ARTICLES

PUBLISHED: 30 JUNE 2017 | VOLUME: 2 | ARTICLE NUMBER: 17104

Pairing of near-ultraviolet solar cells with

electrochromic windows for smart management of
the solar spectrum
Nicholas C. Davy1, Melda Sezen-Edmonds1, Jia Gao1, Xin Lin2, Amy Liu1, Nan Yao3, Antoine Kahn2
and Yueh-Lin Loo1,4*

Current smart window technologies offer dynamic control of the optical transmission of the visible and near-infrared portions
of the solar spectrum to reduce lighting, heating and cooling needs in buildings and to improve occupant comfort. Solar
cells harvesting near-ultraviolet photons could satisfy the unmet need of powering such smart windows over the same
spatial footprint without competing for visible or infrared photons, and without the same aesthetic and design constraints.
Here, we report organic single-junction solar cells that selectively harvest near-ultraviolet photons, produce open-circuit
voltages eclipsing 1.6 V and exhibit scalability in power generation, with active layers (10 cm2 ) substantially larger than those
typical of demonstration organic solar cells (0.04–0.2 cm2 ). Integration of these solar cells with a low-cost, polymer-based
electrochromic window enables intelligent management of the solar spectrum, with near-ultraviolet photons powering the
regulation of visible and near-infrared photons for natural lighting and heating purposes.

H
eating, cooling, and lighting needs in buildings comprise
roughly 40% of the energy demand in the United States1 ,
with heating accounting for nearly twice the energy required
for cooling2,3 . Smart windows can regulate the transmission of
visible and near-infrared light, thereby offsetting building energy
needs by up to 40% (refs 4–6). Among smart window technologies,
electrochromic windows have garnered the most attention because
they offer dynamic control7,8 . Together with temperature and light
sensors, smart windows promise to work in concert with building
climate control and lighting systems to afford significant energy sav-
ings over passive or static alternatives, such as photochromic coat-
ings4 . Electrochromic windows require intermittent or continuous
utility for switching the electrochemical states of ECWs. These
cells produce power that increases linearly with active area, and
can drive similarly sized or larger ECWs over the same spatial
footprint with switching times on the order of seconds. This work
implicates near-UV solar cells as promising power sources for
emerging dual-band smart window technologies to independently
regulate visible and near-IR/infrared wavelengths over the same
spatial footprint (Fig. 1)4,7 . Beyond powering smart windows for
Near-UV Visible Near-IR
2.5
Spectral irradiance (W m−2 nm−1 )

power for operation, necessitating external wiring and complicating

installation2,4,9–11 . Low-cost, visibly transparent photovoltaic cells 2.0 For power
(TPVs) can provide local, renewable power to such electrochromic production For lighting
windows2 without altering building aesthetics or imposing fur- For heating
1.5
ther design constraints. Existing TPV technologies12–14 , however, do
not currently meet this application need because they target the 1.0
same near-infrared (near-IR) wavelengths for power generation that
smart windows aim to regulate dynamically. Selectively harvesting
0.5
near-ultraviolet (near-UV) light for powering smart window tech-
nologies avoids competition for the same spectral range.
0.0
Organic semiconductors, with inherently narrow and tunable 300 400 500 600 700 800 900 1,000
absorption, offer the opportunity for wavelength-specific light Wavelength (nm)
harvesting12,15 . Here, we report solar cells based on contorted
hexabenzocoronene (cHBC) derivatives that selectively harvest Figure 1 | Proposed management of the solar spectrum. ASTM G173
near-UV light, and demonstrate their pairing with polymer-based spectral irradiance at global tilt conditions (grey filled). The colour legend
electrochromic windows (ECWs) to modulate the transmission of (top panel) is scaled to the wavelength on the x-axis. We propose the
visible and near-IR light. These single-junction cHBC-based solar pairing of organic solar cells that absorb near-ultraviolet (near-UV) light to
cells exhibit photovoltages exceeding 1.6 V; the high photovoltages power ECWs that modulate the transmission of visible and near-infrared
of the near-UV cells (relative to typical solar cells) enhances their (near-IR) photons for natural lighting and heating purposes, respectively.

1
Department of Chemical and Biological Engineering, Princeton University, Princeton, New Jersey 08544, USA. 2 Department of Electrical Engineering,
Princeton University, Princeton, New Jersey 08544, USA. 3 Princeton Institute for the Science and Technology of Materials (PRISM), Princeton University,
Princeton, New Jersey 08544, USA. 4 Andlinger Center for Energy and the Environment, Princeton University, Princeton, New Jersey 08544, USA.
*e-mail: lloo@princeton.edu

NATURE ENERGY 2, 17104 (2017) | DOI: 10.1038/nenergy.2017.104 | www.nature.com/natureenergy 1

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
ARTICLES NATURE ENERGY

a Cl Cl a 100
X X
O O 90

Cl Cl 80

70

60
Cl Cl

%T
50
O O
X X
40
Cl Cl
D A1, X = H, 8Cl 30 D/A1
A2, X = Cl, 12Cl
D/A2
b 20
−2.6 eV D/C60
10
−3.1 eV 300 400 500 600 700 800
−3.4 eV Wavelength (nm)
Energy (eV)

−4.0 eV b
D/A1
active layer
−5.3 eV

−6.2 eV
−6.3 eV
−6.4 eV

c
20 D
Absorption coefficient (104 cm−1)

A1
A2
C60
15

10

0
300 400 500
Wavelength (nm)
Figure 2 | Chemical structures and thin-film optoelectronic properties.
a, Chemical structures of contorted hexabenzocoronene derivatives used as
electron donor (D) and electron acceptor (A1 or A2). b, Energy levels of the
highest occupied and lowest unoccupied molecular orbitals of the
constituents determined by ultraviolet photoelectron and inverse
photoemission spectroscopies, respectively. The energy levels of D were
reported previously20 and those of C60 were obtained from the literature25 .
c, Absorption coefficients of thin films of D (green), A1 (red), A2 (blue),
and C60 (black).
buildings and automobiles, solar cells that selectively harvest near-
UV light can benefit a broad range of other applications, such as
transparent displays, wearables, and sensors for internet-of-things.
They can also serve as power sources for vertically integrated
electrochemical devices, and as top cells for multi-junction
photovoltaic systems15–18 .
Optoelectronic characterization of cHBC derivatives
Harvesting near-UV light selectively using organic solar cells
requires a pair of wide-bandgap organic semiconductors with
complementary frontier orbital energy levels19 . Our near-UV solar
cells comprise organic semiconductors from a class of polyaromatic
hydrocarbons based on cHBC20 . Thin films of cHBC exhibit peak
absorption at 386 nm, with an absorption coefficient exceeding
200,000 cm−1 , and an electronic bandgap of 3.0 eV (ref. 20). cHBC
600 700 800
Figure 3 | Active-layer characterization. a, Thin-film transmittance spectra
of active layers comprising D/A1 (red), D/A2 (blue), and D/C60 (black).
b, Photograph of a D/A1 active-layer film on glass.
is thus an excellent starting point for subsequent functionalization
to create derivatives with favourable properties for harvesting near-
UV light in solar cells. The synthesis of these cHBC derivatives
is facile, uses widely available precursors, and has produced a
library of chemically stable molecules with varied optoelectronic
properties20–22 . The chemical structures of the electron donor
(D) and the two halogenated cHBC electron acceptors21
(A1 = 8Cl-cHBC, A2 = 12Cl-cHBC) used in our solar cells
are shown in Fig. 2a. These molecules can be handled in ambient
conditions and deposited as thin films using vacuum thermal
evaporation23,24 , a deposition technique already demonstrated to
produce organic electronic devices commercially18 , due to their
thermal stability (Supplementary Fig. 1). We evaluated the thin-film
energetics of A1 and A2 using ultraviolet photoelectron (UPS)
and inverse photoemission (IPES) spectroscopies (Supplementary
Fig. 2). The frontier orbital energy levels of A1 and A2 are shown
in Fig. 2b along with those of D and C60 obtained from the
literature20,25 . Compared to the lowest unoccupied molecular orbital
(LUMO) energy level of C60 , those of A1 and A2 are raised towards
the vacuum level by 0.9 eV and 0.6 eV, respectively. The energy-level
alignment of D/A1 and D/A2 pairs should thus allow for solar
cells that harvest near-UV photons at energies substantially higher

2 NATURE ENERGY 2, 17104 (2017) | DOI: 10.1038/nenergy.2017.104 | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE ENERGY ARTICLES
a 1.0 b 1.8
0.5
1.6

Current density (mA cm−2)

0.0 D/A1
−0.5
1.4
D/A2
−1.0

Voc (V)
−1.5 1.2

−2.0
1.0
−2.5 D/A1
D/A2 D/C60
−3.0 D/C60 0.8
−3.5
−0.5 0.0 0.5 1.0 1.5 2.0 −4.2 −4.0 −3.8 −3.6 −3.4 −3.2 −3.0
Voltage (V) LUMO (V)

c 40 d

30 Al
BCP
D/A1 A2
Counts

20
D
MoO3
10
Si 30 nm

0
1.60 1.61 1.62 1.63 1.64 1.65
Voc (V)

e 2.0 f 16
D/A1 D/A1
D/A2 D/A2
D/C60 D/C60
Spectral power response (%)
Power density (mW cm−2)

1.5 12

1.0 8

0.5 4

0.0 0
0.0 0.5 1.0 1.5 300 400 500 600 700
Voltage (V) Wavelength (nm)

Figure 4 | Solar cell characterization. a, J–V characteristics of D/A1 (red), D/A2 (blue), and D/C60 (black) planar-heterojunction solar cells, each having
an active area of 0.18 cm2 . Cells were measured under 100 mW cm−2 AM1.5G simulated solar illumination. b, Open-circuit voltage as a function of the
LUMO energy level of the electron acceptor used in each cell type. c, Histogram of V oc across 57 D/A1 solar cells. d, A high-resolution TEM image of the
cross-section of a D/A2 cell on silicon with interfaces marked in red. e,f, Power–voltage dependence (e) and spectral power response (f) of the solar cells.
The spectral power response is the product of spectral responsivity and operating voltage.

than those accessed by D/C60 cells. The optical responses of thin
films of the absorbers are shown in Fig. 2c. Thin films of D,
A1, and A2 exhibit maximum absorption coefficients exceeding
200,000 cm−1 at 372, 403, and 409 nm, respectively. Thus, A1 and
A2 films both complement the optical response of films of D at the
short-wavelength edge of the visible spectrum. With an absorption
coefficient of 131,000 cm−1 , the maximum near-UV absorption by
C60 thin films is not only lower, it occurs at 345 nm where the solar
flux is almost three times weaker than at 400 nm. The absorption
of C60 is also broader and extends into the visible, with an average
absorption coefficient of 45,500 cm−1 at 400–550 nm, and mild
absorption to 700 nm. A1 and A2 are thus stronger absorbers, and
each exhibits more selective absorption of near-UV light than C60 .
400–420 nm light is strongly absorbed by A1 and A2 films.
These wavelengths are barely detectable by the human eye but
represent 36% of the solar flux that is available between 300–420 nm.
Figure 3a shows the transmittance of each active layer. The D/A1
and D/A2 active layers absorb 52 and 56% of light between 300
and 420 nm, respectively, while transmitting 83% and 81% of visible
light. Figure 3b illustrates the transparency of a 25 cm2 D/A1 active
layer on glass.
Performance characterization of near-UV solar cells
Planar-heterojunction solar cells comprising active layers of D with
A1, A2, or C60 were optimized and tested under 1 sun AM1.5G
illumination. The J –V characteristics of the cells are shown in
Fig. 4a and tabulated in Supplementary Table 1. The open-circuit
voltages (Voc ) of D/A1, D/A2, and D/C60 cells are 1.63, 1.46, and
0.82 V, respectively. D/A1 and D/A2 cells both exhibit Voc s that are
higher than any previously reported single-junction solar cell, with
the Voc of D/A1 cells holding the record for any single-junction thin-
film PV technology26,27 . The fact that the Voc of the D/A1 cell is

NATURE ENERGY 2, 17104 (2017) | DOI: 10.1038/nenergy.2017.104 | www.nature.com/natureenergy 3

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
ARTICLES NATURE ENERGY

a 2.0 previous solar cells with Voc s exceeding 1.3 V exhibit fill factors of
0.24 cm2
30–45%, indicating recombination as a bottleneck to operation in
1.5
0.48 cm2 such devices26,30,31 .
1.0 0.96 cm2 Figure 4c details the Voc values recorded for all D/A1 cells
Current density (mA cm−2)

2.22 cm2 tested in this study. The judicious selection of conformationally

0.5 non-planar cHBC derivatives favours the formation of amorphous
pinhole-free films20 , and enables 100% yield of the 57 cells fabri-
0.0 cated. The high-resolution cross-sectional micrograph in Fig. 4d
confirms the amorphous nature of the donor and acceptor layers,
−0.5
and reveals the preservation of a pristine donor–acceptor interface.
−1.0 Reduced visible light absorption by A1 and A2 relative to C60
increases active-layer transparency at the expense of photocurrent;
−1.5 the short-circuit current density (Jsc ) of these cells are almost half
that of the D/C60 cell (Fig. 4a). At 1.3–1.5 mW cm−2 , however, the
−2.0 power densities of D/A1 and D/A2 cells (Fig. 4e) are comparable
−0.5 0.0 0.5 1.0 1.5 2.0
Voltage (V) to that of the D/C60 cell because of their high Voc and fill factor.
With photovoltages at the maximum-power points of 1.15 V and
1.28 V, D/A1 and D/A2 cells produce power densities comparable to
b
1.2
the most transparent near-IR-absorbing organic solar cells12,32 , but
our cells deliver this power at almost twice the photovoltage while
absorbing significantly less solar flux.
1.0
Power density (mW cm−2)

At 100 mW cm−2 AM1.5G illumination, the metrics in Fig. 4a

translate to power-conversion efficiencies (PCEs) of 1.3–1.5%.
0.8
By targeting the near-UV, our cells neglect approximately 93%
of spectral irradiation; the standard metric of PCE is thus not
0.6 informative in quantifying their targeted performance. We instead
define a wavelength-specific efficiency as the product of the spectral
0.4 0.24 cm2 responsivity (Supplementary Fig. 3A) and operating voltage (Fig. 4e)
0.48 cm2 for each type of cell, which we refer to as the spectral power
0.2 0.96 cm2 responsivity (SPR) in Fig. 4f. Conversion of highly energetic near-
2.22 cm2
UV photons to free charge carriers is known to be very inefficient
0.0 due to thermalization losses. For reference, the SPR at 400 nm of
0.0 0.5 1.0 1.5 2.0
the most efficient crystalline silicon solar cells is less than 19%
Voltage (V) (ref. 33). The SPR at 400 nm of cells comprising D/A1 or D/A2 is
c 3.5 12.4 and 14.0%, respectively. With an external quantum efficiency
(Supplementary Fig. 3B) approximately a third that of a champion
3.0 silicon solar cell33 , the SPR of D/A1 and D/A2 cells at 400 nm
remains competitive because their photovoltages are approximately
2.5 a factor of two higher than that of the silicon cell. D/C60 cells have
Photocurrent (mA)

an SPR of 5.9% at 400 nm. Thus, D/A1 and D/A2 cells are twice as
2.0 efficient at harvesting near-UV photons as D/C60 cells.

1.5 Scalability and stability of near-UV solar cells

1.0
0.5
0.0
0.0 0.5 1.0 1.5
Area (cm2)
Figure 5 | Area scaling of solar cells. a–c, J–V characteristics (a),
power-density profiles (b) and power-area dependence (c) of 0.24 cm2
(black), 0.48 cm2 (blue), 0.96 cm2 (red), and 2.22 cm2 (green) D/A1
planar-heterojunction cells fabricated on ITO-coated glass. The dashed line
is a fit of the data with the intercept fixed at the origin, with its linearity
demonstrating quantitative scaling of power generation with active area.
double that of the D/C60 cell is a manifestation of the raised LUMO
energy level of A1 relative to that of C60 . Figure 4b reveals a linear
correlation between Voc and the LUMO energy level of the acceptor.
The 0.81 and 0.64 V difference in Voc between D/A1 and D/A2 cells
relative to D/C60 cells, respectively, is accordingly consistent with the
0.9 and 0.6 eV difference in the LUMO energy levels of A1 and A2
relative to that of C60 . Importantly, the fill factors of the D/A1 and
D/A2 cells are 58 and 60%, respectively. With a few exceptions28,29 ,
Figure 5a,b shows the J –V characteristics and power-density pro-
files of D/A1 planar-heterojunction cells with increasing active areas
of 0.24 cm2 , 0.48 cm2 , 0.96 cm2 and 2.22 cm2 . The performance of
the cells is invariant despite an active-area increase by an order
of magnitude. The cells produce a photocurrent that scales with
area. At constant photovoltage and fill factor, power production
2.0 2.5
thus increases linearly with active area for these cells (Fig. 5c).
Comparatively, near-IR-absorbing organic solar cells exhibit sig-
nificant reductions in performance with increasing areas34 . In the
case of a high-efficiency low-bandgap polymer bulk-heterojunction
solar cell, increasing the active area from 0.27 cm2 to 2 cm2 resulted
in a 20% reduction in fill factor and a 24% loss in power den-
sity, even after significant device optimization34 . We attribute the
scalability of our cells to two factors. First, high-voltage power
has lower resistive losses compared to high-current power because
the total resistive power loss per unit area is proportional to the
square of the current density at the maximum-power point35 . Our
choice to target near-UV light, as opposed to near-IR light, results
in cells that output high-voltage/low-current power. Second, the
active layers in our cells are amorphous and inherently free of
pinholes20 . Contrary to cells that use polycrystalline active layers36 ,
this feature has enabled us to scale up the active area with the same
single-junction architecture while mitigating the shorting of cells.

4 NATURE ENERGY 2, 17104 (2017) | DOI: 10.1038/nenergy.2017.104 | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE ENERGY
a b
1 mm
c
200 µm
Figure 6 | Large-area solar cell with an alternative electrode. a–c, Photograph (a) and optical micrographs (b,c) of D/A1 active layer prepared on a 10 cm2
ITO/Ag-grid/PEDOT:PSS bottom electrode. d, J–V characteristic of 1 cm2 (black) and 10 cm2 (red) D/A1 cells fabricated on an ITO/PEDOT:PSS transport
layer and an ITO/Ag-grid/PEDOT:PSS transport layer, respectively. Cells were measured under 100 mW cm−2 AM1.5G simulated solar illumination.
Although lower charge-carrier mobilities in amorphous thin films
can be a concern23 , charge transport appears not to be the bottleneck
in these planar-heterojunction cells given their very thin (10–20 nm)
active layers37 .
Leveraging the ability to scale D/A1 films, we fabricated a 10 cm2
D/A1 cell on an indium tin oxide (ITO)-coated glass substrate. A
30-nm-thick honeycomb-inspired Ag-grid38 , shown in Fig. 6a–c,
that is 97% transparent between 300–800 nm (Supplementary Fig. 4)
was added to decrease the series resistance associated with ITO
carrying current over extended distances. The J –V characteristics
of this 10 cm2 D/A1 cell are shown in Fig. 6d along with a similarly
constructed 1 cm2 reference cell. We observe comparable perfor-
mance when the cell area is increased from 1 cm2 to 10 cm2 . The
fact that power production is scalable to 10 cm2 with an alternative
electrode is further testament that D/A1 active layers are pinhole-
free and amenable to large-area power production without sacri-
ficing performance metrics. The vast majority of publications on
organic solar cells report performance based on active areas that are
0.04–0.2 cm2 (ref. 39). However, the building blocks of organic solar
modules are solar cells having active areas of 5–10 cm2 (refs 40–42).
Developing organic solar cells that minimize power loss when the
active area is scaled up to sizes relevant for integration into mod-
ules has been investigated in numerous papers and texts40–42 . This
manuscript represents the first study that quantifies performance
scalability of near-UV solar cells, and the first report of organic solar
cells having properties inherently amenable to such scale up.
We subjected encapsulated D/A1 cells to >150 h of continuous
illumination to probe their stability under direct sunlight
(Supplementary Fig. 5). Importantly, the Voc is invariant throughout
this experiment. Because the frontier orbital energy levels of cHBC
acceptors are determined by the extent of halogenation21 , the Voc
invariance in D/A1 cells negates the possibility of photo-induced
chlorine loss from A1, especially at the D/A1 interface. The Jsc
and fill factor suffer an initial decrease, resulting in a power-
density reduction of 27% over the initial 30 h, beyond which the
device characteristics are stable. This burn-in behaviour is typical;
comparable reductions in performance metrics have been reported
in low-bandgap polymer solar cells, even when the UV portion of
the solar spectrum was filtered43–45 . As another example, similarly
constructed planar-heterojunction solar cells comprising boron
subphthalocyanine chloride (SubPc) and C60 , an archetypical
donor/acceptor pair, exhibit a burn-in power loss of 28% after 10 h
of illumination at 1 sun AM1.5G (ref. 46). Thus, despite harvesting
high-energy near-UV photons, D/A1 cells have comparable stability
to other promising organic solar cell technologies15,47 . Extending
the lifetime of organic solar cells is an ongoing area of research
NATURE ENERGY 2, 17104 (2017) | DOI: 10.1038/nenergy.2017.104 | www.nature.com/natureenergy
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
ARTICLES
d
1.0
0.5
Current density (mA cm−2)
0.0
−0.5
−1.0
−1.5 10 cm2
1.0 cm2
−2.0
−0.5 0.0 0.5 1.0 1.5 2.0
Voltage (V)
that is commonly embraced in the organic electronics community;
efforts span the design of new active materials, electrodes and
buffer layers, optimization of device architecture, and introduction
of new encapsulation technologies47 .
Self-powered electrochromic windows
Near-UV solar cells offer the opportunity to integrate power gen-
eration with ECWs that modulate sunlight over the same foot-
print without competing for the same spectral range. As a proof of
concept, we demonstrate the side-by-side48 operation of a 2.25 cm2
ECW using the high-voltage power produced by a 1.38 cm2 D/A1
solar cell (Supplementary Video 1, set-up shown in Supplementary
Fig. 6). Characterization of the operation of an ECW powered by
a D/A1 cell is presented in Fig. 7. Figure 8 offers the optical trans-
mittance of a vertically aligned ECW and PV stack. Compared to
ECWs previously powered with near-IR-absorbing solar cells2,11,49 ,
we demonstrate power generation with near-UV light, leaving near-
IR—in addition to the visible wavelengths—to be regulated by the
ECW for heating and lighting purposes, respectively1,7 .
Figure 7a shows a scheme of the ECW comprising a primary
electrochromic layer of polyaniline that is doped with poly(2-
acrylamido-2-methyl-1-propane-sulfonic acid) (PANI-PAAMPSA)
(refs 50–54), a poly(3,4–ethylenedioxythiophene):poly(styrene-
sulfonate) (PEDOT:PSS) counter electrode, and a gel electrolyte
composed of an aprotic ionic liquid, 1-ethyl-3-methylimidazolium
bis(trifluoromethanesulfonyl)imide (EMIM TFSI) with 5 wt%
fumed silica55 . PEDOT:PSS and PANI-PAAMPSA are bleached
when the former is oxidized and the latter is reduced
(Supplementary Fig. 7A)50,56 . Similarly, both polymers are coloured
when PEDOT:PSS is reduced and PANI-PAAMPSA is oxidized
(Supplementary Fig. 7B). Thus, the complementarity of PEDOT:PSS
and PANI-PAAMPSA enhances the optical contrast of the resulting
ECW56 . Applying a negative or positive bias from a single-
junction D/A1 cell between the PANI-PAAMPSA and PEDOT:PSS
electrodes drives the ECW toward its bleached or coloured
transmittance states, respectively, as illustrated by Fig. 7b. The use
of solution-processable electrochromic materials allows for tuning
of transmittance by varying the thickness of each polymer film at the
outset (Supplementary Fig. 8), with thinner electrochromic layers
resulting in higher bleached-state transmittance. For any given
PANI-PAAMPSA and PEDOT:PSS electrode thickness, the optical
transmittance contrast is further determined by the applied voltages
(Supplementary Fig. 9). In Fig. 7c,d we compare the transmittance
spectra in the bleached and coloured states achievable in an ECW
having PANI-PAAMPSA and PEDOT:PSS of thicknesses 370
and 355 nm, respectively, when powered by operating voltages
5
ARTICLES NATURE ENERGY

a b

+ D/A1
bias

ITO/gla
ss
PANI-P
AAMPS
A
Ionogel −D/A1
1 mm
bias 1 mm
PEDOT:P
SS
ITO/gla
ss

c 70 d 70

−1.2 V
56 56 −0.6 V

42 42

%T
%T

28 28
+0.6 V
+1.2 V

14 14

0 0
300 400 500 600 700 800 900 1,000 300 400 500 600 700 800 900 1,000
Wavelength (nm) Wavelength (nm)

e 70 f 70 6

−1.2 V No bias −1.2 V

60
4
56
50
2
%T (λ = 650 nm)

%T (λ = 650 nm)

42

J (mA cm−2)
40
0
30
28
−2
20
14
10 −4

+1.2 V +1.2 V
0 0 −6
0 2,000 4,000 6,000 8,000 10,000 0 100 200 300 400
Time (s) Time (s)

Figure 7 | Electrochromic window (ECW) characterization. a, Schematic representation of the ECW stack powered by a D/A1 cell in a side-by-side
geometry. The working electrode is composed of PANI-PAAMPSA deposited on ITO/glass, the counter electrode is composed of PEDOT:PSS deposited on
ITO/glass, and the gel electrolyte is EMIM TFSI with 5 wt% fumed silica. b, Photograph of the coloured and bleached states of the ECW powered by a D/A1
solar cell. c, Optical transmittance of the coloured and bleached states of the ECW applied voltages of ±1.2 V. d, Optical transmittance of the coloured and
bleached states of the ECW at applied voltages of ±0.6 V. e, Stability of the transmission contrast of the ECW when cycled between ±1.2 V. f, Kinetics of a
single coloration/bleaching cycle at 650 nm for an applied voltage of ±1.2 V.

characteristic of our solar cells (±1.2 V) and when powered by
operating voltages characteristic of near-IR solar cells (±0.6 V)2,12 .
Having 1.2 V instead of 0.6 V to drive ECW switching leads to a
substantial increase in the transmittance contrast at 650 nm (from
17% to 42%). Our near-UV solar cells are thus more effective at
powering ECWs that regulate visible light than near-IR solar cells.
While one could connect multiple near-IR solar cells in series or use
a tandem cell to access higher operating voltages17 , our near-UV
solar cells obviate the need for such fabrication complexity, allowing
for vertically integrated designs that directly connect single-
junction solar cell and ECW components. Additionally, for ECWs
that regulate near-IR light, near-UV cells are the only appropriate
power source to avoid competition for the same spectral range.
Figure 7e shows the transmission response of the ECW during
repeated potential cycling. The optical transmittance switches stably
and reversibly when the applied bias between electrodes is pulsed
at either positive or negative 1.2 V. When the applied potential is
removed (see Fig. 7e; between 5,000 and 6,000 s), a slow decay
in transmittance from 58% to 50% occurs over 10 min. The
ECW quickly recovers its stable maximum transmittance with a
subsequent pulse of −1.2 V. The kinetics of a single colouring and
bleaching cycle are shown in Fig. 7f, along with the instantaneous
current draw. Within 40 s of the applied potential, the transmittance
of the ECW reaches 95% of the saturated values corresponding to
its bleached and coloured states. Faster ECW switching speeds, if
necessary, can be achieved by replacing the electrochromic polymers

6 NATURE ENERGY 2, 17104 (2017) | DOI: 10.1038/nenergy.2017.104 | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE ENERGY ARTICLES
a Bleaching Colouring coloured and bleached states every hour, as well as the need to
circuit circuit maintain the transmittance while the ECW is in these respective
+
states, the total energy needed by the ECW is approximately
406 mJ cm−2 per 24 h. A D/A1 solar cell thus produces an order of
ITO/gla magnitude more energy in one hour than what is required to operate
ss −
PANI ele the ECW for 24 h. With an energy storage component, our solar-
− Gel elec ctrode
trolyte powered ECW should be able to operate without external power
PEDOT
AI electrod despite solar intermittency and in a wide variety of weather and
ITO/gla e
ss insolation conditions.
+ PV layers AI
ITO/gla
ss
Challenges
Longer-term testing of PV and ECW components under different
b 80 illumination and temperature conditions is needed to inform the
Background design of and extend the lifetime of next-generation devices.
Like other electrochromic technologies11 , the ECW reported here
60 achieves its coloured state using absorbing layers, necessitating
future study of heat management and optimization for combined
device thermal stability. In this regard, smart windows that reflect
−1.2 V light may improve thermal stability and aid heat management of the
40
%T

integrated stack.
Replacement of ITO electrodes with inexpensive alternatives
comprising earth-abundant materials will also be needed for low-
20
cost and sustainable access to smart window technologies. A further
+1.2 V challenge stems from our thin active layers, which necessitates
alternative electrode materials to be ultra-smooth. We speculate that
0 electrode conductivity may be less vital for sufficient near-UV solar
300 400 500 600 700 800 900 1,000
cell function, given the low currents produced by such cells, and the
Wavelength (nm)
conservative energy needs of our ECW. In comparison, electrode
c 80 transmittance in the visible and colour neutrality are probably stiffer
constraints for such smart window applications.

−1.2 V Conclusion
60
We demonstrate the complementary pairing of near-UV solar
cells with a low-cost, single-band ECW option based on water-
40
dispersible conducting polymers. The realization of solar cells that
%T

selectively harvest near-UV light necessitates the use of organic

semiconductor pairs in which each has a wide bandgap. Leveraging
+1.2 V our initial success and new machine-learning techniques58 , efforts
20
are ongoing to predict, and design, synthesize and characterize
accordingly promising organic semiconductors for solar cells
with improved wavelength selectivity and enhanced spectral
0 responsivity. This exciting technological advance also raises new
300 400 500 600 700 800 900 1,000
Wavelength (nm)
scientific questions about the fundamental operational limits of
devices that incorporate wide-bandgap organic semiconductor
Figure 8 | Combined ECW and PV stack optical transmittance. pairs. Simultaneously, future progress is being enabled by time-
a, Combined stack geometry for ECW and D/A1 cell, with the positive (+) resolved spectroscopic studies that elucidate exciton dynamics
and negative (−) circuit connections for powering the coloured and of these materials. Dual-band ECWs are on the horizon1,7,11,50,59 ;
bleached states. b, Optical transmittance of the bleached (red) and vertical integration with optimized near-UV solar cells can enable
coloured (blue) states of the combined stack using air as the background, local, self-contained power generation and expedite deployment of
along with transmittance of the glass/ITO/electrolyte stack as background these energy-saving technologies.
(black). c, Optical transmittance of the bleached (red) and coloured (blue)
states of the combined stack after subtraction of the background spectrum Methods
provided in b. Molecular semiconductors D, A1, A2. D (or cTBFDBC; contorted
0 0 0 0
tetrabenzofuranyldibenzocoronene; dibenzo(3,4:9,10)benzo(4 ,5 )furan(3 ,2 :5,6)-
0 0 0 0 0 0 0 0

employed here with those requiring less charge for switching and/or
the electrolyte to reduce mass transfer limitations2,57 .
When deployed, this technology should be implemented with
an energy storage component, trickle-charged by our solar cell,
to enable uninterrupted operation of the ECW regardless of solar
intermittency. Our D/A1 solar cell produces 4.7 J cm2 of energy per
hour of 1 sun illumination. The total charge and energy needs for a
colouring and bleaching cycle (Fig. 7f) are given in Supplementary
Table 2. At an applied voltage of ±1.2 V, complete colouring and
bleaching of the ECW requires 4.7 and 5.0 mJ cm−2 , respectively,
and maintenance pulses every 10 min each require approximately
1.2 mJ cm−2 . Assuming a need to cycle the ECW between its
benzo(4 ,5 )furan(2 ,3 :7,8)benzo(4 ,5 )furan-(3 ,2 :11,12)coroneno(1,2-
b)benzofuran) was synthesized according to previously reported methods20 .
Carbon tetrabromide (99%), triisopropyl phosphite (95%),
tris(dibenzylideneacetone)dipalladium(0) (97%),
0 0
2-Dicyclohexylphosphino2 ,6 -dimethoxybiphenyl (SPhos, %), potassium
phosphate tribasic (≥98%), 2-benzofuranylboronic acid MIDA ester (97%),
anhydrous dichloromethane (99.9%), dioxane (99.5%), iodine (≥99.8%), and
propylene oxide (≥99%) were all purchased from Sigma-Aldrich and used
as-received. Dichloromethane and toluene were purchased from Fisher
Scientific and used as-received. A1 (or 8Cl-cHBC; 1,3,6,8,13,15,18,20-
0 0 0 0 00 00 00 00
octachlorotrinaphtho(1,2,3,4-fgh:1 ,2 ,3 ,4 -pqr:1 ,2 ,3 ,4 -za1b1)trinaphthylene)
and A2 (or 12Cl-cHBC; 1,3,6,7,8,13,14,15,18,19,20-dodecachlorotrinaptho
0 0 0 0 00 00 00 00
(1,2,3,4-fgh:1 ,2 ,3 ,4 -pqr:1 ,2 ,3 ,4 -za1b1)trinapthylene) were synthesized
according to previously reported methods21 . Bis(triphenylphosphine)

NATURE ENERGY 2, 17104 (2017) | DOI: 10.1038/nenergy.2017.104 | www.nature.com/natureenergy 7

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
ARTICLES
palladium(II) dichloride (≥99.99%), potassium carbonate (99+%), and
3,5-dichlorophenylboronic acid (98%) were purchased from Sigma-Aldrich and
used as-received. 3,4,5-trichlorophenylboronic acid (AstaTech, 98%) was used
as-received.
Thin-film preparation and characterization. C60 (99.5%, NanoC) and D were
purified once using horizontal physical-vapour transport with Ar (99.999%,
Airgas), as the inert carrier gas. A1 and A2 were purified via recrystallization in
toluene. Thin films of C60 , D, A1 and A2 were thermally evaporated at a rate of
1 Å s−1 in a chamber with a pressure < 2 × 10−6 torr. Absorption spectra of thin
films of C60 , D, A1, and A2 on glass slides were collected on an Agilent Cary 5000
UV–Vis–NIR spectrophotometer. Film thicknesses for Figs 2c and 3a were
identical to the thicknesses used in devices: 23 nm (D), 17 nm (A1, A2) and 40 nm
(C60 ). Ultraviolet photoelectron spectroscopy (UPS) and inverse photoelectron
spectroscopy (IPES) measurements were performed at room temperature, in a
dedicated ultrahigh vacuum chamber with a base pressure of <2 × 10−10 torr.
10-nm-thick films of A1 or A2 were evaporated in a thermal evaporator located
in a nitrogen-filled glovebox and transported to the measurement vacuum system
in nitrogen. UPS was performed using a Specs gas discharge lamp operating with
helium, yielding helium I photons at 21.22 eV. The energy resolution in UPS was
0.15 eV. IPES was performed in isochromat mode using a home-built set-up60 ,
with an energy resolution of 0.45 eV. A Helios G3 DualBeam Focused Ion Beam
was used to prepare a cross-section of a D/A2 stack on silicon for imaging with a
Titan Themis 80–300 Cubed Double Cs-corrected Scanning/Transmission
Electron Microscope. We chose A2 for this imaging experiment instead of A1
because it has a higher electron density contrast relative to the donor layer (D).
Solar cell fabrication and testing. MoO3 (99.97%, Sigma-Aldrich)
and bathocuproine (99.99%, Sigma-Aldrich) were used as-received. Glass
substrates (23 mm × 27 mm) pre-patterned with an ITO strip 7.5 mm wide and
23 mm long (20 /square, TFD) were cleaned by sonication in deionized water,
acetone, and isopropyl alcohol, and then dried with nitrogen. The substrates
were then immediately transferred to a nitrogen glovebox for sequential
evaporation of device layers. MoO3 and bathocuproine (BCP) layers were each
5 nm thick, deposited at 1 Å s−1 . After deposition of BCP, 60 nm of Al (99.999%,
Kurt J. Lesker) was evaporated through patterned masks at a rate of 2 Å s−1
to define the active area. The devices were placed under AM1.5G 100 mW cm−2
illumination in a nitrogen-filled glovebox and the current density–voltage (J –V)
characteristics were acquired with a Keithley 2400 (Fig. 4 and Supplementary
Fig. 5) or a Keithley 2602B source measurement unit (Figs 5 and 6), respectively.
J –V and external quantum efficiency (EQE) measurements were performed using
a 300 W xenon arc lamp (Newport Oriel), with filtered monochromatic light
from a Cornerstone 260 1/4 M double grating monochromator (Newport 74125)
for the EQE measurements. A Newport calibrated photodetector (model 71582)
served as the reference cell. The spectral mismatch factor, M, was calculated
to be 0.97, 0.96, and 1.11 for the D/A1, D/A2, and D/C60 cells, respectively.
We fabricated and tested 57, 28, and 24 devices comprising A1, A2, and C60
as the acceptor, respectively. The active area of each device for these experiments
was 0.18 cm2 . The yield on functional devices is 100% with all three acceptors; the
average and standard deviation values for relevant device characteristics are
summarized in Supplementary Table 1. The solar cells exhibit identical J –V
characteristics in forward and reverse directions (Supplementary Fig. 10).
D/A1 cells (active area 0.18 cm2 ) were encapsulated in a nitrogen-filled
glovebox using a glass-on-glass architecture and DELO-KATIOBOND LP686
solvent-free epoxy. The cells remained in the glovebox overnight to ensure full
curing of the epoxy before they were brought out to the ambient. A Keithley 2400
source measurement unit was used to continuously monitor the maximum-power
point and periodically probe the current density–voltage (J –V ) characteristics of
the cells during continuous exposure to AM1.5G 100 mW cm−2 illumination
provided by a 300 W xenon arc lamp (Newport Oriel). The cell stage was
maintained at 21 ◦ C throughout the ageing measurements.
To fabricate cells with areas larger than our standard test cell, ITO-coated
glass substrates (20 /square, Colorado Concept Coatings) having dimensions
23 mm × 27 mm were patterned with an ITO strip having dimensions
12 mm × 27 mm. Photomasks with varying sizes were used during aluminium top
electrode deposition to define device active areas of 0.24, 0.48, 0.96, and 2.22 cm2 .
The substrates used to prepare the 10 cm2 devices reported in Fig. 6 were
obtained by a multi-step patterning process. ITO-coated glass sheets
(20 /square, Colorado Concept Coatings) were cut into 5 cm × 5 cm squares
and cleaned by sonication in soap and deionized water, followed by deionized
water, acetone, isopropanol, and then dried with nitrogen. Photolithography was
then performed to define the bottom electrode (Fig. 6b, light brown) and the
30-nm-thick Ag-grid current collector (Fig. 6b,c, white). These patterned
substrates were inspected for defects using an optical microscope before a final
cleaning via sonication in 2-propanol for 3 min followed by drying with nitrogen.
To create a smooth surface for device layers, the Ag-grid substrates were
8 NATURE ENERGY 2, 17104 (2017) | DOI: 10.1038/nenergy.2017.104 | www.nature.com/natureenergy
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE ENERGY
augmented with a PEDOT:PSS layer per literature38 . PEDOT:PSS (Heraeus,
Clevios pH 1000) was mixed with 5 vol% dimethyl sulfoxide and stirred overnight
in the dark before the dispersion was spin-coated atop the patterned substrates at
4,000 r.p.m. (1,000 r.p.m. ramp) for 40 s. The substrates were then annealed at
140 ◦ C for 30 min in a glovebox before they were transferred to the evaporator
chamber for deposition of device layers as outlined above.
PANI-PAAMPSA synthesis. Aniline (Sigma-Aldrich, 99.5%), ammonium
peroxydisulfate (98.9%, Fisher Scientific), and PAAMPSA (Scientific Polymer
Products, 10.36 wt% in water, reported MW 800 kg mol−1 ) were used in the
synthesis of PANI-PAAMPSA as-purchased. PANI-PAAMPSA was synthesized by
template polymerization of aniline along PAAMPSA at a 1:1 monomer-to-acid
molar ratio in deionized water, as described by Yoo and colleagues54 . Ammonium
peroxydisulfate was added to the mixture dropwise at a 1:0.9
monomer-to-oxidizing agent molar ratio. The reaction mixture was stirred in an
ice bath for 24 h. Acetone was slowly added to precipitate the PANI-PAAMPSA
complex, which was dried under vacuum for 24 h.
Electrochromic window (ECW) fabrication and characterization. A monolayer
of 12-(phosphonododecyl)phosphonic acid (Sigma-Aldrich, 97%) was covalently
bound to pre-cleaned ITO/glass by the tethering-by-aggregation-and-growth
method developed by Schwartz and colleagues61 . A 5 wt% PANI-PAAMPSA
dispersion in water that were stirred for at least 10 days—or as-purchased
PEDOT:PSS (Heraeus, Clevios pH 1000)—were spin-coated onto phosphonic
acid-modified ITO/glass substrates. We varied the spin-coating rates from
700 r.p.m. to 1,000 r.p.m., and deposited 1 to 2 layers of PANI-PAAMPSA and 1
to 4 layers of PEDOT:PSS films to vary the thicknesses of the electrochromic
polymers. These polymer films were dried for 1 min at 100 ◦ C between each
deposition. These films were then vigorously shaken in dichloroacetic acid (DCA,
Sigma-Aldrich, 99%) for 3 min as described by Yoo and colleagues51 . DCA
treatment is performed in order to eliminate any hysteresis in electrochromic
switching due to ion-transport limitation50 . Excess DCA was removed by baking
PANI-PAAMPSA films at 170 ◦ C and PEDOT:PSS films at 140 ◦ C. The films were
kept under vacuum for at least 3 h prior to testing to remove any residual DCA.
Film thicknesses were measured using a Dektak 3.21 profilometer.
To bring PANI-PAAMPSA and PEDOT:PSS to opposite oxidation states,
PEDOT:PSS was oxidized for 200 s at 0.9 V (versus Ag/AgCl) and
PANI-PAAMPSA was reduced for 200 s at −0.6 V (versus Ag/AgCl) in pH 8,
100 mM phosphate buffer solution using CH-Instruments CHI-660
electrochemical workstation. 1-Ethyl-3-methylimidazolium
bis(trifluoromethylsulfonyl)imide (EMIM TFSI, Sigma-Aldrich, >97%) mixed
with 5 wt% fumed silica (Sigma-Aldrich, 0.007 µm) was used as the gel
electrolyte55 . UV–Vis–NIR spectra of the ECWs were collected using an Agilent
8453 UV–Vis–NIR spectrometer while simultaneously performing
chronoamperometry experiments using a CH-Instruments CHI-660
electrochemical workstation to switch between the bleached and the coloured
states of our ECW. Spectra in Fig. 7 were background subtracted, with two
ITO/glass substrates and EMIM TFSI/silica ionogel used as the background.
ECW video methods. A 2.25 cm2 ECW was prepared as described above. A
1.38 cm2 solar cell with D/A1 as the active layers was used to drive the switching
of this ECW between its coloured and bleached states. The cathode and anode of
the solar cell were connected to the PANI-PAAMPSA and PEDOT:PSS
electrodes, respectively, to achieve the coloured state of the ECW. The
connections were reversed to access the bleached state.
Film thicknesses for various ECWs. ECW video: PANI-PAAMPSA: 350 nm,
PEDOT:PSS: 285 nm. Potential map (Fig. 7c,d and Supplementary Fig. 8):
PANI-PAAMPSA: 260 nm, PEDOT:PSS: 320 nm. Stability/cyclability and
charge/energy calculations (Fig. 7e,f and Supplementary Table 2):
PANI-PAAMPSA: 370 nm, PEDOT:PSS: 355 nm. Combined device (Fig. 8):
PANI-PAAMPSA: 245 nm, PEDOT:PSS 200 nm.
Data availability. The data that support the plots within this paper and other
findings of this study are available from the corresponding author upon
reasonable request.
Received 10 February 2017; accepted 30 May 2017;
published 30 June 2017; corrected 18 September 2017
References
1. Korgel, B. A. Materials science: composite for smarter windows. Nature 500,
278–279 (2013).
2. Dyer, A. L. et al. A vertically integrated solar-powered electrochromic window
for energy efficient buildings. Adv. Mater. 26, 4895–4900 (2014).
NATURE ENERGY
3. US Dep. Energy, Energy Efficiency and Renewable Energy 2011
Buildings Energy Data Book (D & R International, Ltd., 2012);
http://go.nature.com/2rI7grb
4. Wang, Y., Runnerstom, E. L. & Milliron, D. J. Switchable materials for smart
windows. Annu. Rev. Chem. Biomol. Eng. 7, 283–304 (2016).
5. Shehabi, A. et al. U.S. energy savings potential from dynamic daylighting
control glazings. Energy Build. 66, 415–423 (2013).
6. DeForest, N. et al. United States energy and CO2 savings potential from
deployment of near-infrared electrochromic window glazings. Build. Environ.
89, 107–117 (2015).
7. Llordés, A., Garcia, G., Gazquez, J. & Milliron, D. J. Tunable near-infrared and
visible-light transmittance in nanocrystal-in-glass composites. Nature 500,
323–326 (2013).
8. Khandelwal, H., Schenning, A. P. H. J. & Debije, M. G. Infrared regulating
smart window based on organic materials. Adv. Energy Mater.
http://dx.doi.org/10.1002/aenm.201602209 (2017).
9. Wu, J.-J. et al. Fast-switching photovoltachromic cells with tunable
transmittance. ACS Nano 3, 2297–2303 (2009).
10. Cannavale, A. et al. Perovskite photovoltachromic cells for building integration.
Energy Environ. Sci. 8, 1578–1584 (2015).
11. Cannavale, A. et al. Forthcoming perspectives of photoelectrochromic devices:
a critical review. Energy Environ. Sci. 9, 2682–2719 (2016).
12. Lunt, R. R. & Bulovic, V. Transparent, near-infrared organic photovoltaic solar
cells for window and energy-scavenging applications. Appl. Phys. Lett. 98,
113305 (2011).
13. Betancur, R. et al. Transparent polymer solar cells employing a layered
light-trapping architecture. Nat. Photon. 7, 995–1000 (2013).
14. Chen, C. et al. Visibly transparent polymer solar cells produced by solution
processing. ACS Nano 6, 7185–7190 (2012).
15. Polman, A. et al. Photovoltaic materials: present effiencies and future
challenges. Science 352, aad4424 (2016).
16. Gao, Y. et al. Homo-tandem polymer solar cells with V oc > 1.8 V for efficient
PV-driven water splitting. Adv. Mater. 28, 3366–3373 (2016).
17. Zhang, X. et al. Flexible all-solution-processed all-plastic multijunction solar
cells for powering electronic devices. Mater. Horiz. 3, 452–459 (2016).
18. Heliatek Heliatek Sets New Organic Photovoltaic World Record Efficiency of
13.2% (2016); http://www.heliatek.com/en/press/press-releases/details/
heliatek-sets-new-organic-photovoltaic-world-record-efficiency-of-13-2
19. Rand, B. P., Burk, D. P. & Forrest, S. R. Offset energies at organic
semiconductor heterojunctions and their influence on the open-circuit voltage
of thin-film solar cells. Phys. Rev. B 75, 115327 (2007).
20. Davy, N. C. et al. Contorted hexabenzocoronenes with extended heterocyclic
moieties improve visible-light absorption and performance in organic solar
cells. Chem. Mater. 28, 673–681 (2016).
21. Hiszpanski, A. M. et al. Halogenation of a nonplanar molecular semiconductor
to tune energy levels and bandgaps for electron transport. Chem. Mater. 27,
1892–1900 (2015).
22. Loo, Y.-L. et al. Unusual molecular conformations in fluorinated, contorted
hexabenzocoronenes. Org. Lett. 12, 4840–4843 (2010).
23. Hiszpanski, A. M. et al. Post-deposition processing methods to induce
preferential orientation in contorted hexabenzocoronene thin films. ACS Nano
7, 294–300 (2013).
24. Hiszpanski, A. M. et al. Tuning polymorphism and orientation in organic
semiconductor thin films via post-deposition processing. J. Am. Chem. Soc.
136, 15749–15756 (2014).
25. Benning, P. J. et al. C60 and C70 fullerenes and potassium fullerides. Phys. Rev. B
45, 6899–6913 (1992).
26. Sullivan, P. et al. An N-ethylated barbituric acid end-capped bithiophene as an
electron-acceptor material in fullerene-free organic photovoltaics. Chem.
Commun. 51, 6222–6225 (2015).
27. Akkerman, Q. A. et al. Strongly emissive perovskite nanocrystal inks for
high-voltage solar cells. Nat. Energy 2, 16194 (2016).
28. Peng, Y. et al. High open-circuit voltage, high fill factor single-junction organic
solar cells. Appl. Phys. Lett. 105, 083304 (2014).
29. Sullivan, P. et al. Halogenated boron subphthalocyanines as light harvesting
electron acceptors in organic photovoltaics. Adv. Energy Mater. 1,
352–355 (2011).
30. Bartynski, A. N. et al. Symmetry-breaking charge transfer in a zinc
chlorodipyrrin acceptor for high open circuit voltage organic photovoltaics.
J. Am. Chem. Soc. 137, 5397–5405 (2015).
31. Bijleveld, J. C., Verstrijden, R. A. M., Wienk, M. M. & Janssen, R. A. J.
Maximizing the open-circuit voltage of polymer: fullerene solar cells.
Appl. Phys. Lett. 97, 85–88 (2010).
32. Suddard-Bangsund, J. et al. Organic salts as a route to energy level control
in low bandgap, high open-circuit voltage organic and transparent solar
cells that approach the excitonic voltage limit. Adv. Energy Mater. 6,
1501659 (2015).
NATURE ENERGY 2, 17104 (2017) | DOI: 10.1038/nenergy.2017.104 | www.nature.com/natureenergy
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
ARTICLES
33. Masuko, K. et al. Achievement of more than 25% conversion heterojunction
solar cell. IEEE J. Photovolt. 4, 1433–1435 (2014).
34. Ben Dkhil, S. et al. Square-centimeter-sized high-efficiency polymer solar cells:
how the processing atmosphere and film quality influence performance at large
scale. Adv. Energy Mater. 6, 1600290 (2016).
35. Choi, S., Potscavage, W. J. & Kippelen, B. Area-scaling of organic solar cells.
J. Appl. Phys. 106, 54507 (2009).
36. Verreet, B., Heremans, P., Stesmans, A. & Rand, B. P. Microcrystalline organic
thin-film solar cells. Adv. Mater. 25, 5504–5507 (2013).
37. Cnops, K. et al. 8.4% efficient fullerene-free organic solar cells exploiting
long-range exciton energy transfer. Nat. Commun. 5, 3406 (2014).
38. Li, Y. et al. High-efficiency robust perovskite solar cells on ultrathin flexible
substrates. Nat. Commun. 7, 10214 (2016).
39. Zhan, C., Zhang, X. & Yao, J. New advances in non-fullerene acceptor based
organic solar cells. RSC Adv. 5, 93002–93026 (2015).
40. Tsai, P. et al. Large-area organic solar cells by accelerated blade coating.
Org. Electron. 22, 166–172 (2015).
41. Agrawal, N. et al. Efficient up-scaling of organic solar cells. Sol. Energy Mater.
Sol. Cells 157, 960–965 (2016).
42. Organic Solar Cells: Fundamentals, Devices, and Upscaling (Pan Stanford
Publishing Pte. Ltd., 2014).
43. Peters, C. H. et al. High Efficiency Polymer solar cells with long operating
lifetimes. Adv. Energy Mater. 1, 491–494 (2011).
44. Peters, C. H. et al. The mechanism of burn-in loss in a high efficiency polymer
solar cell. Adv. Mater. 24, 663–668 (2012).
45. Zhang, Y. et al. PCDTBT based solar cells: one year of operation under
real-world conditions. Sci. Rep. 6, 21632 (2016).
46. Tong, X. et al. Intrinsic burn-in efficiency loss of small-molecule organic
photovoltaic cells due to exciton-induced trap formation. Sol. Energy Mater.
Sol. Cells 118, 116–123 (2013).
47. Cheng, P. & Zhan, X. Stability of organic solar cells: challenges and strategies.
Chem. Soc. Rev. 45, 2544–2582 (2016).
48. Xie, Z. et al. Integrated smart electrochromic windows for energy saving and
storage applications. Chem. Commun. 50, 608–610 (2014).
49. Amasawa, E., Sasagawa, N., Kimura, M. & Taya, M. Design of a new
energy-harvesting electrochromic window based on an organic polymeric dye,
a cobalt couple, and PProDOT-Me2 . Adv. Energy Mater. 4, 1400379 (2014).
50. Tarver, J. D., Yoo, J. E. & Loo, Y.-L. Polyaniline exhibiting stable and reversible
switching in the visible extending into the near-IR in aqueous media. Chem.
Mater. 22, 2333–2340 (2010).
51. Yoo, J. E. et al. Directly patternable, highly conducting polymers for broad
applications in organic electronics. Proc. Natl Acad. Sci. USA 107,
5712–5717 (2010).
52. Yoo, J. E. et al. Polymer conductivity through particle connectivity.
Chem. Mater. 21, 1948–1954 (2009).
53. Yoo, J. E., Bucholz, T. L., Jung, S. & Loo, Y.-L. Narrowing the size distribution of
the polymer acid improves PANI conductivity. J. Mater. Chem. 18,
3129–3135 (2008).
54. Yoo, J. E. et al. Improving the electrical conductivity of polymer acid-doped
polyaniline by controlling the template molecular weight. J. Mater. Chem. 17,
1268–1275 (2007).
55. Horowitz, A. I., Wang, Y. & Panzer, M. J. Reclamation and reuse of ionic liquids
from silica-based ionogels using spontaneous water-driven separation. Green
Chem. 15, 3414–3420 (2013).
56. Lin, T. H. & Ho, K. C. A. complementary electrochromic device based on
polyaniline and poly(3,4-ethylenedioxythiophene). Sol. Energy Mater. Sol. Cells
90, 506–520 (2006).
57. Tarver, J. & Loo, Y.-L. Mesostructures of polyaniline films affect
polyelectrochromic Switching. Chem. Mater. 23, 4402–4409 (2011).
58. Hachmann, J. et al. The Harvard clean energy project: large-scale
computational screening and design of organic photovoltaics on the world
community grid. J. Phys. Chem. Lett. 2, 2241–2251 (2011).
59. Arroyave, F. A. & Reynolds, J. R. Dioxypyrrole-based polymers via
dehalogenation polycondensation using various electrophilic halogen sources.
Macromolecules 45, 5842–5849 (2012).
60. Wu, C. I., Hirose, Y., Sirringhaus, H. & Kahn, A. Electron-hole interaction
energy in the organic molecular semiconductor PTCDA. Chem. Phys. Lett. 272,
43–47 (1997).
61. Hanson, E. L. et al. Bonding self-assembled, compact organophosphonate
monolayers to the native oxide surface of silicon. J. Am. Chem. Soc. 125,
16074–16080 (2003).
Acknowledgements
N.C.D. acknowledges financial support from the National Science Foundation Graduate
Research Fellowship Program under Grant No. DGE 1148900. We (N.C.D., M.S.-E. and
Y.-L.L.) acknowledge support from NSF MRSEC funding through Princeton Center for
Complex Materials (DMR-1420541) and the Wilke Family Fund administered by the
9
ARTICLES NATURE ENERGY

School of Engineering and Applied Science at Princeton University. A.K. acknowledges
support from the National Science Foundation under Grant No. DMR-1506097. We
thank M. Panzer and H. Qin of Tufts University for valuable discussions regarding
ionogels, and G. Man of Princeton University for assistance with UPS/IPES analysis.
Author contributions
N.C.D., M.S.-E. and Y.-L.L. wrote the manuscript; N.C.D. synthesized molecular
semiconductors; X.L. and A.K. provided UPS/IPES measurements and analysis; N.C.D.
and J.G. fabricated and characterized PV cells; M.S.-E. synthesized PANI-PAAMPSA;
M.S.-E. and A.L. fabricated and characterized ECWs; M.S.-E. and N.D. integrated ECW
and PV components for solar-powered ECW demonstration. N.Y. conducted TEM
sample preparation and imaging. All authors discussed the results and reviewed
the manuscript.
Additional information
Supplementary information is available for this paper.
Reprints and permissions information is available at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to Y.-L.L.
How to cite this article: Davy, N. C. et al. Pairing of near-ultraviolet solar cells with
electrochromic windows for smart management of the solar spectrum. Nat. Energy 2,
17104 (2017).
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
Competing interests
The authors declare no competing financial interests.

10 NATURE ENERGY 2, 17104 (2017) | DOI: 10.1038/nenergy.2017.104 | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Author Correction
DOI: 10.1038/s41560-017-0017-8

Author Correction: Pairing of near-ultraviolet solar cells with electrochromic windows

for smart management of the solar spectrum
Nicholas C. Davy1, Melda Sezen-Edmonds1, Jia Gao1, Xin Lin2, Amy Liu1, Nan Yao3, Antoine Kahn2 and Yueh-Lin Loo1,4*

Nature Energy 2, 17104 (2017); published online 30 June 2017; corrected online 18 September 2017.

In the version of this Article originally published, Fig. 4f was incorrect. This error has now been corrected (see new figure below).

a 1.0 b 1.8
0.5
1.6
Current density (mA cm−2)

0.0 D/A1
−0.5
1.4
D/A2
−1.0
−1.5 Voc (V) 1.2

−2.0
1.0
−2.5 D/A1
D/A2 D/C60
−3.0 D/C60 0.8
−3.5
−0.5 0.0 0.5 1.0 1.5 2.0 −4.2 −4.0 −3.8 −3.6 −3.4 −3.2 −3.0
Voltage (V) LUMO (V)

c 40 d

30 Al
BCP
D/A1 A2
Counts

20
D
MoO3
10
Si 30 nm

0
1.60 1.61 1.62 1.63 1.64 1.65
Voc (V)

e 2.0 f 16
D/A1 D/A1
D/A2 D/A2
D/C60 D/C60
Spectral power response (%)
Power density (mW cm−2)

1.5 12

1.0 8

0.5 4

0.0 0
0.0 0.5 1.0 1.5 300 400 500 600 700
Voltage (V) Wavelength (nm)

Nature Energy | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.