NSF Workshop:
Emerging Opportunities of Nanoscience to Energy Conversion and Storage
P. Braun, D. Carlson, V. Klimov, J. Michl, and A.J.
Nozik
Background
Solar
cells represent a potentially very important source of carbon-free energy.
However, in order to make them competitive with traditional energy sources, the
cost-to-efficiency ratio of photovoltaics must be reduced appreciably [1]. Cost
considerations have been a strong driver for the development of non-silicon
photovoltaic devices that are based on, e.g., polymers (plastic cells) [2] and
dye-synthesized porous metal oxides (GrŠtzel cells) [3]. Increases in
efficiency have typically relied upon evolutionary improvements of material
quality (for both Si and non-Si systems) and/or device engineering aspects
including, e.g., the use of tandem architectures. However, other approaches can
lead potentially to major increases in photovoltaic performance through the use
of new principles for conversion of solar energy into electricity. One such
approach involves the use of multiple exciton generation (MEG) from a single
photon that can greatly increase the photocurrent of solar cells [3, 4].
In the MEG process, absorption of a single photon produces
multiple electron-hole pairs (excitons) and hence the internal quantum
efficiency (QE) for converting photons into charge carriers becomes greater
than 100%. The phenomenon of
multiple electron-hole pair generation from single photons has been known in
bulk semiconductors since the 1950s [5] and has been explained in terms of
impact ionization [3-5]. However, because of restrictions imposed by energy and
momentum conservation and extremely fast intraband relaxation that competes
with impact ionization, carrier multiplication (CM) efficiencies in bulk
materials are quite low, particularly in the range of energies that are
relevant to solar-energy conversion. For example, in bulk Si p-n junctions, the
quantum yield reaches only 130% at photon energies of 5 eV—an energy
value that is outside of the solar spectrum (ref)
Carrier
multiplication in nanocrystals: Proof-of-principle experiments
As
was first suggested by Nozik [6], the CM efficiency might be enhanced in
nanoscale semiconductor particles [semiconductor nanocrystals (NC)] due to
reduced rates of intraband relaxation and enhanced Coulomb interactions.
Experimentally, high-efficiency CM in NCs was first demonstrated by Schaller
and Klimov [7] who utilized significant differences in the recombination
dynamics of single excitons and multiexcitons in order to detect CM and
quantify its efficiency. Single excitons decay via relatively slow radiative
recombination (in CdSe and PbSe NCs, e.g., it occurs with time constants of
tens and hundreds of nanoseconds, respectively [8 - 10]), while multiexcitons
decay on a much faster, picosecond timescale because of Auger recombination
[11-13]. As a result of this difference, the generation of multiexcitons can be
detected via a fast decay component in NC carrier population dynamics, while
the CM efficiency can be calculated from the ratio of the signal amplitude at
short times after excitation (before the Auger decay of the multiexcitons
occurs) to the amplitude of the slow single-exciton background (after
recombination of the multiexcitons is completed) (Fig. 1, inset). Using this approach the experiments in
ref. 7 demonstrated that absorption of a single photon by a PbSe NC could
produce up to 2.2 excitons, which translated into 220% internal QE. Nozik and
co-workers confirmed these results in similar studies of PbSe NCs, in which
they detected QEs up to 300% (3 excitons per absorbed photon) [14]. They also
demonstrated high-efficiency CM in NCs of
two other lead-salt compounds, PbS and PbTe, and they found that the
carrier multiplication process is essentially instantaneous (less than the
instrument response time of their systems which was 250 fs); because of this
the carrier multiplication effect was termed Òmultiple exciton generation
(MEG)Ó to reflect the fact that the CM process is instantaneous and is not
driven by the impact ionization mechanism present in bulk semiconductors.
Generality
of MEG and MEG quantum efficiency limits
These first
observations of highly efficient MEG in NCs raise questions regarding the
generality of this phenomenon to other materials and the potential limits on
photon-to-exciton conversion efficiency. One specific issue is: How important
are the unique features of lead salts (such as mirror symmetry between the
conduction and valence bands resulting from nearly identical electron and hole
masses) for obtaining highly efficient CM?
This issue
has been recently addressed by Klimov and co-workers in a comparative study of
CdSe and PbSe NCs [15]. These two types of NCs have distinctly different
electronic structures and carrier relaxation behavior. However, despite these
differences, both materials show comparable MEG efficiencies for similar excess
energies above the MEG threshold (Fig. 1), which is indicative of the
generality of this phenomenon to quantum-confined, semiconductor NCs. Further,
it is observed that CdSe NCs have a lower MEG activation threshold than PbSe
NCs: ~2.5Eg
vs. ~2.9Eg (Eg is the NC energy gap). This result can be explained in
terms of
simple
carrier effective-mass arguments and, specifically, by the difference in the
distribution of the photon excess energy (the energy in the excess of the
energy gap) between the conduction and the valence band [15]. The spectral
dependence of the MEG efficiency is almost linear above the MEG threshold with
a slope of ca. 110% per energy gap, which is nearly identical for both CdSe and
PbSe NCs (Fig. 1). In addition to this fast growth, the remarkable result of
these measurements was the observation that in the case of PbSe NCs, QE
approaches 700% (7 excitons per single absorbed photon) at photon energy of 7.8Eg, which corresponds to the ultimate
limit allowed by energy conservation
for this excitation wavelength (Fig. 2).
MEG
mechanism
CM in bulk materials has been
traditionally explained in terms of impact ionization, which is the inverse of
Auger recombination. However, the first experiments on carrier multiplication
in NCs [7, 14] as well as more recent studies in Ref. 16 indicate a significant
disparity between time constants of Auger recombination and MEG in
nanocrystalline materials. While the Auger decay is characterized by
tens-to-hundreds of picosecond time scales, MEG occurs with much faster time
constants. Specifically, direct measurements of the multiexciton population
build-up indicate that the MEG time constant is shorter than 50-to-200 fs [16],
strongly suggesting that it is an instantaneous process.
To
explain very fast generation of multiexcitons in the MEG process, Nozik, Efros,
and co-workers proposed a Òcoherent superpositionÓ model, in which the combined
single exciton/multiexciton system is photo-excited through its single-exciton
component and then experiences coherent oscillations between various resonant
exciton and multiexciton states [14]. This mechanism predicts an oscillatory
buildup of the multiexciton population if the dephasing of the combined
single-exciton/multiexciton wavefunction occurs primarily via its multiexcitonic
component.
An
alternative model for high-efficiency CM was recently proposed in Ref. 16. It
explains this effect in terms of direct (instantaneous) photo-generation of multiexcitons via virtual single-exciton states. This process also
relies on confinement-enhanced Coulomb coupling between single excitons and
multiexcitons and also takes advantage of a large spectral density of
high-energy single- and multiexciton resonances in nano-sized semiconductor
crystals. In this model, CM is a second-order process, which can be described
in the framework of second-order perturbation theory. This model produces
tens-of-percent CM efficiencies assuming the presence of only a single virtual
exciton resonance; and these values rapidly increase in the case of the multiple
resonances that exist in real NC systems. An interesting feature of this model
is that high-energy, single excitons are involved in the CM process as
ÒvirtualÓ but not ÒrealÓ states, which may explain why NC ionization (i.e.,
ejection of carriers from NCs) does not occur despite the use of high pump
photon energies.
Current
challenges
One current
challenge for practical applications of MEG is the extraction of charges from
NCs on time scales that are faster than Auger recombination and other non-radiative
and radiaitive channels. While the kinetics and the mechanism of charge
separation in NC systems are not well understood, several researchers have
demonstrated that charge separation at the interface between NCs and, e.g.,
surface adsorbed electron acceptors or porous TiO2 is highly
efficient and occurs with sub-picosecond to picosecond time constants, which
are significantly faster than those of Auger and radiative decay.
Other
challenges are associated with poor electronic conductivity of NC solids,
photo-corrosion of NCs, and the toxicity of materials that so far have been
utilized in studies of MEG. However, most of these problems are not conceptual
but rather are associated with material quality issues and design aspects of
NC-based photovoltaic structures.
1. ÒBasic
Research Needs for Solar Energy UtilizationÓ, Report of the U.S. DOE/Basic
Energy Sciences Workshop, April 18-21, 2005; available online at:
http://www.sc.doe.gov/bes/reports/list.html
2. G.
Yu, J. Gao, J. C. Himmelen, F. Wudl, and A. J. Heeger, Science 270, 1789 (1995).
3. M.
GrŠtzel, Nature 414, 338
(2001).
4. P.
T. Landsberg, H. Nussbaumer, and G. J. Willeke, Appl. Phys. 74,
1451 (1993).
5. S.
Kolodinski, J. H. Werner, T. Wittchen, and H. J. Queisser, Appl. Phys. Lett. 63, 2405 (1993).
6. S.
Koc, Czech J. Phys. B 7, 91 (1957).
7. A.
J. Nozik, Physica E 14, 115 (2002).
8. R.
D. Schaller, V. I. Klimov, Phys. Rev. Lett. 92, 186601 (2004).
9. S.
A. Crooker, T. Barrick, J. A. Hollingsworth, and V. I. Klimov, Appl. Phys. Lett. 82, 2793 (2003).
10. B. L. Wehrenberg, C. J. Wang,
and P. Guyot-Sionnest, J. Phys. Chem. B 106, 10634 (2002).
11. H. Du, C. Chen, R. Krishnan, T.
D. Krauss, J. M. Harbold, F. W. Wise, M. G. Thomas, and J. Silcox, Nano
Lett. 2, 1321 (2002).
12. V.
I. Klimov, A. A. Mikhailovsky, D. W. McBranch, C. A. Leatherdale, and M. G.
Bawendi, Science 287, 1011
(2000).
13. R.
D. Schaller, M. A. Petruska, and V. I. Klimov, J. Phys. Chem. B 107, 13765 (2003).
14. V. I.
Klimov, J. Phys. Chem. B 104,
6112 (2000).
15. R.
Ellingson, M. C. Beard, J. C. Johnson, P. Yu, O. I. Micic, A. J. Nozik, A.
Shabaev, and A. L. Efros, Nano Lett. 5,
865-71 (2005).
16. R. D.
Schaller, M. A. Petruska, and V. I. Klimov, Appl. Phys. Lett. 87 (December 12, 2005).
17. R.
D. Schaller, V. M. Agranovich, and V. I. Klimov, Nature Phys. 1, 189
(2005).
2.2 Plasmonics
and Photonic Crystals for Photon Management
Photonic crystals have the potential to strongly modulate,
focus, and concentrate light in ways that might increase the efficiency of PV
devices which may reduce the need for solar tracking and increase the
efficiency of PV structures.
Importantly, there are a number of rapid and low cost fabrication routes
for photonic crystals increasing the possibility of photonic crystals for large
area PV devices.
The field intensity within a 3-D photonic crystal is highly
nonuniform, and strongly dependent on the details of the structure (Figure
3). It may be possible to use the
dependence of these field intensities to spatially modulate the absorption of
light. We can imagine that through
appropriate theory and simulation, it may be possible to design a 3-D structure
that can modulate, focus, and enhance the absorption of light in a fashion that
would increase the efficiency of a PV device.
Figure 3. Lower left, simulated field
intensity map for light of a specific wavelength and incident direction
inside a selenium three-dimensional inverse opal structure. Background, SEM micrograph of the
photonic crystal.
It is possible that simple photonic crystals will not be
sufficient for modulating light.
It may be necessary to imbed features within the photonic crystal to
focus light to specific regions.
Such features can be directly written within a photonic band gap crystal
using a direct multiphoton writing [2].
Simple z-bend waveguide structures have been formed via this approach
(Figure 4), and more complex structures designed explicitly for a PV device
could also be formed. Success in
this area will require close coupling between theory and experiment.
Figure 4. Waveguide structure written through
multiphoton polymerization within a 3D self-assembled photonic crystal.
[1] P. V. Braun, R. W. Zehner, C. A. White, M. K. Weldon, C.
Kloc, S. S. Patel and P. Wiltzius, Adv. Mater. 2001, 13, 721.
[2] W. Lee, S. A. Pruzinsky, P. V. Braun, Adv. Mater. 2002, 14, 271.
Presently,
there are no materials that satisfy both the transparency and conductivity
requirements that are needed for electrodes in a parallel multijunction organic
photovoltaic cell. Transparent electrodes based on Indium Tin Oxide (ITO) thin
films have been widely used in single junction OPV cells because the
transparency window of ITO is well matched to the solar spectrum. However, the
DC conductivity in ITO is at least two orders of magnitude lower than that of a
metal such as Ag and its sheet-resistance contributes significantly to losses,
especially in large-area high-efficiency cells. One possibility is to replace
the ITO electrode with optically transparent nanopatterned metal films. This
will have the following benefits:
¤
The lower
sheet-resistance will reduce losses in large-area OPV cells.
¤
Replacement of the
brittle ITO metal-oxide will allow OPV cells to withstand larger deformations
associated with substrate bending, leading to more rugged cells.
¤
Since the ITO electrode
accounts for as much as 50% of the cost of an OPV cell, its replacement by a
thin metal film will result in substantial cost savings.
¤
In parallel
multijunction OPV cells where electrodes are required in between adjacent
organic subcells, sputtered ITO electrodes are of limited use since the
sputtering process inflicts damage on the underlying organic cells, resulting
in dramatically decreased efficiencies. Metal deposition, on the other hand, is
well suited for the fabrication of high efficiency cells.
It was recently discovered that metal
films patterned with arrays of nanoscale holes transmit light with efficiencies
much higher than expected from purely geometric considerations [T. W. Ebbessen,
H. J. Lezec, H. F. Ghaemi, T. Thio, and P. A. Wolff, Nature, 391, 667, 1998].
Since nanoscale holes have little effect on the in-plane DC conduction, such
films retain their low sheet-resistance, resulting in a unique combination of
optical transparency and metallic sheet-resistance. While
transmissivities of only a few percent were demonstrated until now [F. I.
Baida, and D. Van Labeke, Physical Review B, 62, 16100, 2003], simulations have
indicated vastly improved structures that transmit up to 80% of the incident
radiation. [H. Shin, P. Catrysse, and S. Fan, Physical Review B, 72 085436,
2005] We believe there are important opportunities in exploiting
nano-structured metal films for PV applications and especially for
parallel-connected multijunction OPV cells.
2.3 Singlet Fission
Singlet
fission in a molecular chromophore would produce two triplet excitons in two
neighboring chromophores from a single exciton generated by the absorption of a
single photon of sufficient energy by a molecule, and is a molecular analog of
multiple exciton generation (MEG) by photon absorption in a semiconductor. Each triplet exciton would then produce
an electron-hole pair, doubling the photocurrent. In principle, the two chromophores on which the triplet
excitations are created could have different structures, in which case the
usual term would not be singlet fission but quantum cutting.
The
use of singlet fission or quantum cutting
represents a possible disruptive dye-sensitized solar cell technology
that would surpass the 32 % power conversion efficiency restriction imposed on
a dye-sensitized Graetzel cell by the Shockley-Queisser limit and would replace
it by a 47% theoretical efficiency
limit for a singlet-fission sensitizer used in tandem with an ordinary sensitizer
(quantum dots or dyes) that would absorb photons of energies intermediate
between the lowest triplet and singlet states of the sensitizer (Fig. 5). For
electricity production, the best choice of excitation energies would be about
1.1 eV for the triplet and 2.2 eV for the singlet, and for solar hydrogen
splitting, the numbers would be somewhat different, depending on the specific
design used. The forbidden nature
of singlet-triplet absorption would permit photons of intermediate 1.1 - 2.2 eV
energies to pass the singlet-fission sensitizer zone and proceed to be absorbed
by the ordinary sensitizer.
Transparency in a large intermediate region endows molecules with an
advantage over semiconductors, in which the singlet and triplet levels are
nearly degenerate, such that there hardly are any photons of intermediate
energies.
Although
attractive in principle, the notion of using singlet fission for improving the
efficiency of excitonic solar cells is totally unproven. Singlet fission has been observed
accidentally on half-a-dozen organic chromophores, mostly in crystals or
polymers, and it is not obvious how to design molecules that undergo the
process efficiently. The inverse
process, triplet-triplet annihilation, is much better known, and tends to be
efficient and exothermic. It
generates a single excited species starting with two, and seems to occur
whenever diffusion brings two triplet molecules into close proximity.
Figure
5. Energy level diagram of
nanocrystalline solar cell based on a singlet fission chromophore (C1)
in optical series but parallel current flow with a conventional dye or
semiconductor chromophore (C2).
The
requirements that would need to be met in order for a singlet-splitting
sensitizer to be efficient are severe.
Its lowest excited singlet state needs to undergo singlet fission faster
than any other process that might compete, such as fluorescence, internal
conversion into the ground state, intersystem crossing into the triplet state,
or injection of an electron into the conduction band of the semiconductor on
which is adsorbed. In its turn,
electron injection by the triplet state must be faster than any other competing
process, such as intersystem crossing into the ground state or triplet-triplet
annihilation. Furthermore, the
singlet-fission sensitizer will have to meet the usual conditions that are
already well recognized for any ordinary excitonic solar cell sensitizer, such
as a high absorption coefficient in the ground state, good adsorption on the
semiconductor, fast kinetics of hole transfer from oxidized dye to a relay,
slow recombination, and stability upon irradiation.
Some
of the structural requirements for singlet-fission sensitizers are recognized
relatively easily. On an absolute
energy scale, the T1 state has to be positioned just above the
bottom of the conduction band of the semiconductor, to permit rapid electron
injection, and this can be fine-tuned by introduction of suitable
substituents. Intersystem crossing
rate can be reduced by minimizing spin-orbit coupling, and it will be best to
avoid heavy atoms in the molecular design. Internal conversion to the ground state will be slowed down
by molecular rigidity in the adsorbed state, and internal rotors and
low-frequency modes would be best avoided. Fast electron injection will be favored by strong coupling
to states of the semiconductor, and it is likely that chemisorption will be
preferable to physisorption, and that the triplet state should be of
intramolecular charge-transfer nature.
Injection from the initially excited singlet state could be discouraged
if it is of locally excited nature or if it is spatially separated from the
semiconductor surface by suitable nanoengineering.
Avoidance
of triplet-triplet annihilation process is of particular concern, given that is
normally very fast, and that the two triplets are created in immediate
vicinity. The coupling of two
triplets can in principle produce an overall singlet, a triplet, or a
quintet. Three of the possibly
competing processes are unlikely to represent serious problems. The formation of the ground state
singlets of the two chromophores, as well as the formation of one ground state
singlet and one triplet, would be highly exothermic and likely to be slow due
to the energy gap law. The
formation of a ground state singlet and a quintet is likely to be strongly
endothermic and negligible.
However, two of the processes are a real concern. They are the formation of a singlet
ground state plus an excited singlet state and the formation of a ground state
plus an upper triplet state. It
would appear best to make sure that they are both sufficiently endothermic and
therefore very slow at room temperature.
This would lead to a highly unusual requirement for the excitation energies
from the ground state: E(T2) $ E(S1) > 2E(T1). This is much more severe than the above
identified condition E(S1) $ 2E(T1) alone.
Meeting these requirements by molecular
design encounters problems whose solution requires fundamental knowledge that
is presently not available. The
leading two that are new and specific to singlet-fission sensitizers are (i)
how to adjust the energy levels into a highly unusual arrangement, (ii) how to
optimize the degree of coupling between chromophores to maximize the rate of
singlet fission. Perhaps a
sensitizer molecule that meets these in addition to all the other requirements
cannot be found. However, the
potential payoff is so high that it seems advisable to invest some effort into
the fundamental research needed. A
search for a solution to problem (i) will have to start with the use of
molecular quantum theory to identify structural classes likely to contain
molecules with the desired arrangement of energy levels, continue to numerical
calculations for a scan of a large number of representatives, identification of
the most promising candidates for synthesis, higher level calculations for
these candidates, their synthesis, and experimental determination of the
positions of their T1, T2, and S1 levels. A search for a solution to problem (ii)
will involve the development of a detailed theory of the singlet fission (and
triplet-triplet annihilation) process that will identify the optimal molecular
structure for connecting two chromophores, synthesis of dimers or higher
oligomers, and a search for evidence of singlet exciton fission and a
measurement of its efficiency.
Then, after it is understood how to design good sensitizers for singlet
fission, one will need to be concerned with the other issues identified above,
such as adjusting absolute level positions, optimizing the electron injection
speed for the triplet and minimizing it for the singlet, etc.
2.4 Hole
Conductor Relay in Dye-Senitized Solar Cells
The
iodide ion relay currently used in the best performing excitonic solar cells is
more strongly reducing than it needs to be and as a result, the voltage
produced by the cell is more than half a volt lower than it could be. This accounts for a large fraction of
the difference between the Shockley-Queisser limit and the 11% efficiency
actually attained. Much effort has
gone into a search for a relay with a better positioned redox potential, but
none have been found. The very
fast hole-transfer kinetics of the iodide ion, and the conversion of the
resulting iodine atom into the negatively charged triiodide ion, which
discourages recombination, are advantages that are very hard to beat.
Still,
the effort ought to be continued.
For engineering reasons, there would be considerable advantages in
replacing a liquid solution with a solid hole-conducting polymer, and rapid
kinetics for hole transfer from the sensitizer to the polymer could perhaps be
secured by a covalent link between them.
The discovery of new polymers with high hole conductivity whose redox
potential can be adjusted to a desired level should be a high priority.
2.5 PV
Industry Perspective
In the first few decades after the first silicon solar cell
was made at Bell Labs in 1954, the photovoltaic (PV) industry grew at a rate of
about 18% annually due largely to the use of PV systems in remote locations
(such as telecommunication relay sites).
However, in the last six years, the market growth has increased to ~ 35%
per year due to the rapid growth of grid-connected PV systems, which in turn
has been largely driven by government-supported programs in Japan and
Germany. Silicon technology has
dominated the PV industry since its inception, and in 2005 about 65% of all
solar cells were made from polycrystalline (or multicrystalline) silicon, ~ 24%
from monocrystalline silicon and ~ 4% from ribbon silicon. While conversion efficiencies as high
as 24.7% have been obtained in the laboratory for silicon solar cells, the best
efficiencies for commercial PV modules are in the range of 17 – 18% (the
efficiency limit for a silicon solar cell is ~ 29%). A number of companies are commercializing solar cells based
on other materials such as amorphous silicon, microcrystalline silicon, cadmium
telluride, copper-indium-gallium-diselenide (CIGS), gallium arsenide (and
related compounds) and dye-sensitized (Graetzel) solar cells. Thin film CIGS solar cells have been
fabricated with conversion efficiencies as high as 19.5% while efficiencies as
high as 39% have been demonstrated for GaInP/Ga(In)As/Ge triple-junction cells
operating at a concentration of 236 suns.
Thin film solar cells are being used in consumer products and in some
building-integrated applications, while PV concentrator systems are being
tested in grid-connected arrays located in high solar insolation areas. However, most industry experts believe
that crystalline silicon PV technology will continue to dominate the
terrestrial market for at least the next decade. It is likely that by 2020, silicon PV module efficiencies
will exceed 20% and module prices will fall below $1/Wp.
Photovoltaics
has the potential to become a major energy source within the next several
decades, but technology development must be accelerated in order for PV to
provide a viable alternative to fossil fuels (which are being rapidly depleted)
and to have a significant impact as soon as possible on carbon dioxide
emissions and climate change .
Nanotechnology is an attractive option that may lead to new low-cost,
high performance solar cells that could displace crystalline silicon solar
cells in the next decade, and thus lead to a more rapid deployment of
large-scale PV systems and the widespread availability of low-cost, clean PV
electricity.