites: theoretical perspectives. J. Mater. Chem. A 2, 9091–
9098 (2014).
102. Xiao, M. et al. A Fast Deposition-Crystallization Procedure for Highly Efficient Lead Iodide Perovskite Thin-
Film Solar Cells. Angew. Chem. Int. Ed. 53, 9898–9903 (2014).
103. Zhou, H. et al. Interface engineering of highly efficient perovskite solar cells. Science 345, 542–546 (2014).
104. Guldi, D. M. & Prato, M. Excited-State Properties of C60 Fullerene Derivatives. Acc. Chem. Res. 33, 695–703
(2000).
105. Li, C.-Z. et al. Doping of Fullerenes via Anion-Induced Electron Transfer and Its Implication for Surfactant
Facilitated High Performance Polymer Solar Cells. Adv. Mater. 25, 4425–4430 (2013).
106. Weber, C. D., Bradley, C. & Lonergan, M. C. Solution phase n-doping of C60 and PCBM using
tetrabutylammonium fluoride. J. Mater. Chem. A 2, 303–307 (2013).
107. Popescu, L. M., Hof, P. van ’t, Sieval, A. B., Jonkman, H. T. & Hummelen, J. C. Thienyl analog of 1-(3-
methoxycarbonyl)propyl-1-phenyl-[6,6]-methanofullerene for bulk heterojunction photovoltaic devices in
combination with polythiophenes. Appl. Phys. Lett. 89, 213507 (2006).
108. Kumar, P. P. & Yashonath, S. Ionic conduction in the solid state. J. Chem. Sci. 118, 135–154 (2006).
109. Lehovec, K. Space‐Charge Layer and Distribution of Lattice Defects at the Surface of Ionic Crystals. J. Chem.
Phys. 21, 1123–1128 (1953).
110. Chua, L. O. Memristor-The missing circuit element. IEEE Trans. Circuit Theory 18, 507–519 (1971).
111. Strukov, D. B., Snider, G. S., Stewart, D. R. & Williams, R. S. The missing memristor found. Nature 453, 80–
83 (2008).
112. Xiao, Z. et al. Giant switchable photovoltaic effect in organometal trihalide perovskite devices. Nat. Mater.
advance online publication, (2014).
113. Guo, X. et al. Polymer solar cells with enhanced fill factors. Nat. Photonics 7, 825–833 (2013).
114. Liu, P. et al. Interfacial electronic structure at the CH3NH3PbI3/MoOx interface. Appl. Phys. Lett. 106, 193903
(2015).
63
115. Kaltenbrunner, M. et al. Flexible high power-per-weight perovskite solar cells with chromium oxide-metal
contacts for improved stability in air. Nat. Mater. 14, 1032–1039 (2015).
116. Kato, Y. et al. Silver Iodide Formation in Methyl Ammonium Lead Iodide Perovskite Solar Cells with Silver
Top Electrodes. Adv. Mater. Interfaces 2, n/a-n/a (2015).
117. Li, X. et al. Improved performance and stability of perovskite solar cells by crystal crosslinking with
alkylphosphonic acid ω-ammonium chlorides. Nat. Chem. 7, 703–711 (2015).
118. Leijtens, T. et al. Stability of Metal Halide Perovskite Solar Cells. Adv. Energy Mater. 5, n/a-n/a (2015).
119. Mei, A. et al. A hole-conductor–free, fully printable mesoscopic perovskite solar cell with high stability.
Science 345, 295–298 (2014).
120. Jung, M.-C., Raga, S. R., Ono, L. K. & Qi, Y. Substantial improvement of perovskite solar cells stability by
pinhole-free hole transport layer with doping engineering. Sci. Rep. 5, 9863 (2015).
121. Mailoa, J. P. et al. A 2-terminal perovskite/silicon multijunction solar cell enabled by a silicon tunnel junction.
Appl. Phys. Lett. 106, 121105 (2015).
122. Kranz, L. et al. High-Efficiency Polycrystalline Thin Film Tandem Solar Cells. J. Phys. Chem. Lett. 6, 2676–
2681 (2015).
123. Jeon, N. J. et al. Compositional engineering of perovskite materials for high-performance solar cells. Nature
517, 476–480 (2015).
124. Im, J.-H., Jang, I.-H., Pellet, N., Grätzel, M. & Park, N.-G. Growth of CH3NH3PbI3 cuboids with controlled
size for high-efficiency perovskite solar cells. Nat. Nanotechnol. 9, 927–932 (2014).
125. Heo, J. H. et al. Efficient inorganic-organic hybrid heterojunction solar cells containing perovskite compound
and polymeric hole conductors. Nat. Photonics 7, 486–491 (2013).
126. Yang, W. S. et al. High-performance photovoltaic perovskite layers fabricated through intramolecular
exchange. Science 348, 1234–1237 (2015).
127. Cappel, U. B., Daeneke, T. & Bach, U. Oxygen-Induced Doping of Spiro-MeOTAD in Solid-State Dye-
Sensitized Solar Cells and Its Impact on Device Performance. Nano Lett. 12, 4925–4931 (2012).
128. Nguyen, W. H., Bailie, C. D., Unger, E. L. & McGehee, M. D. Enhancing the Hole-Conductivity of Spiro-
OMeTAD without Oxygen or Lithium Salts by Using Spiro(TFSI)2 in Perovskite and Dye-Sensitized Solar
Cells. J. Am. Chem. Soc. 136, 10996–11001 (2014).
129. Hawash, Z., Ono, L. K., Raga, S. R., Lee, M. V. & Qi, Y. Air-Exposure Induced Dopant Redistribution and
Energy Level Shifts in Spin-Coated Spiro-MeOTAD Films. Chem. Mater. 27, 562–569 (2015).
130. Pattanasattayavong, P. et al. Electric field-induced hole transport in copper(I) thiocyanate (CuSCN) thin-films
processed from solution at room temperature. Chem. Commun. 49, 4154–4156 (2013).
131. Ito, S. et al. Carbon-Double-Bond-Free Printed Solar Cells from TiO2/CH3NH3PbI3/CuSCN/Au: Structural
Control and Photoaging Effects. ChemPhysChem 15, 1194–1200 (2014).
132. Müller, C. D. et al. Multi-colour organic light-emitting displays by solution processing. Nature 421, 829–833
(2003).
133. Müller, D. C. et al. Efficient Blue Organic Light-Emitting Diodes with Graded Hole-Transport Layers.
ChemPhysChem 1, 207–211 (2000).
134. Jhuo, H.-J. et al. Inverted perovskite solar cells with inserted cross-linked electron-blocking interlayers for
performance enhancement. J. Mater. Chem. A 3, (2015).
135. Deng, Y. et al. Scalable fabrication of efficient organolead trihalide perovskite solar cells with doctor-bladed
active layers. Energy Environ. Sci. 8, (2015).
136. Hayashi, N., Nishio, R. & TAKADA, S. Composition, film using the composition, charge transport layer,
organic electroluminescence device, and method for forming charge transport layer. (2012).
137. Chiang, T.-Y. et al. Functional p-Type, Polymerized Organic Electrode Interlayer in CH3NH3PbI3
Perovskite/Fullerene Planar Heterojunction Hybrid Solar Cells. ACS Appl. Mater. Interfaces 7, 24973–24981
(2015).
138. Lee, J.-H. & Kim, J.-J. Interfacial doping for efficient charge injection in organic semiconductors. Phys. Status
Solidi A 209, 1399–1413 (2012).
139. Kröger, M. et al. Role of the deep-lying electronic states of MoO3 in the enhancement of hole-injection in
organic thin films. Appl. Phys. Lett. 95, 123301 (2009).
140. Meyer, J. et al. Transition Metal Oxides for Organic Electronics: Energetics, Device Physics and Applications.
Adv. Mater. 24, 5408–5427 (2012).
141. Xu, J. et al. Perovskite-fullerene hybrid materials suppress hysteresis in planar diodes. Nat. Commun. 6, (2015).
142. Zhou, X. et al. Enhanced Hole Injection into Amorphous Hole-Transport Layers of Organic Light-Emitting
Diodes Using Controlled p-Type Doping. Adv. Funct. Mater. 11, 310–314 (2001).
64
143. Leem, D.-S. et al. Low driving voltage and high stability organic light-emitting diodes with rhenium oxide-
doped hole transporting layer. Appl. Phys. Lett. 91, 11113 (2007).
144. Matsushima, T., Kinoshita, Y. & Murata, H. Formation of Ohmic hole injection by inserting an ultrathin layer
of molybdenum trioxide between indium tin oxide and organic hole-transporting layers. Appl. Phys. Lett. 91,
253504 (2007).
145. Zhao, Y. et al. Transition metal oxides on organic semiconductors. Org. Electron. 15, 871–877 (2014).
146. Ng, T.-W., Lo, M.-F., Yang, Q.-D., Fung, M.-K. & Lee, C.-S. Near-Infrared Electric Power Generation
Through Sub-Energy-Gap Absorption in an Organic–Inorganic Composite. Adv. Funct. Mater. 22, 3035–3042
(2012).
147. Yoo, S.-J. et al. Formation of perfect ohmic contact at indium tin oxide/N,N′-di(naphthalene-1-yl)-N,N′-
diphenyl-benzidine interface using ReO3. Sci. Rep. 4, (2014).
148. Matsushima, T. et al. Interfacial charge transfer and charge generation in organic electronic devices. Org.
Electron. 12, 520–528 (2011).
149. Burgelman, M., Nollet, P. & Degrave, S. Modelling polycrystalline semiconductor solar cells. Thin Solid Films
361–362, 527–532 (2000).
150. Wang, X. et al. Tandem colloidal quantum dot solar cells employing a graded recombination layer. Nat.
Photonics 5, 480–484 (2011).
151. Li, X. et al. A vacuum flash–assisted solution process for high-efficiency large-area perovskite solar cells.
Science aaf8060 (2016). doi:10.1126/science.aaf8060
152. NREL. Research Cell Record Efficiency Chart. Available at:
http://www.nrel.gov/pv/assets/images/efficiency_chart.jpg. (Accessed: 4th November 2016)
65
Appendices
1 Methods
1.1 Anhydrate lead acetate protocol. 10 g lead (II) acetate trihydrate (Sigma-Aldrich,
99.99%) was dissolved in 10 ml of acetic acid anhydride at ~80˚C and distilled off at ~40
˚C under vacuum. The solid residue is anhydrate lead acetate.
1.2 Perovskite solution preparation. The dehydrated lead acetate (Pb(Ac)2) and
methylammonium iodide (MAI) (Dyesol, 99%+) were dissolved in DMF (N,N-
Dimethylformamide, Sigma-Aldrich, 99.9%) with the molar ratio 1:3 to form the
perovskite precursor solution. To obtain ultrathin films and thick films, we tune the
perovskite concentration between 0.2 M and 1 M.
1.3 Perovskite-PCBM hybrid solution preparation. PCBM (Nano-C, 99.5%) is mixed into
the perovskite solution. In typical procedure, the PCBM-perovskite weight ratios are
between 1:100 and 1:10. Specifically, PCBM can be dissolved into chlorobenzene first,
and then mixed with perovskite solution before spin coating. The solution is kept at 70˚C
before spinning. For low mixture ratio, the miscibility of mixture solution is good and can
be stabilized at room temperature; for high ratio mixtures approaching 1:10 and beyond,
the solution needs to be used quickly after mixing.
1.4 Planar perovskite film and device fabrication (Chapter 3 and 4). A thin TiO2 compact
layer was first formed on FTO substrate using magnetron sputtering (~50 nm, Kurt J.
Lesker, 99.9%) followed by a low-concentration TiCl4 treatment for interfacial contact
improvement: soak in 120 mM TiCl4 aqueous solution at 70˚C for 0.5 hour followed by
annealing at 500˚C for 0.5 hour. Perovskite-PCBM hybrid solid films were deposited on
pre-heated TiO2 substrate using spin-coating at 3000~5000 rpm for 60 seconds in a
nitrogen glovebox. During the spin-coating, the film turned to dark brown, implying that
the perovskite crystallization was almost done. The hybrid solid film was then heated for
10 minutes at 70˚C to remove the residual solvent. For a control planar heterojunction
device, pure perovskite solution is deposited on TiO2 substrate in the same way. No excess
acetate (Ac-) and methylammonium (MA-) were found in the final films. For bilayer
control devices, PCBM in chlorobenzene (~20 mg ml-1) was spin cast on a TiCl4-treated
TiO2 substrate and then annealed at 70˚C for 10 minutes before spin-coating the perovskite
on top. A thin PCBM layer (< 30 nm) between TiO2 and perovskite is formed. Hole
66
transfer layer was deposited by spin-coating of Spiro-OMeTAD (Borun Chemical, 99%+)
solution following the doping procedure reported below in Appendix 1.6. Top contact was
50 nm thermally evaporated gold through the shadow mask under 10-7 torr vacuum using
an Angstrom Engineering deposition system.
1.5 Planar perovskite device fabrication (Chapter 5). A thin TiO2 compact layer is first
formed on FTO substrates using atomic layer deposition (ALD) (~10 nm, Cambridge
Nanotech Savannah S100) using tetrakis-dimethyl-amido titanium and H2O as precursors.
A low-concentration TiCl4 treatment is used for interfacial improvement. The substrates
are soaked in TiCl4 aqueous solution (120 mM, 70˚C) for 0.5 hour and then annealed at
500˚C for 0.5 hour. PCBM ([6,6]-phenyl-C61-butyric acid methyl ester, Nano-C, 99.5%)
in chlorobenzene (~20 mg ml-1) is spin cast on the TiO2 substrates and then annealed at
70˚C for 10 minutes before spin-coating the perovskite on top. Anhydrate lead acetate
(Pb(C2H3O2)2). and methylammonium iodide (MAI) (Dyesol, 99%+) are dissolved in DMF
(N,N-Dimethylformamide, Sigma-Aldrich, 99.9%) with the molar ratio 1:3 to form the
perovskite precursor solution (1~1.3 M) and kept at 70˚C. Perovskite precursor is mixed
with 20 ul PCBM in chlorobenzene (30 mg ml-1) and deposited by spin-casting
(3000~5000 rpm for 60 seconds) on pre-heated TiO2-PCBM substrates in a nitrogen
glovebox. The film is annealed at 75˚C for 5 minutes and then 100˚C for 15 minutes. For
control devices using chemical doping, the hole transfer layer was deposited by spin-
coating the mixture solution of Spiro-MeOTAD (Borun Chemical, 99%+), dopant Li-TFSI
(Bis(trifluoromethane)sulfonimide lithium salt) and additive tBP (4-tert-butylpyridine)
following below Appendix 1.6. For the interface doping devices, the hole transport layer
VNPB (Lumtec, 95%+) in anhydrous toluene (3 mg ml-1) is spin cast (3000~4000 rpm,
30s) on perovskite film, followed by processing for the thermal crosslinking (120˚C for
20minutes and 150˚C for 10 minutes). The interface doping layer is 10 nm MoO3,
evaporated under 10-7 torr vacuum (Angstrom Engineering deposition system). After that,
the samples are kept at 40 ˚C in the evaporation chamber for 10 minutes. The top metal
contact is gold (50 nm) deposited through a shadow mask. Encapsulation is done using the
UV cured epoxy (Ossila) in conjunction with a glass coverslip.
1.6 Spiro-MeOTAD doping protocol. Spiro-MeOTAD was dissolved in chlorobenzene (63
mg ml-1). Then tBP (4-tert-butylpyridine) was added as additive ( 20 μl ml-1). Dopant Li-
67
TFSI (Bis(trifluoromethane)sulfonimide lithium salt) (170 mg ml-1 in acetonitrile) was
finally added into the prepared Spiro-MeOTAD solution (70 μl ml-1).
2 Characterizations
2.1 Steady-state photovoltaic performance and hysteresis-effect characterization. The
active area of devices is determined by an optical aperture (area 0.049 cm2) placed before
the device. The AM1.5 solar simulator (ScienceTech) is class A (<25% spectrum
mismatch) and the spectral mismatching factor was characterized using a Newport
calibrated reference Si solar cell. The spectral mismatching factor was used in every
reported performance. The illumination intensity on devices was calibrated using a Melles–
Griot power meter to be 1sun (100 mW cm− 2). The final accuracy of the solar-to-
electricity measurements was estimated to be ± 5%. Steady-state performance was
measured using a Keithley 2400 SourceMeter. A standard testing process is as follows:
first the steady-state open-circuit voltage VOC(t) is measured by fixing the current to zero;
then short-circuit current JSC(t) is measured by setting the voltage to zero; thirdly, the
forward- and reverse-scanning instantaneous J-V curves are measured with a scanning rate
of 0.2 V s-1 and the voltage of maximum power point (MPP) is determined. The J-V
voltage scanning range is 1.1~1.2 times the steady-state open-circuit voltage. The
hysteresis factor of J-V curves is quantified using Equation (5-1). Finally, the steady-state
power conversion efficiency (PCE(t)) is measured by setting the bias at the maximum
power point and tracking the output steady-state current for a certain duration. To avoid the
overestimation due to the hysteresis effect, the figure of merit of photovoltaic performance
is only determined by the steady-state efficiency.
2.2 External Quantum Efficiency (EQE) spectra is measured by aligning the cell to
monochromatic illumination (a 400W Xe lamp passing through a monochromator and
appropriate cut-off filters). The active area was defined by the optical aperture before the
cell, and the power was calibrated with UV-IR photodetectors (Newport 818-UV and
Newport 838-IR). A solar simulator at 1 sun intensity provided the light bias. The
monochromatic beam was chopped at 220Hz. The response of the cell was measured with
a pre-amplifier (Lakeshore) connected to a lock-in amplifier (Stanford Research 830) at
short circuit conditions.
2.3 Conductive atomic force microscope characterization (CAFM, Chapter 4). Scanning
probe microscopy experiments were carried out in a commercial ultrahigh-vacuum atomic
68
force microscope (UHV bean-deflection AFM, Omicron) using Cr/Pt-coated silicon
cantilevers (Budget Sensor, Multi75E-G). All the measurements were performed at a
background pressure of < 2 x 10–10 Torr after transferring the samples from ambient
without any additional treatment. Contact-mode AFM images and 2D current maps are
simultaneously obtained with the tip in contact with the surface (loading force ~1 nN)
applying fixed bias voltages. The I-V curves were acquired in the conductive AFM regime
from various locations of the sample surfaces applying a linear bias ramp with a rate of
~0.5 V s-1.
2.4 X-ray diffraction (XRD) measurements were performed at room temperature with a
Rigaku Miniflex 2-circle diffractometer operating in Bragg–Brentano scanning mode, with
angular resolution of 0.01 degrees and Cu-K radiation (0.154056 nm wavelength).
2.5 Ultraviolet photoelectron spectroscopy (UPS) was carried out using He I (21.22 eV)
photon lines from a discharge lamp.
2.6 XPS study on stoichiometry. X-ray photoelectron spectroscopy (XPS) is carried out using
a Thermo Scientific K-Alpha spectrometer. Core level spectra of Pb-4f, I-3d, O-1s, N-1s
and C-1s with a pass energy of 75 eV. The elemental composition was calculated based on
integrated counts of respective peaks. The curves were fitted using Gaussian functions with
1.5 eV FWHM. For comparison of different samples, all spectra were normalized to Pb
signal.
2.7 UV-Vis-IR absorption was measured using a PerkinElmer LAMBDA 950
Spectrophotometer.
2.8 Transient photoluminescence (PL) was carried out using TCSPC function of a HORIBA
FLuorolog-3 Spectrofluorometer. Samples were tested in N2 ambient.
