Efficient hot-electron transfer by a plasmon-induced interfacial charge-transfer transition

See allHide authors and affiliations

Science  07 Aug 2015:
Vol. 349, Issue 6248, pp. 632-635
DOI: 10.1126/science.aac5443

Improving electron harvesting

Small metal nanostructures generate electrons from light by creating surface plasmons, which can transfer “hot electrons” to a semiconductor. The efficiency of this process, however, is often low because of electron-electron scattering. Wu et al. demonstrate a pathway that allows the plasmon to directly excite an electron in a strongly coupled semiconductor acceptor (see the Perspective by Kale). Cadmiun selenide nanorods bearing gold nanoparticles on their ends strongly damped plasmons via interfacial electron transfer with a quantum efficiency exceeding 24%.

Science, this issue p. 632; see also p. 587


Plasmon-induced hot-electron transfer from metal nanostructures is a potential new paradigm for solar energy conversion; however, the reported efficiencies of devices based on this concept are often low because of the loss of hot electrons via ultrafast electron-electron scattering. We propose a pathway, called the plasmon-induced interfacial charge-transfer transition (PICTT), that enables the decay of a plasmon by directly exciting an electron from the metal to a strongly coupled acceptor. We demonstrated this concept in cadmium selenide nanorods with gold tips, in which the gold plasmon was strongly damped by cadmium selenide through interfacial electron transfer. The quantum efficiency of the PICTT process was high (>24%), independent of excitation photon energy over a ~1–electron volt range, and dependent on the excitation polarization.

The surface plasmon resonance (SPR) of metallic nanostructures has been widely used to improve the efficiency of photovoltaics (15), photocatalysis (6, 7), and photodetectors (8, 9), either by increasing light absorption through enhanced local fields near the metal nanostructures (10) or by plasmon-induced charge transfer from the excited metal (1113). The latter enhancement mechanism suggests the possibility of using plasmonic metal nanostructures as light absorbers with broad spectral tunability, large absorption cross sections, superior long-term stability, and low-cost colloidal synthesis (14, 15). Thus far, all reported plasmon-induced charge-separation processes have been believed to occur through a conventional plasmon-induced hot-electron transfer (PHET) mechanism (Fig. 1A). During PHET, a plasmon decays into a hot electron–hole pair within the metal via Landau damping on time scales of a few to tens of femtoseconds (1618); this is followed by the transfer of the hot electron into adjacent semiconductors or molecules. Hot-electron transfer competes with electron relaxation through rapid electron-electron scattering in the metal’s conduction band (CB) on time scales of hundreds of femtoseconds (1820). Efficient PHET requires interfacial charge separation on an even faster time scale, which is difficult to realize in many semiconductor-metal hybrid materials. Thus, the reported efficiencies for devices based on plasmon-induced charge-separation concepts are too low for practical applications (68).

Fig. 1 Metal-to-semiconductor charge-separation pathways.

(A) Conventional PHET mechanism, in which a photoexcited plasmon (SP, blue ellipsoid) in the metal decays into a hot electron–hole pair (solid and open red circles in the dotted ellipsoids) through Landau damping, followed by injection of the hot electron into the CB of the semiconductor. The electron-hole pair has a broad distribution of initial electron and hole energies; only two are shown for clarity. (B) Optical excitation of an electron in the metal directly into the CB of the semiconductor through the DICTT pathway. (C) The newly demonstrated PICTT pathway, where the plasmon decays by directly creating an electron in the CB of the semiconductor and a hole in the metal. VB is the semiconductor valence band and hv indicates the excitation photons.

Metal-to-semiconductor hot-electron transfer efficiencies can be enhanced if the competition with ultrafast electron-electron scattering in the metal can be avoided. One approach is to create a direct metal-to-semiconductor interfacial charge-transfer transition (DICTT) that can be directly excited to promote an electron from the metal into the semiconductor CB (Fig. 1B). Such transitions between metal adatoms and semiconductor electrodes have been reported (21), as have metal-to-adsorbate resonances for CO adsorbed on Pt nanoparticles (NPs) (22, 23) and Cs atoms adsorbed on Cu(111) (24, 25). However, these interfacial transitions are often too weak as compared with bulk metal transitions or plasmon bands (2225) and cannot serve as efficient light-harvesting pathways. Ideally, a desirable photoinduced hot-electron transfer pathway would combine the strong light-absorbing power of plasmonic transitions with the superior charge-separation properties of the DICTT mechanism (Fig. 1C). In this plasmon-induced metal-to-semiconductor interfacial charge-transfer transition (PICTT) pathway, the metal plasmon serves as a light absorber, but strong interdomain coupling and mixing of the metal and semiconductor levels lead to a new plasmon decay pathway: the direct generation of an electron in the semiconductor and an electron hole in the metal. This model is an extension of the chemical interface damping model that has been proposed to account for adsorbate-induced broadening of the plasmon bands of metal NPs (2628).

We proposed and experimentally demonstrated the PICCT pathway in colloidal quantum-confined CdSe-Au nanorod (NR) heterostructures. Strong Au-CdSe interactions led to strong damping of the plasmon band in the Au tip. The proposed pathway was verified by observing highly efficient plasmon-induced Au-to-CdSe charge separation with >24% quantum efficiencies upon excitation of the Au tip. Measurements of transient absorption anisotropy showed more efficient Au-to-CdSe charge transfer when the plasmon was excited along the NR axis, consistent with the PICTT mechanism. The charge-separation efficiencies were independent of excitation photon energy, a result that is inconsistent with the conventional PHET mechanism and supports the proposed PICTT pathway.

Colloidal CdSe-Au NRs were synthesized according to a published procedure (29, 30). Representative transmission electron microscope (TEM) images of CdSe and CdSe-Au NRs (Fig. 2A and figs. S1 and S3) showed well-defined dumbbell-like morphologies for CdSe-Au, with two Au NPs at both ends of a single-crystalline CdSe NR. Static absorption spectra of CdSe and CdSe-Au NRs dispersed in chloroform are shown in Fig. 2B. The discrete absorption peaks of CdSe NRs at ~480 and ~582 nm are attributed to the 1Π and 1Σ exciton bands, respectively, arising from quantum confinement in the radial direction (31).

Fig. 2 Plasmon-induced metal-to-semiconductor charge-transfer transition in CdSe-Au NRs.

(A) A representative TEM image of CdSe-Au NRs (inset: a representative high-resolution TEM image). (B) Absorption spectra of CdSe NRs, CdSe-Au NRs, and CdSe QD-Au dimers dispersed in chloroform. The gray dashed line is the difference spectrum between the absorptions of the CdSe-Au NRs and the CdSe QD-Au dimers. (C) Absorption spectra (with the y axis plotted on a logarithmic scale and shifted by +0.01 to avoid negative values) of (i) isolated Au NPs and (ii) CdSe and CdSe-Au NRs with their first excitonic peak positions at 555, 582, and 605 nm. The absorption spectra of CdSe-Au NRs show an onset at ~1450 nm (0.85 eV). (D) Schematic electronic structure of a CdSe-Au NR, composed of a strongly damped Au tip with broadened electronic levels and a central region with relatively unperturbed discrete levels (1σe, 1πe, 1σh). The green dashed arrow indicates the interband transition in the visible (Vis) spectrum, the red arrow indicates the intraband transition in the IR spectrum, and the green solid arrow indicates electron transport in the NR.

Compared with free Au NPs, the SPR band of the Au tips (diameter, ~4.1 nm) was strongly damped and showed a continuous absorption feature extending to the near-infrared (IR) spectrum, consistent with previous observations (29, 32). It has been suggested that such an extreme change in the SPR band cannot be accounted for by dielectric effects alone; rather, it requires strong electronic interactions between the CdSe and Au domains through some mechanism that is yet to be understood (32). We observed ultrafast quenching (<100 fs) of excitons in the CdSe NR by the Au tips (fig. S5), consistent with previous results for related CdSe-Au NRs (33). In contrast to the pronounced broadening of the Au plasmon band, changes in the CdSe exciton bands were not apparent (Fig. 2B). The CdSe NR was much longer than the strongly interacting metal/semiconductor interface region; the ends of the CdSe NR (at the CdSe/Au interface) were probably strongly perturbed, whereas the center was relatively unperturbed (34).

To mimic the tip region of the CdSe-Au NRs, we synthesized CdSe quantum dot (QD)–Au dimers using CdSe QDs with a lowest-energy exciton band similar to that of the CdSe NRs (fig. S2). In these dimers, the excitonic peaks of the QDs and the Au plasmon band were completely damped (Fig. 2B), showing a continuous absorption feature that closely matched the feature observed in the CdSe-Au NRs. Subtracting this feature from the CdSe-Au NR absorption spectrum revealed slightly blue-shifted exciton bands relative to free CdSe NRs (Fig. 2B), which can be attributed to NR etching during the growth of the Au tips (29). Thus, the electronic structure of CdSe-Au NRs can be viewed as a combination of the strongly interacting tip region, which resembles the CdSe QD-Au dimers (with strongly damped plasmon and exciton bands), and the center part, which is similar to the unperturbed CdSe NRs (Fig. 2D).

The broad, featureless, near-IR absorption spectra of the CdSe-Au NRs showed an onset at ~1450 nm (0.85 eV) for three NRs with 1Σ exciton bands at 555, 582, and 605 nm (with a corresponding shift of ~150 meV in the CB-edge position) (Fig. 2C). This suggests that the transition cannot be attributed to the DICTT mechanism (Fig. 1B), which should have an onset wavelength that shifts with the CdSe CB edge. Instead, we attribute the broad near-IR absorption feature to a strongly damped Au plasmon caused by the strong mixing of Au and CdSe electronic levels. Such strong interaction activates the PICTT plasmon decay pathway (Fig. 1C), which is not possible in isolated Au NPs.

In the PICTT pathway, the damped plasmon decays via direct excitation of an interfacial electron-hole pair (with an electron in CdSe and a hole in Au). Direct evidence to support this proposed mechanism was obtained through ultrafast transient absorption (TA) studies. In these studies, a pump laser with a photon energy below the CdSe band gap was used to excite the Au tip, and the electron transferred to the CdSe domain was probed through the bleaching of the 1Σ exciton band in the visible spectrum (caused by state filling of the 1σe level) and/or the 1σe to 1πe intraband absorption in the mid-IR spectrum (Fig. 2D) (35). The assignment of these spectral signatures was confirmed by comparing transient spectra of CdSe NRs and CdSe-benzoquinone (electron-acceptor) complexes.

The kinetics of the 1Σ exciton bleach at ~580 nm and the intraband absorption at ~3000 nm matched closely in free CdSe NRs, and electron transfer from CdSe to adsorbed electron acceptors led to a faster decay of both spectral features in the NR electron-acceptor complexes (fig. S4). The TA spectra of CdSe-Au NRs (Fig. 3A) showed a pronounced 1Σ-exciton-bleach feature at ~575 nm, indicating the formation of CdSe CB electrons through the excitation of the damped Au plasmon band at 800 nm. The bleach overlapped with a broad positive TA feature that was also present in the TA spectra of CdSe QD-Au dimers (fig. S6). This broad feature was subtracted from the total TA signal to obtain the TA spectra (fig. S8) and the kinetics of the 1Σ exciton bleach (Fig. 3C and fig. S8). Excitation at 800 nm also generated an intraband absorption feature at 3000 nm. The formation and decay kinetics of this signal and the 1Σ-exciton-bleach signal agreed well (Fig. 3B), further confirming the presence of CB electrons in CdSe.

Fig. 3 Plasmon-induced charge separation in CdSe-Au NRs.

(A) Two-dimensional pseudo-color plot of TA spectra for CdSe-Au NRs at 800-nm excitation [x axis, probe wavelength; y axis, pump-probe delay; colors, change in absorbance (ΔAbs), shown in milli–optical density units (mOD)]. (B) Intraband absorption (probed at ~3000 nm, red circles) and 1Σ-exciton-bleach (~580 nm, green dashed line) kinetics of CdSe-Au NRs after 800-nm excitation. A negligible intraband absorption signal is apparent in a control sample of a mixture of CdSe NRs and Au nanoparticles (gray dashed line). The black solid line is a multiexponential fit of the kinetics.

These intra- and interband signals were absent in control samples composed of a mixture of CdSe NRs and Au NPs (Fig. 3B and fig. S7). Fitting these kinetics yielded a formation time of ~20 ± 10 fs and a biexponential decay with a half-life of ~1.45 ± 0.15 ps, which corresponded to plasmon-induced hot-electron transfer and charge-recombination times, respectively (35). Such an ultrafast charge-separation time is consistent with the PICTT mechanism, in which the decay of a plasmon instantaneously generates an electron in CdSe (giving rise to the observed inter- and intraband features) and a hole in Au near the CdSe/Au interface. The injected electron quickly relaxes back to the Au with a recombination time of 1.45 ps, which indicates a negligible band-bending–induced recombination barrier at the CdSe/Au interface.

The transient quantum yields (QY) of Au-to-CdSe charge separation as a function of excitation wavelength are shown in Fig. 4. These QY values were determined by the peak amplitude of the CdSe intraband absorption signal (which averaged between 0.2 and 0.4 ps; see the supplementary materials for details). The pump wavelength (energy) was tuned over a >1-eV range below the band gap of CdSe NRs: 670 nm (1.85 eV), 710 nm (1.75 eV), 750 nm (1.65 eV), 800 nm (1.55 eV), 1160 nm (1.07 eV), 1340 nm (0.92 eV), and 1550 nm (0.80 eV). Within experimental errors, the measured QY values were constant (~24%) above ~0.85 eV (Fig. 4), the onset of the near-IR absorption feature shown in Fig. 2B. Below the onset, no measurable electron signals were observed because of a lack of photon absorption.

Fig. 4 Quantum yield of wavelength-independent hot-electron transfer.

QY values of plasmon-induced charge separation as a function of excitation photon energies are shown (red open circles and green solid triangles, measured with PbS and Cd3P2 QDs as calibration samples, respectively; see the supplementary materials for details), with predictions according to various Fowler models: Eq. 1(blue dashed line), Eq. 2 (green dashed line), and Eq. 3 (purple dashed line). The black solid line is a step function with an onset at ~0.85 eV.

In the conventional PHET mechanism (Fig. 1A), the excitation energy dependence of the charge-separation QY has been shown to follow Fowler’s equation (36)

Embedded Image(1)

where ℏω is the energy of the excitation photon, EB is the barrier height between the metal and the semiconductor, and EF is the Fermi energy of the metal. Most devices based on hot-electron transfer reported to date have been shown to follow this model (8, 3739). For small NPs, QY values of metal-to-semiconductor photoemission generally follow the same functional form, but they can be enhanced by a factor C through a geometric effect (40, 41) and a lowering of the escape barrier (42)

Embedded Image(2)

Recently, it has also been found that the momentum conservation requirement can be relaxed if electrons only scatter at the semiconductor/metal interface (43). Under this condition, QY is determined by

Embedded Image(3)

The predicted QY values according to Eqs. 1 to 3 are shown in Fig. 4. For Eq. 2, we chose the value of C so that the predicted QY at 1.85 eV agreed with the experimental results. The estimated value of EB was 0.7 to 1.4 V because of uncertainty in the reported band-edge positions (30). We used a value of 0.85 V to allow more convenient comparison with the experimental results. The Fowler-type conventional hot-electron transfer models predict an increasing QY at higher excitation energies (because of the increase in hot electrons with energies above the semiconductor CB edge) and are inconsistent with our experimental data.

The measured QY values are consistent with the PICTT pathway. In this pathway, the plasmon decays by direct excitation of an electron from Au to CdSe, and the QY is independent of the excess energy of the electron above the CB edge. The strongly damped plasmon bandwidth is probably dominated by homogeneous broadening: As long as the excitation energy is above the absorption threshold (~0.85 eV), the same plasmon is excited, and therefore the charge-separation QY is independent of excitation energy. The estimated full width at half maximum of the damped plasmon band, ~1.6 eV (fig. S11), corresponds to a plasmon dephasing time of ~0.8 fs, which is consistent with the observed fast hot-electron transfer time. Furthermore, the observed QY is about an order of magnitude higher than the reported value for CdS-Au NRs (2.75%) (35). In CdS-Au NRs, the Au plasmon band is weakly perturbed, and the plasmon-induced hot-electron transfer occurs through the conventional PHET mechanism (Fig. 1A), where the competition of hot-electron transfer with ultrafast relaxation reduces its efficiency. Compared with CdSe-Au NRs, the hot-electron transfer time in CdS-Au NRs is noticeably slower (90 ± 20 fs) (fig. S9). The microscopic origin of the dramatic differences in plasmon damping and hot-electron transfer between CdS-Au and CdSe-Au NRs remains unclear and will be examined in future studies.

The observed QY values in CdSe-Au NRs are less than unity for at least two reasons. First, PICTT is probably not the exclusive decay channel for the strongly damped plasmon. The plasmon could also be damped in the Au domain, thereby proceeding via the less efficient conventional PHET pathway (35). Second, the electron generated in CdSe could either quickly relax back into the Au or escape into the center of the CdSe NR (and recombine on the 1.4-ps time scale); only the latter process was detected in our measurements. Thus, the measured QY values represent a lower limit of plasmon-induced electron transfer in this system.

Further insight into the nature of the PICTT pathway can also be obtained through polarization-dependent TA studies. Plasmons polarized in the direction parallel to the NR may be more strongly coupled to the CdSe than those polarized in the direction perpendicular to the NR. Because the optical transition in the CdSe NRs is strongly polarized (31, 44), the TA signal should depend on the relative polarization of the pump and the probe beams. This polarization dependence can be quantified by the anisotropy (r) of the TA signals, so that r = (SHHSVH) / (SHH + 2SVH), where SHH and SVH are TA signals with the polarizations of the pump and probe beams parallel (SHH) or perpendicular (SVH) to each other. For band-edge (590-nm) excitation of free CdSe NRs (without Au tips), the 1Σ-exciton-bleach signal amplitude for SHH was larger than for SVH (Fig. 5A). The calculated anisotropy was 0.12 and showed negligible decay over 10 ps (Fig. 5A, inset), which indicates that the band-edge–absorption transition dipole has 70% axial and 30% radial components (fig. S12; see the supplementary materials for details), consistent with previous reports (45). The intraband transition between 1σe and 1πe showed negligible anisotropy (fig. S13). Thus, we used the 1Σ exciton bleach to probe anisotropy in CdSe-Au NRs. For 800-nm excitation of CdSe-Au NRs, the resulting 1Σ-exciton-bleach signal amplitude for SHH was larger than for SVH, with an anisotropy value of 0.10 (Fig. 5B). This result suggests a more efficient hot-electron transfer by plasmons polarized parallel to the NR axis, consistent with the PICTT mechanism. In contrast, in CdS-Au NRs where the conventional PHET mechanism dominates (35), we detected negligible anisotropy in the CdS 1Σ-exciton-bleach signal generated by exciting the Au plasmon band (fig. S15).

Fig. 5 Transient absorption anisotropy of CdSe and CdSe-Au NRs.

(A) 1Σ-exciton-bleach kinetics (probed at ~580 nm) in free CdSe NRs after band-edge (590-nm) excitation, with pump and probe beams with parallel (HH, horizontal pump and horizontal probe, red solid line) and perpendicular (VH, vertical pump and horizontal probe, blue dashed line) polarizations. Calculated anisotropy, r, is shown in the inset as a function of pump-probe delay. (B) 1Σ-exciton-bleach kinetics (probed at ~575 nm) in CdSe-Au NRs after 800-nm excitation, with pump and probe beams with parallel (HH, red solid line) and perpendicular (VH, blue dashed line) polarizations. r is shown in the inset.

Finally, we showed that in the presence of sacrificial electron donors (S2–), the electrons injected into CdSe could be used to reduce methyl viologen, a well-known redox mediator for solar fuel generation (46), with a steady-state QY of >0.75% (fig. S18; see the supplementary materials for details). In light of the high QY values measured for transient charge separation (>24%), the relatively low steady-state QY can be attributed to ultrafast charge recombination. Charge recombination can be retarded by creating a built-in electric field at the semiconductor/metal interface using longer NRs (>100s nm) or by applying a large external bias (8).

Strong mixing of Au and TiO2 electronic levels has been reported in a recent computational study of Au-cluster–sensitized TiO2 NPs (47). Strongly broadened Au plasmon bands and efficient plasmon-induced hot-electron transfer were also observed in Au-NP–sensitized TiO2 nanocrystalline thin films (11). Thus, the PICTT mechanism reported here is potentially a general phenomenon at metal/semiconductor and/or metal/molecule interfaces. PICTT may present a new opportunity to circumvent energy-loss channels in metal nanostructures and greatly increase the efficiencies of devices based on plasmonic light-absorption materials.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S18

Tables S1 to S5

References (4872)


  1. Materials and methods are available as supplementary materials on Science Online.
  2. ACKNOWLEDGMENTS: This work was funded by the U.S. Department of Energy, Office of Basic Energy Sciences, Solar Photochemistry Program (grant DE-FG02-12ER16347). J.R.M. acknowledges financial support from NSF (grant CHE-1213758). Scanning transmission electron microscope and energy-dispersive x-ray spectroscope images were acquired using an FEI Tecnai Osiris electron microscope purchased with support from NSF (grant EPS-1004083).
View Abstract

Stay Connected to Science

Navigate This Article