Professional Documents
Culture Documents
pubs.acs.org/NanoLett
ABSTRACT: A rapid increase in anthropogenic emission of greenhouse gases, mainly carbon dioxide, has been a growing cause
for concern. While photocatalytic reduction of carbon dioxide (CO2) into solar fuels can provide a solution, lack of insight into
energetic pathways governing photocatalysis has impeded study. Here, we utilize measurements of electronic density of states
(DOS), using scanning tunneling microscopy/spectroscopy (STM/STS), to identify energy levels responsible for photocatalytic
reduction of CO2−water in an artificial photosynthetic process. We introduce desired states in titanium dioxide (TiO2)
nanoparticles, using metal dopants or semiconductor nanocrystals, and the designed catalysts were used for selective reduction of
CO2 into hydrocarbons, alcohols, and aldehydes. Using a simple model, we provide insights into the photophysics governing this
multielectron reduction and design a new composite photocatalyst based on overlapping energy states of TiO2 and copper
indium sulfide (CIS) nanocrystals. These nanoparticles demonstrate the highest selectivity for ethane (>70%) and a higher
efficiency of converting ultraviolet radiation into fuels (4.3%) using concentrated sunlight (>4 Sun illumination), compared with
platinum-doped TiO2 nanoparticles (2.1%), and utilize hot electrons to tune the solar fuel from alkanes to aldehydes. These
results can have important implications for the development of new inexpensive photocatalysts with tuned activity and selectivity.
KEYWORDS: Artificial photosynthesis, hot electrons, quantum dots, selective catalysis
© 2014 American Chemical Society 597 dx.doi.org/10.1021/nl403783d | Nano Lett. 2014, 14, 597−603
Nano Letters Letter
Figure 1. Energy levels responsible for photocatalysis in unmodified TiO2 nanoparticles. (a) Schematic of the proposed artificial photosynthesis in
TiO2-QD based catalysts for selectively producing alkanes, alcohols, and aldehydes. (b) CO2−water photocatalytic reduction data, for unmodified
TiO2 nanoparticles, under different intensities of simulated solar irradiation (AM1.5 spectrum = 1 Sun). The chemical composition of different
catalytic products (hydrogen H2, methane CH4, ethane C2H6, acetaldehyde CH3CHO, and higher hydrocarbons including propane, butane, and
paraffins CxHy) is indicated using color coded bar graphs, on a per electron mole basis. The colored asterisk sign indicates the highest yields for
respective products obtained in our measurements. (c) Internal quantum efficiency and selectivity of solar fuels (for ethane production) under
different intensities of solar irradiance. The internal quantum efficiency (QE, percent) is computed based on the fraction of sunlight above the energy
bandgap of TiO2 nanoparticles and the total catalytic yield of different products obtained, per electron mole basis (since different products require
different number of electrons in the multielectron reduction process). Unmodified TiO2 exhibits low activity and selectivity for CO2−water
reduction. The symbols represent the average internal QE over several experiments, and the lines represent the highest and lowest yields measured.
(d) STS data showing defect levels (∼0.5−1 eV below conduction band) in TiO2 nanoparticles, likely responsible for photocatalytic activity (see
discussion on advantages of ambient measurements in the Supporting Information). The solid line shows the first derivative of current with voltage
(dI/dV) or the density of electronic states, the dashed line represents the current−voltage spectrum, and levels indicated on the right represent the
relative reduction potential values for different fuel products formed from CO2−water reduction.15 The work function of the Pt−Ir tip and the
electron affinity of TiO2 semiconductor were used to represent the data on an absolute energy level, with respect to vacuum. Standard hydrogen
electrode potential was used to represent NHE and energetic levels (w.r.t. vacuum) on the same scale. (e) Cross-section and 3D map of STS data
collected on 15 nm unmodified TiO2 nanoparticles. The current−voltage spectrum (I−V) shown here is used to obtain the density of electronic
states (dI/dV) based on tunneling probability (or current) from the Pt−Ir tip. The work function of the tip and the electron affinity of TiO2
semiconductor were used to obtain energy values with respect to vacuum level (0 eV).
Figure 2. Effect of platinum metal doping on TiO2 energy levels and photocatalytic rates. (a) CO2−water photocatalytic reduction products obtained
using doped TiO2−Pt (1 wt %, see Methods in the Supporting Information) catalyst illuminated with different intensities of AM1.5 simulated solar
irradiation. The solar irradiance was increased from half up to 4.6 times the incident radiation. The increase in sunlight illumination results in change
in product yields due to filling up of lower energy levels and likely slower cooling of hot carriers, resulting in products with higher reduction
potentials. (b) Internal quantum efficiency (percent) and selectivity (for ethane hydrocarbon fuel) obtained using TiO2−Pt (1 wt %) photocatalyst.
(c) STS spectrum collected with 0.5 wt % Pt metal dopant in TiO2 nanoparticles and (d) 1 wt % Pt doped TiO2 semiconductor nanoparticle. While
the 0.5 wt % Pt metal dopant does not cause significant changes in the electronic DOS of the TiO2 semiconductor (compared to Figure S1 in
Supporting Information), 1 wt % platinum clearly shows increased DOS between the bandgap, extending all of the way through the bandgap. This
caused likely trapping, and eventually recombination, of the photogenerated electron and hole excited in the doped semiconductor (as shown by
arrows in d). This enhanced recombination on platinum sites is likely responsible for reduced photocatalytic yields on increasing the light intensity
(as shown in b).
for the selective generation of solar fuels. Therefore, these ∼1.5−1.7 eV, when compared to CO2 reduction without
semiconductor photocatalysts with different band alignments proton (H+) assistance):15
and/or bandgap energies can yield selective products on
illumination with band-edge photons. Moreover, careful tuning H 2O + h+ → H+ + ·OH
of energy states can help in utilizing hot electrons (especially
CO2 + 8H+ + 8e− → CH4 + 2H 2O
with solar concentrators) to change the selective solar fuel from
low reduction potential alkane to aldehyde (which requires CO2 + 6H+ + 6e− → CH3OH + H 2O
higher energy electrons for reduction). Therefore, we can
utilize modifications on TiO2 semiconductors, including the CO2 + 4H+ + 4e− → HCHO + H 2O
attachment of semiconductor quantum dots, to tune the
energetic alignment of band-edge states. Since chemical where e− and h+ represents photogenerated electrons and
modification of the Fermi level (FL) in semiconductor holes, respectively. Moreover, since different solar fuels
(alkanes, alcohols, and aldehydes) require a different number
nanoparticles can have important consequences on recombi-
of photogenerated electrons, their respective efficiencies are
nation dynamics of photogenerated charges14 and density of
normalized to the number of electron moles of fuel produced
electronic states which participate in charge transfer for (see Methods in the Supporting Information).
photoreduction of adsorbed species, we carried out careful We used correlated STM and scanning tunneling spectros-
photophysical studies on unmodified (Figure 1b,c) and copy (STS)16,17 for measurement of single nanoparticle
modified TiO2 semiconductor nanoparticles (Figures 2−5). electronic density of states (see Methods in Supporting
The energetic position of these states can affect the reduction Information). Unmodified TiO2 nanoparticles demonstrated
potential (Figure 1d) and hence likely product of simultaneous low catalytic rates for reducing CO2−water under simulated
reduction of CO2−water (because simultaneous or synergistic AM1.5 illumination (represented as 1 Sun) and very low
reduction of CO2−water does not require a large overpotential selectivity toward specific solar fuels resulting in a mixture of
599 dx.doi.org/10.1021/nl403783d | Nano Lett. 2014, 14, 597−603
Nano Letters Letter
Figure 3. TiO2−Cu2−xO composite photocatalyst. (a) Chemical analysis of photocatalytic products obtained using TiO2−Cu2−xO (0 < x < 1)
composite catalyst (see Methods in Supporting Information), for CO2−water reduction, on illuminating with simulated solar radiation. (b) TiO2−
Cu2−xO photocatalyst shows higher yields (internal QE) for CO2−water reduction than unmodified TiO2, but variation in oxidation state of copper
results in change in semiconductor bandgap (Figure S2 in the Supporting Information) and traps on the surface, leading to enhanced recombination,
lower yields, and unselective product formation. Due to the large number of defect states present on the surface, the highest catalytic yields are
obtained under low light intensities (here 0.5 Sun illumination). On increasing the photogeneration rate (or solar irradiance), while the yield
decreases (explained using our photophysical model, Figure 5c), the filling of low energy electron states (c) leads to more products with a higher
reduction potential (acetaldehyde and higher hydrocarbons CxHy), thereby increasing the observed selectivity. (c) STS data of the composite
photocatalyst showing the overlapping energy levels of Cu2−xO and the TiO2 semiconductor.
alkanes, aldehydes, and alcohols (Figure 1b,c). However, the Chemical doping of nanoparticle can be used to selectively
presence of some defect states in TiO2 nanoparticles ∼0.5−1 tune the FL and introduce energetic states with-in the bandgap
eV below the conduction band (Figure 1d) accelerated the of a semiconductor.17,19 To utilize this, we introduced some
yield of CO2−water reduction reaction (Figure 1b,c). The role well-studied metal dopants in the TiO2 semiconductor to
of these defect states has also been studied by other understand their role in photocatalysis. We first introduced
groups,7,9−12 and nominally unmodified TiO2 nanoparticles varying amounts of platinum (Pt) dopant in TiO2 to optimize
with a larger number of defect states showed higher catalytic the Pt loading and obtain highest catalytic rates for solar fuel
activity (see Figure S1 in the Supporting Information). These production. While there has been variability in literature on the
defect states are likely electron traps14 since they are amount of Pt metal doping to obtain the highest photocatalytic
energetically closer to the conduction band (∼0.5−1 eV) rates,7,12,20 we obtained the highest yield (23.1%, Figure 2a,b)
than the valence band (∼2.2−2.7 eV, Figure 1d). Therefore, for 1 wt % (percentage by weight) Pt under 1 Sun (AM1.5)
multiphonon emission leading to relaxation of photogenerated simulated irradiation. On studying the effect of metal doping on
energetic states, we did not observe any additional states in the
charges in these sub-bandgap states18 supports their role as
electronic DOS of 0.5 wt % Pt-doped TiO2 nanoparticles, when
electron traps. While adsorption of oxygen (or CO2 here) is the
compared to unmodified TiO2 (Figure 2c compared to Figure
rate-limiting step in photocatalytic reactions and oxygen
1d,S1). However, the analysis of STS data on 1 wt % Pt-doped
vacancies with similar energies (∼0.5−1 eV below conduction TiO2 clearly revealed bandtail states within the TiO2 semi-
band) have been predicted to catalyze the reaction,7 our conductor bandgap (Figure 2d). Since the photocatalytic
measurements suggest that photogenerated electron trapping activity of CO2−water reduction was enhanced several fold
can also lead to similar adsorption of reactant species. Similar due to the introduction of Pt metal (Figure 2a), it is likely
findings also reported by other groups7−13 can likely explain the caused due to an increase in adsorption of oxygen atom on the
improved photocatalytic activity due to TiO2 modifications, Pt interface.12 But as we increased the photogeneration rate by
leading to visible light absorption and improved catalytic increasing the light intensity to enhance the photocatalytic
activity in TiO2 nanostructures.7,9,13 Therefore, three-dimen- conversion, we saw a decrease in catalytic yields (Figure 2b).
sional mapping of electronic-DOS in TiO2 nanoparticles (using This observation can be explained since the bandtail states
individual current−voltage scans as shown in Figure 1e) helped extended throughout the entire Pt-doped TiO2 bandgap, likely
us in spatially correlating these electronic states responsible for facilitating the trapping and recombination of both photo-
accelerated photocatalytic yields for CO2−water reduction. generated electron and holes (Figure 2d). Therefore, while the
600 dx.doi.org/10.1021/nl403783d | Nano Lett. 2014, 14, 597−603
Nano Letters Letter
Figure 4. Characterization of our new photocatalyst TiO2−CIS. (a) Chemical composition and (b) photocatalytic yields and selectivity for CO2−
water reduction obtained using our designed TiO2−CIS photocatalyst. The new TiO2−CIS catalyst (see Methods in Supporting Information)
combines high selectivity (>70% selectivity for ethane production) with high yield (>5% for 5 Sun illumination) and represents an important step
toward rational design of new catalyst using energy level tuning, as shown in this photocatalysis data. (c) Electronic DOS tuning can be easily
achieved by tuning the energy levels of the quantum confined CIS nanoparticle, as shown in STS data of the composite nanoparticle. 1.7 eV bandgap
CIS nanocrystals (∼3.8 nm in size) were attached to TiO2 nanoparticles resulting in desired energy levels below the conduction band of TiO2
semiconductor and possible tuning the products formed from photocatalytic reduction of CO2−water.
Pt dopant may increase the adsorption of reactant molecules,12 trap which is likely responsible for an enhancement in
enhancing the photocatalysis rates (Figure 2a), it also increases recombination rate of photogenerated electron and results in
the recombination of photogenerated electrons and holes, lower photocatalytic yields at higher photogeneration rates
leading to an overall reduction in observed photogenerated (Figure 3c). To develop a more stable catalyst (see stability
hydrogen yields. This kind of metal-mediated enhanced measurements in Supporting Information) which can produce a
recombination is not uncommon in semiconductor−metal specific product by using the energetic alignment of states, we
interfaces.21 synthesized copper−indium sulfide (CIS) nanocrystals22
To prevent metal dopants from acting as recombination sites, attached to TiO2 nanoparticles.
we incorporated other inexpensive copper-based nanoparticles, We studied the electronic DOS of our new composite
with staggered energy levels (Figure 3c, Figure 4c), to add the nanoparticle photocatalyst (TiO2−CIS, using single nano-
desired electronic energy states in TiO2 composite photo- particle scanning tunneling spectroscopy (Figure 4c). The
catalyst. This can also tune the energetic pathway and hence the Fermi level in the TiO2−CIS quantum dots was expected to
likely products formed from CO2−water reduction. To develop favor multielectron reduction to alkane hydrocarbon fuels
a robust and stable photocatalyst (see stability measurements in (Figure 4c). When we conducted photocatalytic reduction of
Supporting Information), we utilized copper-oxide nano- CO2−water, we observed evolution of hydrocarbon gases like
particles grown on surface of TiO2 as “dopants”. The new ethane, propane, and so forth, as shown in Figure 4a. However,
composite nanoparticle has combined energetic states from the the most striking result was that CIS nanoparticles with a
TiO2 nanoparticle and the Cu2−xO (0 < x < 1) semiconductor bandgap ∼1.7 eV demonstrated very high selectivity (>70%)
(Figure 3c). On illumination with AM1.5 light spectrum, there for the formation of ethane gas (Figure 4a,b), as predicted by
is a rapid increase in CO2−water reduction, as compared to our study (Figure 4c). Moreover, when we increased the
unmodified TiO2 photocatalyst (Figure 3a,b compared to photon flux and the lower energy states get filled up, hot
Figure 1b,c). However, the overall incident photon to fuel electrons favor formation of acetaldehyde, and the predominant
conversion efficiency (Figure 3c) is low because of variable product switches from alkanes to aldehyde (Figure 4a,b). With
oxidation state of copper-oxide nanoparticles. Since Cu2O high selectivity, hot-electron tunability of the reaction product,
(bandgap 2.1 eV) nanoparticles can be oxidized to CuO combined with yield of ∼5% for this multielectron reduction
(bandgap 1.2 eV) nanoparticles, we have observed Cu2−xO process using a small concentrator (∼5 Sun illumination, Figure
nanoparticles with bandgap ranging from 1.87 to 1.37 eV 4b), this new inexpensive photocatalyst shows a lot of promise
(Supplementary Figure S2). This change in oxidation state of for improved catalytic reduction of CO2−water into selective
the Cu2−xO semiconductor can give rise to a charged defect or solar fuels (see stability measurements in the Supporting
601 dx.doi.org/10.1021/nl403783d | Nano Lett. 2014, 14, 597−603
Nano Letters Letter
Figure 5. Photophysics in TiO2−CIS nanoparticle catalyst. (a) Current sensing atomic force microscope (CS-AFM) spectrum of a single TiO2−CIS
nanoparticle in two regions marked “1” (red curve, TiO2 showing electron injection) and “2” (black curve, CIS showing presence of holes). The inset
shows comparable current spectrum of unmodified TiO2 nanoparticle using gold coated current sensing AFM tip. (b) CS-AFM image of a single
photocatalyst nanoparticle. The image is acquired by applying a small bias between the gold coated AFM tip, and the nanoparticles dispersed on
conductive indium−tin−oxide (ITO) substrate. Cross-section slices of the 3D spectrum obtained clearly shows spatial separation of injected
electrons and holes in the composite TiO2−CIS photocatalyst. (c) Schematic of the proposed model explaining the photophysics of the
multielectron reduction on photocatalyst. n and p represent photogenerated electron and hole concentration per nanoparticle, on illumination with
photons above the TiO2 bandgap. Nd represents the total number of defects/dopants per nanoparticle, and the dashed line represents recombination
pathways for photogenerated (with rate constant α) and trapped electrons (with rate constant α) with bandedge photogenerated holes in the
photocatalyst. (d) Relationship between photocatalytic rate of solar fuel produced by CO2−water reduction and the rate of photogeneration
(expressed as comparable solar irradiance). The decrease in photocatalytic rate of TiO2−Pt doped nanoparticles and the superlinear increase in
TiO2−CIS solar fuel production rate is explained well using our simple model.
Information). Moreover, this new class of quantum dot (QD- (rate rt) in dopant/defect centers and back to conduction band
TiO2) based composite photocatalysts developed here allows (rate rdt), and recombination between electrons in dopant/
careful tuning of electronic DOS in the composite photo- defect centers and photogenerated holes (rate constant β)
catalyst, by utilizing quantum confinement effect and hot- change the steady state concentration of electrons and hence
electron states.23−25 rate of photocatalytic reduction (R = r × n, since reduction is
To improve our insights into the mechanism governing high the rate determining step). Using charge neutrality (p = n + nd)
efficiency of photocatalytic yields in the new TiO2−CIS and charge balance for electrons (dp/dt = g − αpn − rtn(1 −
nanoparticle, we utilized current sensing atomic force nd/Nd) + rdtnd − rn = 0 at steady state) and holes (dp/dt = g −
microscopy (CS-AFM) (Figure 5a,b,S3 see Methods in αpn − βpnd − rn)), we get a relationship between the
Supporting Information). We observed a clear spatial photogeneration rate (g) and the rate of photocatalysis (R).
separation of electron and hole, as shown in Figure 5a and b. Since the photocatalysis rate depends on rate of recombination
This spatial separation of photogenerated charges can lead to (α,β), trapping (rt,rdt), and fraction of electrons in defects (nd,
longer lifetimes (due smaller overlap of electron and hole wave compared to n), we identified four regimes for change in
functions) and therefore likely higher photocatalytic yields. We photocatalysis rate (r × n) with rate of photogeneration (g =
developed a simple model (Figure 5c) to explain the αn(n + nd) + βnd(n + nd) + rn). In the first regime, we identified
photophysics governing the photogeneration, recombination, when α ≫ β (recombination between photogenerated electron
charge trapping, and the multielectron photocatalytic reduction and hole) and n ≫ nd, g ∝ n2, and R(= r × n) ∝ g0.5
of adsorbed species in these dopant/defect centers on TiO2 (bimolecular recombination) or expected square root depend-
nanoparticles. Upon irradiation with light above the TiO2 ence between rate of photocatalysis and photogeneration rate.
nanoparticle bandgap (photogeneration rate g), photogener- In photocatalysis mediated by TiO2 nanoparticles, since traps
ated electrons (n) and holes (p) are excited in the conduction and defects play an important role, we also modeled (β ≫ α, g
and valence “band” of the nanoparticle. A fraction of ≈ βnd(n + nd)) trap mediated recombination. The second class
photogenerated electrons is transferred/trapped (nd) in differ- of results predict when n > nd (Nd is small and/or rt is slow), a
ent dopant/defect centers (total number Nd) in these linear relationship exists between photocatalytic rate (R = r ×
nanoparticles. Different photophysical events like recombina- n) and incident photon rate (g ≈ βndn, nd is constant,
tion between photogenerated electrons and holes (rate unimolecular recombination, Figure 1c observed in TiO2
constant α), transfer/trapping of photogenerated electrons nanoparticles). In cases when Nd is large (compared to n,p)
602 dx.doi.org/10.1021/nl403783d | Nano Lett. 2014, 14, 597−603
Nano Letters
■
Letter
■
2012, 134, 11276.
(21) Robbins, D. J.; Dean, P. J. Adv. Phys. 1978, 27, 499.
ASSOCIATED CONTENT (22) Panthani, M. G.; Akhavan, V.; Goodfellow, B.; Schmidtke, J. P.;
*
S Supporting Information
Dunn, L.; Dodabalapur, A.; Barbara, P. F.; Korgel, B. A. J. Am. Chem.
Soc. 2008, 130, 16770.
General methods, measuring photocatalytic activity, X-ray (23) Nozik, A. J. Annu. Rev. Phys. Chem. 2001, 52, 193.
diffraction, X-ray photoelectron spectroscopy, STM and STS, (24) Peng, X. G.; Manna, L.; Wang, W. D.; Wickham, J.; Scher, E.;
CSAFM, TEM and optical characterization for copper indium Kadavanich, A.; Alivisatos, A. P. Nature 2000, 404, 59.
sulfide nanoparticles, advantage of STM/CSAFM character- (25) Talapin, D. V.; Lee, J. S.; Kovalenko, M. V.; Shevchenko, E. V.
ization in air, photocorrosion studies of our TiO2−CIS catalyst, Chem. Rev. 2010, 110, 389.
and supplementary figures. This material is available free of
charge via the Internet at http://pubs.acs.org.
■ AUTHOR INFORMATION
Corresponding Author
*E-mail: pnagpal@colorado.edu.
Author Contributions
V.S. and I.J.C.B. contributed equally.
Notes
The authors declare no competing financial interest.
603 dx.doi.org/10.1021/nl403783d | Nano Lett. 2014, 14, 597−603