Multiple Exciton Generation For Photoelectrochemical Hydrogen Evolution Reactions With Quantum Yields Exceeding 100%

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

ARTICLES

PUBLISHED: 3 APRIL 2017 | VOLUME: 2 | ARTICLE NUMBER: 17052

Multiple exciton generation for


photoelectrochemical hydrogen evolution
reactions with quantum yields exceeding 100%
Yong Yan1,2*, Ryan W. Crisp1,3, Jing Gu1,4, Boris D. Chernomordik1, Gregory F. Pach1,5,
Ashley R. Marshall1,5, John A. Turner1 and Matthew C. Beard1*

Multiple exciton generation (MEG) in quantum dots (QDs) has the potential to greatly increase the power conversion efficiency
in solar cells and in solar-fuel production. During the MEG process, two electron–hole pairs (excitons) are created from the
absorption of one high-energy photon, bypassing hot-carrier cooling via phonon emission. Here we demonstrate that extra
carriers produced via MEG can be used to drive a chemical reaction with quantum efficiency above 100%. We developed a
lead sulfide (PbS) QD photoelectrochemical cell that is able to drive hydrogen evolution from aqueous Na2 S solution with a
peak external quantum efficiency exceeding 100%. QD photoelectrodes that were measured all demonstrated MEG when the
incident photon energy was larger than 2.7 times the bandgap energy. Our results demonstrate a new direction in exploring
high-efficiency approaches to solar fuels.

S
olar-to-chemical energy conversion for energy storage, 100% due to MEG have been observed in PbSe QD solar cells6 , PbSe
fuels and feedstocks, particularly to address greenhouse gas nanorod-based solar cells7 and PbTe QD solar cells8 .
emissions, is one of the primary goals for scientists in the field In addition to photovoltaic applications, quantum-confined
of solar energy generation1,2 . High solar conversion efficiency is key semiconductors are being explored for solar-fuel generation. In
for a viable technology as that directly impacts the land area to be that case, the photocarriers initiate chemical reactions at a
covered and ultimately impacts the cost of the system. The main loss nanocrystal/liquid interface. For example, systems that exhibit
limiting the conversion efficiency in solar conversion systems is that high EQE (60–90%) for photoelectrochemical (PEC) hydrogen
photons with energies greater than the semiconductor bandgap (E g ) generation (EQEhy , defined as two times the number of H2
produce hot carriers that relax via electron–phonon scattering and molecules divided by the number of incident photons) have been
the subsequent phonon dissipation reduces the energy conversion developed that utilize CdSe9,10 , CdS11 and PbS12,13 QDs. The EQEhy
efficiency3 . Multiple exciton generation (MEG), within quantum- has approached 100% for photoelectrodes with 2 or 5 layers of
confined semiconductor nanocrystals (also referred to as quantum ZnS-coated CdSe in mesoporous TiO2 (ref. 14); however, an EQEhy
dots, QDs), describes a process where absorption of a high-energy greater than unity for solar hydrogen generation has not been
photon produces hot carriers that cool via the generation of experimentally demonstrated. An EQEhy > 100% not only has
additional electron–hole pairs (excitons) rather than the generation practical significance towards highly efficient solar-fuel production,
of heat and thus, overcome part of the losses associated with but also represents a fundamental insight towards a new direction
hot-carrier cooling. The result is that the quantum efficiency (the for the capture and conversion of solar energy to fuels. Over 100%
number of excitons produced as a fraction of the number of photons EQEhy in the UV–visible range, which represents the major energy
absorbed) is greater than 100% for those wavelengths that can pro- portion of solar irradiation, is of particular significance.
duce hot carriers with sufficient energy to drive the MEG process. In this study, we develop an approach that exhibits an
Quantum efficiencies exceeding 100% resulting from MEG have EQEhy exceeding 100% in the UV–visible range using PbS QD
been observed spectroscopically4 , as well as in the photocurrent of photoelectrodes. The system consists of two separated electrodes
appropriately designed QD solar cells. For example, a photovoltaic and is constructed so that hydrogen generation is accomplished
system has been developed with an internal quantum efficiency with no external voltage bias. One electrode consists of a conductive
(IQE) close to 200% via chemically attaching a single layer of PbS QD layer deposited on top of a FTO/TiO2 (fluorine-doped
PbS (E g = 0.96 eV) QDs on a single-crystal-anatase (001) TiO2 tin oxide) dielectric stack, while the second electrode consists
electrode5 . However, since only a single layer of QDs is employed of a platinum (Pt) mesh and is kept in the dark. Light is
the amount of light absorbed was limited and the external quantum absorbed at the photoanode within the QD layer producing free
efficiency (EQE) (defined as the number of electrons circulating electrons and holes. Photogenerated holes oxidize sulfide to sulfur
in the cell per second divided by the number of photons incident (nS2− + 2(n − 1)h+ →S2n− ) at the QD/solution interface (1M Na2 S
on the photoelectrode per second) was lower than 1% (ref. 5). aqueous solution), while electrons make their way to the Pt electrode
Subsequently, enhanced photocurrents with EQEs greater than where they can reduce H+ to H2 (g) (2H+ + 2e− →H2 ). Therefore,

1
Chemistry and Nanoscience Center, National Renewable Energy Laboratory, Golden, Colorado 80401, USA. 2 Department of Chemistry and Environmental
Science, New Jersey Institute of Technology, Newark, New Jersey 07102, USA. 3 Department of Physics, Colorado School of Mines, Golden, Colorado
80401, USA. 4 Department of Chemistry and Biochemistry, San Diego State University, San Diego, California 92182, USA. 5 Department of Chemistry and
Biochemistry, University of Colorado, Boulder, Colorado 80309, USA. *e-mail: yong.yan@njit.edu; matt.beard@nrel.gov

NATURE ENERGY 2, 17052 (2017) | DOI: 10.1038/nenergy.2017.52 | www.nature.com/natureenergy 1


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

a b
e− e−
e−
TiO2 QD
Glass
e− e− FTO hν
TiO2 e−
H2O

Energy versus NHE (V)


PbS QDs e−
−0.42
H2
S2− Ef Eg −0.40

Salt bridge
H+/H2
pH = 7 + S2−/S2−
n
(phosphate buffer) h hν

hν h+
+
H
Sn2−
H2O H+
Pt H2O
Na+ 1 M Na2S

EQEhy = 2NH2 /Nhv > 100%


e hν

c d

PbS PbS H+/H2 MEG Eg


TiO2 TiO2
FTO S2−/S2−
n
FTO

500 nm

Figure 1 | MEG PEC cell illustration. a, Schematic of the PEC apparatus. The QD layers make up the active region of the photoanode. Photogenerated holes
oxidize sulfide in one compartment while electrons reduce hydrogen in the other compartment at the dark Pt cathode. A salt bridge connects the two
compartments and transports H+ and Na+ . b, The energy levels of the QD photoelectrode (Eg is the QD bandgap) with respect to normal hydrogen
electrode, NHE (V). Electrons leave the QD layer through TiO2 with enough chemical potential to reduce H+ to H2 while holes can oxidize sulfide.
c,d, Scanning electron micrographs of typical PbS QD photoelectrodes with QD layer thickness of 265 nm (c) and 370 nm (d). The scale bar is the same for
both images. e, For high-energy photons absorbed within individual QDs, MEG produces two or more electron–hole pairs that can participate in the
oxidation and reduction reactions. Using a chemical reaction at the QD surface can store the extra carriers directly in chemical bonds.

the overall reaction is to directly photosplit H2 S to H2 and S (which power intensity is kept at ∼10 µW cm−2 using neutral density
further reacts to form S2n− ). Splitting of H2 S is thermodynamically filters and accurately determined by placing a calibrated silicon
endoergic by 33 kJ mol−1 . Inexpensive and clean ways to remove photodiode or a germanium photodiode in the place of the QD
H2 S from both natural sources and chemical processes are being photoelectrode prior to adding solution. Light is directed onto
sought15 . Photosplitting H2 S can be economical and has the added the PbS/Na2 S (aq) interface and following illumination hydrogen
benefit of the production of H2 gas, which can be used as either evolves at the Pt electrode. The amount of hydrogen, collected
a chemical precursor or energy source. We note that in our from the cathode, is determined by gas chromatography (Shimadzu
experimental demonstration, developed here, the energy necessary GC-2010 Plus, Tracera) with a Carbxen 1010 PLOT column and
to split H2 S is provided by a combination of the chemical bias (via a barrier discharge ionization detector. Since the electrodes are
a pH difference in each compartment) and the photovoltage of the shorted during the entire PEC reaction there is no external bias
QD PbS photoelectrode. applied to the system; therefore, only the photovoltage provides the
necessary driving force (which is reduced from the 33 kJ mol−1 by a
Photoelectrochemical performance chemical bias).
We fabricated QD photoelectrodes with E g values of 0.85, 0.92 To directly evaluate the efficiency of PEC hydrogen generation,
and 1.08 eV by depositing the QDs from solution using a layer-by- we measured EQEhy , at specific wavelengths. Since it takes two
layer approach on top of a patterned sol-gel TiO2 /FTO-coated glass electrons to produce one hydrogen molecule, the EQEhy is equal to
slide. In the layer-by-layer process we first employ PbI2 dissolved two times the number of H2 molecules divided by the number of
in dimethylformamide to displace the oleate ligands and then incident photons, or,
use 3-mercaptopropionic acid to deposit the final 3–4 QD layers 2nH2 NA hc
 
(refs 16,17 and see Methods). The resulting QD film can vary EQEhy (λ) = (1)
between 250 and 500 nm depending on the number of layer-by-layer I0 At λ
cycles (see Supplementary Information for exact film thicknesses). where nH2 is the number of moles of hydrogen gas produced,
The PbI2 /dimethylformamide treatment produces films with very N A is Avogadro’s constant, I 0 is the incident light intensity, A is
little residual organic material, and results in I− -terminated QDs16 . the illumination area, t is the illumination time, λ is the incident
The finished QD photoelectrodes are immersed into an aqueous wavelength, h is Planck’s constant, and c is the speed of light.
1 M Na2 S solution and electrically connected to a Pt electrode The EQEhy under incident wavelengths from 380 nm to 500 nm is
that is submersed in a phosphate buffer (pH = 7) in a separate shown in Fig. 2a. For each PbS QD bandgap we fabricated three
compartment. A salt bridge connects the two compartments to photoelectrodes and measured their response separately obtaining
complete the circuit (Fig. 1a). The gas-tight PEC cell is illumi- less than a 5% difference indicating good reproducibility and
nated with a xenon lamp employing 20 nm band-pass filters reliability of our system. In the wavelength range that MEG would
at wavelengths ranging from 380 nm to 1,500 nm. The incident be expected for photoelectrodes with E g of 0.85 eV, at 450 nm

2 NATURE ENERGY 2, 17052 (2017) | DOI: 10.1038/nenergy.2017.52 | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE ENERGY ARTICLES
a 120 120 c
0.85 eV
0.92 eV
100
110 1.08 eV 110 0.85 eV

Absorbance (a.u.)
80 0.92 eV
1.08 eV
EQEhy (%)

100 100

IPCE (%)
IPCE (%)
60
90 90
40

80 80 20

70 70 0
380 400 420 440 460 480 500 400 800 1,200 1,600
Wavelength (nm) Wavelength (nm)

b
λ = 400 nm
Io = 10 μW cm−2

5 μA cm−2
Current density

1.08 eV

0.92 eV

0.85 eV

0 100 200 300 400 500


Time (s)

Figure 2 | PEC performance characteristics. a, EQE (filled symbols) for hydrogen generation induced by PbS QD photoelectrodes with Eg of 0.85 eV (red
circles), 0.92 eV (green squares) and 1.08 eV (black triangles) in Na2 S aqueous electrolyte. IPCE (open symbols) at the range of 380–500 nm is also
illustrated for comparison. (Error bars generated from standard deviations of 5-times measurement, see Supplementary Note 1.) b, Typical steady-state
chopped-light current measurement of the PbS QD photoelectrodes (scale is 5 µA cm−2 ). c, The symbols show the IPCE of PbS QD photoelectrodes at Eg
of 0.85 eV (red circles), 0.92 eV (green squares) and 1.08 eV (black triangles). The dashed lines are the absorption spectra of the QDs in solution.

(3.24E g ), 420 nm (3.47E g ), 400 nm (3.65E g ) and 380 nm (3.83E g ), photoelectrodes. A good match near the first exciton transition
or for photoelectrodes with E g of 0.92 eV, at 400 nm (3.37E g ) and energy demonstrates that the photocurrent is proportional to the
380 nm (3.55E g ), the EQEhy are all above 100% with the highest amount of light absorbed by the QDs. An IPCE >100% is observed
value of 114 ± 1.3% for E g of 0.85 eV at incident light of 380 nm. in the 450–380 nm range, reaching 118 ± 2.7% for QDs with
As far as we know, this is the first time that hydrogen has been E g of 0.85 eV. For QD photoelectrodes with E g of 0.92 eV, the
produced photoelectrochemically under visible light illumination IPCE reached 109 ± 2.6% at 380 nm (Fig. 2a,c). All of these
with a quantum yield greater than 100%. results demonstrate that we have developed a PbS QD system
In addition to the EQEhy , we also measure the incident photon- that can photoelectrochemically produce hydrogen with an EQE
to-current efficiency (IPCE), which is the number of electrons exceeding 100%. Note for these results we have quantified both
delivered per second divided by the number of incident photons per the photogenerated hydrogen gas and the photogenerated electrons
second and is determined from directly and show that both independently demonstrate MEG in a
J hc
  PEC cell.
IPCE = (2)
I0 e λ MEG comparison in PEC cells and solar cells
where J is the photocurrent density (µW cm−2 ) and e is Coulombs The enhanced IPCE or EQEhy , reported here is comparable to results
per electron. The Faradaic yield (FY) relates the EQEhy to the IPCE, reported for QD solar cells with known MEG. For instance, we
EQEhy = IPCE × FY, and is given by previously reported a PbSe QD solar cell with an EQE reaching
2nH2 114% at incident wavelength of 380 nm (ref. 6). EQEs of 122% using
FY = F (3) narrow-bandgap PbSe nanorods7 , and exceeding 120% for PbTe QD
JAt solar cells have also been reported8 .
where F is Faraday’s constant (Coulombs per mole of electrons, Interestingly, the EQEhy of our PbS QDs PEC system is larger
C mol−1 ). With our set-up under an applied external bias in the dark, than that of the photon-to-electron EQEPV of a PbS QD photovoltaic
we find the FY for hydrogen generation is >97% (Supplementary cell. For example, the maximum EQEPV of the PbS QD solar cell
Figs 1 and 2). Therefore, the EQEhy is approximately equal to the (E g = 1.3 eV) was reported at ∼80% for wavelengths between 430
IPCE. We compare (Fig. 2a open and filled symbols) the IPCE to and 500 nm (ref. 16). At shorter wavelengths, the EQEPV sharply
the EQEhy at each wavelength and find good agreement confirming decreases towards zero. The difference is more obvious in the
that FY >95%. In the following we approximate the EQEhy by blue wavelength range, less noticeable for redder wavelengths and
the IPCE. almost disappears for wavelengths longer than about 650 nm. The
To determine the IPCE we measured the photocurrent using differences between PV and PEC behaviour stem from the different
mechanically chopped light and Fig. 2b displays the typical light direction that light impinges on the devices. For PV operation
response profiles for 400 nm wavelength illumination used to light enters through the glass/FTO/TiO2 interface while for the
calculate the IPCE. In Fig. 2c the IPCE from 380 nm to 1,500 nm PEC operation light is incident on the PbS QD/Na2 S interface.
is compared to the absorption spectra of the respective QD To investigate, we illuminated a photoelectrode from the ‘back’

NATURE ENERGY 2, 17052 (2017) | DOI: 10.1038/nenergy.2017.52 | www.nature.com/natureenergy 3


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

a
120 Eg = 0.85 eV Front side

hν 100
110


100 80

APCE (%)
IPCE (%)

90 60

80
8 0 40


hν 0.85 eV
70
7 0 20
0.92 eV
Back side 1.08 eV
60
6 0 0

400 450 500 550 600 650 2.0 2.4 2.8 3.2

Wavelength (nm) Photo energy (eV)

Figure 3 | IPCE front and back illumination. Comparison of the IPCE


b
(QD Eg : 0.85 eV) for front (filled circles) and back (open circles)
illuminations and for two thicknesses of the QD layers (blue = 265 nm and 100
red = 370 nm). (Insets: demonstrations of light illumination pathway.)
80

APCE (%)
(FTO side) similar to the illumination method that a QD solar
60 hνth /Eg = 2.7
cell experiences. In that case we observe a decreasing IPCE in the
range of 420 to 380 nm (Fig. 3) similar to what is observed in PbS 40
QD solar cells. A maximum IPCE was obtained at 420 nm, but
the value at this wavelength is significantly lower than the IPCE at 20
420 nm when illuminated from the front, and is lower than 100%
(noted as 85%, Fig. 3a, red). Even though illumination through the 0
FTO/TiO2 is most common for QD solar cells, we find that such
1.6 2.0 2.4 2.8 3.2 3.6
‘back’ illumination is detrimental to achieving high IPCEs for PEC
cells where clearly ‘front’ illumination is advantageous. hν/Eg
There are at least three factors that can be responsible for the
Figure 4 | Absorbed photon-to-current efficiency measurements. a, APCE
different behaviours observed when photoexcited from the front or
as a function of illumination photon energy (eV). b, APCE as a function of
back. First, the active wavelength range needed to observe MEG is
incident photo energy divided by QD bandgap energy, or multiples of the
interfered with by the absorption of the TiO2 or FTO layers along
bandgap (hν/Eg ). (Blue solid lines, linear fitting hν/Eg from 1.6 to 2.5 and
the path of the incident photons prior to reaching the QD layer,
from 2.8 to 3.6.). The threshold photon energy (hν th /Eg ) is about 2.7.
since FTO and TiO2 begin to absorb light for wavelengths less
than 420 nm. This produces a drop in EQE values at wavelengths
shorter than 420 nm. When we illuminate from the PbS QD side any For back illumination, holes that do not undergo interfacial
photon losses arise from reflections from the cell window interfaces recombination must transport through the bulk of the PbS QD
and the sulfide solution rather than the FTO/TiO2 . Neither the film to oxidize sulfide within the solution and thereby complete
solvent nor the window absorb within the range from 350 to 600 nm the circuit. Under front illumination electrons must transport
(Supplementary Fig. 3). As a result, in the wavelength range from though the PbS QD film and reach the TiO2 to be collected. Bulk
500 nm to 380 nm (Fig. 2a), we observe a continuously increasing recombination that occurs within the QD layer would increase
IPCE or EQEhy , rather than a sharp decrease for wavelengths for thicker QD films. Therefore, we expect that the performance
shorter than 420 nm. Such an observation also matches the expected difference between back and front illumination would depend on
absorption spectra of the QDs (see Fig. 2c for comparison of the QD the PbS QD thickness. We compare front and back illuminations for
solution absorption spectra (dashed lines) to IPCE). PbS QD films that are 265 nm (Fig. 3, blue) and 370 nm thick (red).
A second reason that the different illumination pathways could The IPCE of front illumination remains indistinguishable; however,
give different EQEs is due to the different reflection losses between the back illumination shows a significant difference: the 370 nm
the glass/FTO and the PbS QD layer. The IPCE obtained from ‘back’ thickness exhibits ∼10% lower efficiency than the 265-nm-thick
illumination is significantly lower than the ‘front’ illumination even QD layer due to the longer hole-diffusion path length. The IPCE
in the wavelength range where TiO2 and FTO are not expected to differences are also prominent in the blue wavelength range while
interfere (no absorption of FTO/ TiO2 was observed from 420 nm less obvious in red. The differences corroborate our predication that
to 600 nm). The reflection losses at the QD/solution interface are carriers (particularly holes) lost inside the PbS QD layer largely
expected to be smaller than the sum of reflections incurred at the depending on the carrier’s path length. Undoubtedly, our current
glass/solution, glass/FTO, FTO/TiO2 and TiO2 /PbS interfaces. PEC system shows significant advantages over QDs solar cells in
Finally, the concentration of photogenerated carriers is highest terms of obtaining high EQE.
at the QD/TiO2 interface in the case of ‘back’ illumination and The internal quantum efficiency (IQEhy ) of our current system
at the QD/solution interface in the case of ‘front’ illumination. can be determined by normalizing the EQEhy to the measured
This is due to the high extinction coefficient of the PbS QDs. At absorbance of the QD photoelectrode and the sulfide solution,
the QD/solution interface, the holes undergo a quick irreversible
reaction of sulfide oxidation and are removed from the QD solid, IQEhy = EQEhy /A = EQEhy /(1 − R) ∼
= IPCE/(1 − R) = APCE (4)
reducing recombination losses. In the case of ‘back’ illumination,
however, a higher fraction of the photogenerated carriers are lost where A includes absorbance from the solution and QDs, defined
due to recombination at the back ohmic contact. as: A = APbS + ASolution . In the wavelength range of 380 nm–750 nm,

4 NATURE ENERGY 2, 17052 (2017) | DOI: 10.1038/nenergy.2017.52 | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE ENERGY ARTICLES
no absorption of ASolution has been observed (Supplementary Fig. 3). Synthesis of PbS QDs from CdS QDs. PbS with different bandgap energies, E g
In addition, IQE may be easily characterized by APCE (absorbed (given by the position of the first exciton peak), of 1.08 eV, 0.92 eV, and 0.85 eV
photon-to-current efficiency) due to the near-unity Faradaic was synthesized by the cation exchange of CdS. CdS was also synthesized following
our previous procedure16,17,26 . Pb-precursor solution was synthesized by heating
efficiency for hydrogen generation in our system. Here, we plot the OLA (5 ml) and PbCl2 (1.5 mmol) in a 50 ml three-neck flask to 140 ◦ C for 30 min
APCE versus the incident photon energy E = hv in Fig. 4a and versus under nitrogen flow until a white, turbid solution was achieved. Then, the solution
hv/E g in Fig. 4b (the photon energy normalized to the bandgap was cooled or heated to a desired temperature for the injection of CdS QDs.
of the QD). Interestingly, the APCE, not adjusted for solution ab-
sorbance or reflections from the cell window, remained nearly con- PbS-1.08 eV. The synthesis of PbS-1.08 eV was based on our previously reported
stant for each QD sample at 78% ± 4% for excitation photon energy procedure17 . When the Pb-precursor reached 120 ◦ C, 1 ml of CdS QD
(∼58 mg ml−1 ) in ODE was swiftly injected. After 20 s, the reaction was
at an absolute value at 1.6 eV up to 2.2 eV. This is in agreement with quenched by a water bath. Hexane (5 ml) and OA (4 ml) were added when the
the IQE observed in PbSe and PbS QD solar cells6 . Although the reaction bath reached 70 and 40 ◦ C respectively. The washed QD solutions were
QD photoelectrode at E g = 1.08 eV did not show IQE or APCE over further isolated from the reactants by filtration with a 0.45 µm PTFE filter
100%, a clear APCE increase was also observed at around 3.0 eV. followed by centrifuging at 7,500 r.p.m. The PbS QDs were washed 2–3 times
Surprisingly, it is very clear in Fig. 4b, that the hv/E g changes in using hexane and ethanol, and then redispersed in octane. The CdS QDs were
slope with the APCE are almost all observed at around 2.7. It is synthesized by heating CdO (2.2 mmol, 0.28 g), 5.7 mmol OA (1.8 g) and 8 g of
ODE to 260 ◦ C for 20 min and allowed to cool to 30 ◦ C when a (NH4 )2 S:OLA
indicative that in our PEC system, the MEG effect is observed for precursor solution was injected and stirred for 1 h without heating. The sulfur
photon energies that are larger than 2.7E g , this observation is very precursor was prepared by stirring a solution of 80 µl (NH4 )2 S aqueous solution
similar to most of the reported value of isolated PbS QDs18 . with 10 ml of OLA for 20 min.

Conclusions PbS-0.92 eV. The synthesis of PbS-0.92 eV was based on our previously reported
Theoretically, MEG is beneficial towards obtaining a higher limit procedure17 . When the Pb-precursor reached 190 ◦ C, 1 ml of CdS QDs
(∼36 mg ml−1 ) in ODE was swiftly injected. All other conditions and treatments
of solar-fuel generation efficiency19,20 . The overall solar-to-chemical were identical to those given for the PbS-1.08 eV synthesis. CdS QDs were
conversion efficiency, ηrxn = J tot V rxn /P in , where J tot is total current synthesized by heating CdO (1 mmol, 0.128 g), 3 mmol OA (0.942 g) and 15 g of
density, V rxn is fixed at the voltage necessary to drive the chemical ODE to 260 ◦ C for 20 min, then the temperature was set to 250 ◦ C, 1 ml of the
reaction, and P in is the input power density from the solar irradiance. sulfur precursor was injected and the solution was maintained at 240 ◦ C for
The total current is Rdetermined by the EQE of the system and 20 min. For these QDs, the sulfur precursor was prepared by dissolving S powder
the solar flux, Jtot = hvmin EQE(hv)0(hv)dhv, where 0(hv) is the at a concentration of 0.5 M in ODE at 130 ◦ C.
solar flux. Here hvmin is the minimum energy needed to initiate PbS-0.85 eV. The synthesis of PbS-0.85 eV was based on our previously reported
the chemical reaction and includes any overpotential. Since V rxn is procedure17 . When the Pb-precursor reached 190 ◦ C, 1 ml of CdS QDs
fixed by the desired photochemical reaction to increase ηrxn one (∼20 mg ml−1 ) in ODE was swiftly injected and the temperature was maintained
needs to increase J tot . There are two ways to increase J tot : decrease at 180 ◦ C. Twenty minutes later, the reaction was quenched by a water bath. The
the overpotential such that hvmin is closer to the thermodynamic remainder of the reaction conditions and treatments were identical the
limit; and utilize MEG to increase EQE. Finding an inexpensive PbS-0.92 eV synthesis. Larger CdS QDs were synthesized as in the PbS-0.92 eV
synthesis but the S-precursor was injected at 265 ◦ C and the solution was
and efficient catalyst, which is a necessary and important research maintained at 250 ◦ C; ∼13 min later, additional S-precursor was added drop-wise
direction, can decrease the overpotential. The results presented until the desired size (∼5.5 nm diameter, 450 nm first exciton peak wavelength)
herein show that, under the appropriate conditions, the EQE can was achieved.
be over unity for a range of wavelengths and can thus increase J tot ,
demonstrating a parallel avenue for increasing the solar-to-chemical Photoelectrode fabrication. The final QD photoelectrode was a thick,
all-inorganic, air-stable conductive QD film formed on a patterned sol-gel
conversion efficiency, ηrxn . TiO2 /FTO-coated glass slide, which was fabricated on the basis of our previously
The generation of solar fuels is receiving significant interest1,2 . reported procedure16 . First, FTO/glass substrates were cleaned vigorously with
However, the low efficiency of low-cost materials and the high cost ethanol and UV–ozone-treated prior to depositing the TiO2 layer. Within 10 min
of the current set of high-efficiency materials for PEC hydrogen of UV–ozone treatment, 70 µl TiO2 sol-gel was deposited on the substrate and
generation are major concerns for the development of solar spun at 1,400 r.p.m. for 30 s. The FTO contact pads were cleaned using ethanol
fuels1,21–24 . In our results presented here, we show that it is possible prior to drying at 115 ◦ C and annealing at 450 ◦ C for 30 min. The resulting films
were stored in air for at least 1 day. The TiO2 sol-gel was prepared by mixing in
to capture multiple-exciton-generated electrons in a fuel-forming air 5 ml anhydrous ethanol, 2 drops hydrochloric acid, and 125 µl deionized water
reaction. It is encouraging for the future design and development and stirred while adding drop-wise 375 µl titanium ethoxide. The resulting clear
of PEC systems that MEG may be utilized to photosplit water to liquid was stirred for 48 h in a nitrogen atmosphere and then stored in a freezer.
generate hydrogen fuels with the goal of surpassing the Shockley– Second, the TiO2 -coated substrates were immersed in a 15 mg ml−1 solution of
Queisser limit. Effective water oxidation as a counter reaction is the desired QDs in hexane. The substrate was removed and immersed in a
undeniably another non-trivial step to investigate and was not 10 mM Pbl2/DMF solution for 30–60 s. The cycle was repeated to build up a thick
QD film. A post-ligand treatment with neat ACN was employed to help remove
considered here. Two or more photoelectrodes connected in series residual DMF. A mixture of 20 vol.% DMF/ACN can also solvate PbI2 and a
or in parallel with one able to take advantage of MEG is a potential photoelectrode made from this solution performed nearly as well as those with
area of future research. Finally, to make the largest impact the MEG PbI2 in DMF for the ligand treatment solvent. Typically, photoelectrodes used
onset should be as close to twice the bandgap as possible. Research 10–40 layers of the PbI2 -treated QDs and then were followed by 3–4 layers of
on new heterostructured QD systems that show onset for MEG QDs treated with 10% 3-mercaptopropionic acid in methanol, rinsing twice with
much closer to the 2E g limit is underway17,25 . methanol and drying with nitrogen. All photoelectrodes presented here were
fabricated at room temperature (23.9–26.7 ◦ C) and relative humidity that
fluctuated between 16 and 20%. The resulting QD layer of the photoelectrode
Methods varied between 250 and 500 nm depending on the number of
Materials for quantum-dot synthesis. PbCl2 (99.999%), oleylamine (OLA, tech. layer-by-layer cycles.
grade, 70%), CdO (99.99%), oleic acid (OA, tech. grade, 90%), 1-octadecene
(ODE, tech. grade, 90%), sulfur powder (99.98%), ammonium sulfide (40–48 wt% Photoelectrode assembly and active area determination. The top edge of the
in H2 O), tetrachloroethylene (TCE, ≥99.9%), hexane (99.9%), octane (99.9%), PbS and TiO2 layer on the photoelectrode was removed by a blade. This exposed
acetonitrile (ACN, grade ≥99.93%), N ,N -dimethylformamide (DMF, FTO conductive area and the copper wire were glued together through
TraceSELECT, for inorganic trace analysis, ≥99.99995% (trace metals analysis)) conductive silver epoxy (PELCO colloid silver). The edge of the above-mentioned
and ethanol (≥99.5%) were purchased from Sigma Aldrich. Lead oxide (PbO, photoelectrode and the conductive connection area was then further completely
99.99%) was purchased from Alfa Aesar. All of the chemicals were used protected via an insulation epoxy (Loctite 9462 Hysol). Further, a resistance epoxy
as received. (Loctite E-120 HP), cured at room temperature overnight, was then applied to

NATURE ENERGY 2, 17052 (2017) | DOI: 10.1038/nenergy.2017.52 | www.nature.com/natureenergy 5


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

prevent the epoxy layer from etching. The final active area of the photoelectrode Received 12 October 2016; accepted 3 March 2017;
was accurately determined by a HP Scanjet 7650 in combination with ImageJ published 3 April 2017
software according to a previously reported method with slight modification27 . In
this method, the electrode surface was aligned parallel to the scanner glass and References
was scanned to obtain an electrode surface image. The number of black pixels of 1. Lewis, N. S. Research opportunities to advance solar energy utilization. Science
the measured active area can be extracted by the ImageJ software and such pixel 351, aad1920 (2016).
number was then converted to the actual area through a standard relationship 2. Turner, J. A. Sustainable hydrogen production. Science 305, 972–974 (2004).
that was previously built (892,802 pixels per square centimetre). 3. Ross, R. T. & Nozik, A. J. Efficiency of hot-carrier solar energy converters.
J. Appl. Phys. 53, 3813–3818 (1982).
Incident light and hydrogen gas calibrations. The incident light came from 4. Klimov, V. I. Multicarrier interactions in semiconductor nanocrystals in
white light passed through a monochromator, and a neutral density filter was also relation to the phenomena of Auger recombination and carrier multiplication.
applied to adjust the light intensity at ∼10 µW cm−2 , where typically a 75 W Xe Annu. Rev. Condens. Matter Phys. 5, 285–316 (2014).
arc lamp was used as the source. The PEC system was placed in a dark room to 5. Sambur, J. B., Novet, T. & Parkinson, B. A. Multiple exciton collection in a
avoid interfering illumination. We used a silicon photodiode for the sensitized photovoltaic system. Science 330, 63–66 (2010).
380–1,060 nm range and a germanium photodiode for 1,060–1,500 nm. The 6. Semonin, O. E. et al. Peak external photocurrent quantum efficiency exceeding
calibrated photodiodes all have NIST traceable calibration, and have a similar 100% via MEG in a quantum dot solar cell. Science 334, 1530–1533 (2011).
area to the test photoelectrodes. We also applied a Newport UV-818-L detector 7. Davis, N. J. L. K. et al. Multiple-exciton generation in lead selenide nanorod
(NIST traced calibration) to compare the light intensity results at the 380–650 nm solar cells with external quantum efficiencies exceeding 120%. Nat. Commun.
range with the Si photodiode. The current measured from the Newport UV-818-L 6, 8259 (2015).
detector and the Si photodiode has less than 0.2% difference in the range of 8. Böhm, M. L. et al. Lead telluride quantum dot solar cells displaying external
380 nm to 500 nm, while the difference was manifested to 0.5% at the range from quantum efficiencies exceeding 120%. Nano Lett. 15, 7987–7993 (2015).
500 nm to 650 nm. The specific light intensity inside the PEC cell, at the specific 9. Han, Z., Qiu, F., Eisenberg, R., Holland, P. L. & Krauss, T. D. Robust
position where the photoelectrode will be placed, was accurately determined by photogeneration of H2 in water using semiconductor nanocrystals and a nickel
the current generated by the calibrated photodiode. Note that the light intensity catalyst. Science 338, 1321–1324 (2012).
measurement might correspond to ∼3% systematic error under our 10. Seol, M., Jang, J.-W., Cho, S., Lee, J. S. & Yong, K. Highly efficient and stable
measurement conditions due to the refractive index difference between the cadmium chalcogenide quantum dot/ZnO nanowires for photoelectrochemical
glass/air interface of the photodiode and the glass/solution interface of the hydrogen generation. Chem. Mater. 25, 184–189 (2012).
photoelectrode (note: light passes through the air/glass interface and then 11. Kim, H., Seol, M., Lee, J. & Yong, K. Highly efficient photoelectrochemical
through the glass/solution interface to reach the photodiode surface in solution). hydrogen generation using hierarchical ZnO/WOx nanowires cosensitized with
Such difference can be illustrated according to the equation: CdSe/CdS. J. Phys. Chem. C 115, 25429–25436 (2011).
12. Trevisan, R. et al. Harnessing infrared photons for photoelectrochemical
R = [n1 − n2 ]2 /[n1 + n2 ]2 hydrogen generation. A PbS quantum dot based ‘‘quasi-artificial leaf’’.
J. Phys. Chem. Lett. 4, 141–146 (2012).
in which R is reflectivity, n1 is the refractive index of glass and n2 is the refractive 13. Yan, H. et al. Visible-light-driven hydrogen production with extremely
index of solution in the photoelectrode case or air in the photodiode case. We can high quantum efficiency on Pt–PdS/CdS photocatalyst. J. Catalys. 266,
approximate the reflective losses via n1 = 1.47 and n2 = 1.33 (for Na2 S solution, 165–168 (2009).
approximated by water) or n2 = 1 (for air). We calculate Rsolu = 0.3% and 14. Lai, L.-H., Gomulya, W., Protesescu, L., Kovalenko, M. V. & Loi, M. A. High
Rair = 3.4%. These two numbers indicated that (∼3%) more light might reach the performance photoelectrochemical hydrogen generation and solar cells with a
photoelectrode surface than the photodiode surface. double type II heterojunction. Phys. Chem. Chem. Phys. 16, 7531–7537 (2014).
Hydrogen gas was measured with a Shimadzu GC-2010 Plus Gas 15. Ma, G. et al. Direct splitting of H2 S into H2 and S on CdS-based photocatalyst
Chromatograph (Tracera) with a highly sensitive barrier discharge ionization under visible light irradiation. J. Catalys. 260, 134–140 (2008).
detector and the carrying gas was 99.9999% ultrapure helium gas to ensure a two 16. Crisp, R. W. et al. Metal halide solid-state surface treatment for high efficiency
order higher detection limit on hydrogen than other common detectors in gas PbS and PbSe QD solar cells. Sci. Rep. 5, 9945 (2015).
chromatography. The standard hydrogen gas was generated with a sealed 17. Zhang, J. et al. Preparation of Cd/Pb chalcogenide heterostructured janus
two-compartment electrochemical cell filled with 1 M H2 SO4 and Pt as both particles via controllable cation exchange. ACS Nano. 9, 8157–8164 (2015).
cathode and anode. The theoretical hydrogen concentration was plotted against 18. Stewart, J. T. et al. Comparison of carrier multiplication yields in PbS and PbSe
the detection area of GC, as shown in Supplementary Fig. 1 and Supplementary nanocrystals: the role of competing energy-loss processes. Nano Lett. 12,
Table 1. A highly correlated linear curve was obtained between the theoretical 622–628 (2012).
hydrogen concentration and the detection area of GC, with a linear correlation 19. Shockley, W. & Queisser, H. J. Detailed balance limit of efficiency of p–n
coefficiency of 0.999952. And this calibrated equation was applied for the future junction solar cells. J. Appl. Phys. 32, 510–519 (1961).
accurate amount determination of hydrogen gas. To reduce any systematic 20. Hanna, M. C. & Nozik, A. J. Solar conversion efficiency of photovoltaic and
errors associated with different compartments we replaced only the anode with photoelectrolysis cells with carrier multiplication absorbers. J. Appl. Phys. 100,
our QD photoelectrode; the cathode compartment (where H2 evolves) remained 074510 (2006).
the same. 21. Khaselev, O. & Turner, J. A. A monolithic photovoltaic-photoelectrochemical
device for hydrogen production via water splitting. Science 280,
425–427 (1998).
Scanning electron microscopy. Scanning electron microscopy (SEM)
22. Gu, J. et al. p-Type CuRhO2 as a self-healing photoelectrode for water
measurement was conducted to obtain the thickness of the PbS films. The surface
reduction under visible light. J. Am. Chem. Soc. 136, 830–833 (2014).
morphology was analysed using a JEOL JSM 7600F field-emission scanning
23. Gu, J. et al. Water reduction by a p-GaInP2 photoelectrode stabilized by an
electron microscope operated at 5 kV. First, the film on these photoelectrodes was
amorphous TiO2 coating and a molecular cobalt catalyst. Nat. Mater. 15,
fabricated as follows: we used the methods (as described from photoelectrode
456–460 (2015).
synthesis) to make PbS film from one batch of PbS solution at a specific bandgap
24. White, J. L. et al. Light-driven heterogeneous reduction of carbon dioxide:
deposited onto a ∼3 cm × 3 cm as-prepared glass first. Then each photoelectrode
photocatalysts and photoelectrodes. Chem. Rev. 115, 12888–12935 (2015).
(in small size) was made from this large-area (3 cm × 3 cm) as-prepared
25. Cirloganu, C. M. et al. Enhanced carrier multiplication in engineered
film-coated glass, which was carefully cut into the specific size (approximately
quasi-type-II quantum dots. Nat. Commun. 5, 4148 (2014).
1 cm2 ) as noted in Supplementary Table 2. Note that the thickness results of the
26. Chernomordik, B. D., Marshall, A. R., Pach, G. F., Luther, J. M. & Beard, M. C.
film from the above table are listed below: for example PbS-0.85 eV, thicknesses
Quantum dot solar cell fabrication protocols. Chem. Mater. 29,
of electrodes 1, 2 and 3 (originated from the same large electrode), measured
189–198 (2017).
using the SEM mentioned here, were very similar, noted as 265 nm; PbS-0.92 eV,
27. Gu, J. et al. A graded catalytic-protective layer for an efficient and stable
thicknesses of electrodes 4–6 are also almost the same, noted as 255 nm;
water-splitting photocathode. Nat. Energy 2, 16192 (2017).
PbS-1.05 eV, thickness of electrodes 7–9 are also the same, noted as 270 nm.
A thicker electrode at bandgap 0.85 eV was also intentionally made to investigate
the IPCE difference from back and front illumination as shown in Fig. 3; SEM
(Fig. 1c,d) indicated a thickness of 370 nm for the film. Acknowledgements
We would like to thank B. To and C. Xiao from NREL for the scanning electron
micrographs. A. J. Nozik, J. Luther, D. Kroupa and Y. Yang are thanked for their helpful
Data availability. The data that support the plots within this paper and other discussions. Y.Y. would like to acknowledge the support from the startup fund at New
findings of this study are available from the corresponding authors on request. Jersey Institute of Technology. Y.Y., R.W.C., B.D.C., G.F.P., A.R.M. and M.C.B.

6 NATURE ENERGY 2, 17052 (2017) | DOI: 10.1038/nenergy.2017.52 | www.nature.com/natureenergy

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE ENERGY ARTICLES
acknowledge support from the Center for Advanced Solar Photophysics, an Energy Additional information
Frontier Research Center funded by the Office of Basic Energy Sciences within the Office Supplementary information is available for this paper.
of Science. J.G. and J.A.T. acknowledge the solar photochemistry programme within the Reprints and permissions information is available at www.nature.com/reprints.
division of Chemical Sciences, Geosciences, and Biosciences of the Office of Basic Energy
Sciences within the Office of Science. All work is supported by the Department of Energy Correspondence and requests for materials should be addressed to Y.Y. or M.C.B.
under contract No. DE-AC36-08GO28308 to NREL. How to cite this article: Yan, Y. et al. Multiple exciton generation for
photoelectrochemical hydrogen evolution reactions with quantum yields exceeding
Author contributions 100%. Nat. Energy 2, 17052 (2017).
Y.Y. and M.C.B. conceived the experiments and led the project; Y.Y. and J.G. carried out Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
the photoelectrode synthesis and characterization and photoelectrochemical published maps and institutional affiliations.
investigation; R.W.C., B.D.C., G.F.P. and A.R.M. carried out quantum-dot synthesis and
film preparation; Y.Y., J.A.T. and M.C.B. analysed the data; Y.Y. and M.C.B. wrote the Competing interests
manuscript with input and discussion from all authors. The authors declare no competing financial interests.

NATURE ENERGY 2, 17052 (2017) | DOI: 10.1038/nenergy.2017.52 | www.nature.com/natureenergy 7


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

You might also like