Size Matters: Cocatalyst Size Effect on Charge Transfer and

Dec 13, 2017 - A set of Ni decorated CdSe@CdS nanorods with different tip size were examined, and an optimal metal domain size of 5.2 nm was obtained...
0 downloads 15 Views 10MB Size
Download PDF
Letter Cite This: Nano Lett. 2018, 18, 357−364

pubs.acs.org/NanoLett

Size Matters: Cocatalyst Size Effect on Charge Transfer and Photocatalytic Activity Yifat Nakibli,† Yair Mazal,‡ Yonatan Dubi,‡ Maria Wac̈ htler,*,§ and Lilac Amirav*,† †

Schulich Faculty of Chemistry, Technion − Israel Institute of Technology, Haifa 32000, Israel Department of Chemistry and the Ilse Katz center for Nanoscale Science and Technology, Ben-Gurion University of the Negev, Beer-Sheva 8410501, Israel § Department Functional Interfaces, Leibniz Institute of Photonic Technology Jena, Albert-Einstein-Straße 9, 07745 Jena, Germany ‡

S Supporting Information *

ABSTRACT: Hybrid semiconductor−metallic nanostructures play an important role in a wide range of applications and are key components in photocatalysis. Here we reveal that the nature of a nanojunction formed between a semiconductor nanorod and metal nanoparticle is sensitive to the size of the metal component. This is reflected in the activity toward hydrogen production, emission quantum yields, and the efficiency of charge separation which is determined by transient absorption spectroscopy. A set of Ni decorated CdSe@CdS nanorods with different tip size were examined, and an optimal metal domain size of 5.2 nm was obtained. Remarkably, charge separation time constants were found to be nonvariant with metal tip size. It is proposed that electron transfer mechanism encompasses two consecutive but separate processes: slow charge migration along the rod toward the interface, followed by fast interface crossing of the electron from the semiconductor into the metal phase. The first migration step dominates the time constant for the charge separation process and is not affected by the metal size. The efficiency of charge separation on the other hand was found to be sensitive to metal size. It is suggested that Coulomb blockade charging energy and a size-dependent Schottky barrier contribute to the metal size effect on charge transfer probability across the semiconductor− metal nanojunction. These two opposing trends result in an optimal metal size domain for the cocatalyst. This work is expected to benefit a broad range of applications utilizing semiconductor−metal nanocomposites. KEYWORDS: Semiconductor metal hybrids, nanojunction, photocatalysis, catalyst size, hydrogen, nanoheterostructures, transient absorption

T

production, the photocatalyst design should only include a single cocatalytic site per each segment of the semiconductor capable of light excitation.3 This is to ensure that intermediates of the catalytic reaction are formed at close proximity and can complete the reaction during the life span of the intermediate. In addition, utilization of complex multicomponent heterostructure systems that are designed for improved charge separation requires precise control over spatial location of the catalyst.12,13 The importance of metal size and shape is well recognized from heterogeneous catalysis, where the metal−liquid interface is critical for the promotion of the reaction.14,15 However, the mechanism of photocatalytic reactions relies on the characteristics of an additional interface that is formed between the semiconductor and the metal. This interface will dictate the efficiency and dynamics of photoinduced charge transfer from

he solar-driven photocatalytic splitting of water into hydrogen and oxygen is a potential source of clean and renewable fuels. However, four decades of global research have proven this multistep reaction to be highly challenging. The design of effective artificial photocatalytic systems will depend on the ability to correlate the photocatalyst structure, composition, and morphology with its activity. The advantages of sculpting photocatalysts on the nanoscale were demonstrated recently with perfect 100% photon-to-hydrogen production efficiency, which was obtained for the photocatalytic water splitting reduction half reaction using a well-controlled nanoparticle-based artificial system.1,2 Though great care was devoted to the characteristics of the semiconductor component and its role in dictating the photocatalyst activity, only recently efforts were directed to understanding the effects of the metal cocatalyst morphology, in particularly metal location, number of catalyst domains,3,4 and catalyst shape 5 and size. 6−11 These aspects were demonstrated to be vital for optimization of the materials potential in promoting the reaction of interest. For example, it was found that for a multielectron reaction such as hydrogen © 2017 American Chemical Society

Received: September 29, 2017 Revised: December 4, 2017 Published: December 13, 2017 357

DOI: 10.1021/acs.nanolett.7b04210 Nano Lett. 2018, 18, 357−364

Letter

Nano Letters

adjustments and fine-tuning of the nickel precursor concentration (absolute and relative to the rods), reaction temperature and time, as well as the ligands and solvent that were used. In particularly, the nickel precursor concentration was found to be the key parameter for controlling the Ni tip size. Increasing the concentration enabled us to tune the Ni size from 2.3 to 10.1 nm (synthetic conditions are detailed in the SI) with a reasonable size distribution with a standard deviation of 0.6 nm for most samples, and up to 2.5 nm for the larger tips. All sizes of tips were synthesized in a one-step reaction. Figure 1 shows transmission electron microscope (TEM) micrographs of rods with five different tip sizes 2.3, 3.1, 5.2, 8.9, and 10.1 nm. This narrow tip size distribution, which is narrower than the typical distribution obtained for Pt tipping, enabled us to examine the effect of the metal domain size on photoinduced charge transfer dynamics and the activity for hydrogen production. The high level of control over deposition location and number of metal domains, enabled us to disentangle the size dependency from other surface coverage effects. The activity toward hydrogen production was examined for a set of Ni decorated CdSe@CdS nanorods, with different metal domain size. Because absolute activity of any given sample is strongly related to the rod morphology, throughout the course of this research the utilized rods were 50 nm long with 2.3 nm seed size. Great care was devoted to maintain all parameters as identical with the exception of the Ni tip size. Solutions containing about 1015 rods suspended in water with isopropyl alcohol (10% by volume) acting as a hole scavenger (electron donor) at neutral pH conditions were placed in a custom-built gastight reaction cell purged with argon. The samples were then illuminated with a 455 nm LED adjusted to 50 mW (equivalent to a photon flux of 1.15 × 1017 photons/sec). The evolving hydrogen was analyzed using an online gas chromatograph equipped with a thermal conductivity detector. Operation in continuous flow mode allowed for direct determination of the gas production rate. The apparent quantum efficiency of the sample, which is defined as QE = 2NH2/Nhυ, was determined by quantifying the amount of evolved hydrogen at a given photon flux. The results for the apparent quantum efficiency, as obtained from the experiment for a set of different Ni tip sizes, are presented in Figure 2 in dark green. It is assumed that only photons absorbed by the semiconductor portion lead to hydrogen generation. As the metal component itself absorbs light, rods that are decorated with metal tips absorb smaller portion of the illuminated light and generate fewer charge carriers. Hence, the weight of the metal contribution to the absorption, which increases with the size of the metal tip, is filtered from the data, in order to allow genuine comparison between the activities of different samples. Quantum yields corrected for this internal filter effect, accounting for the metal absorption, are presented in Figure 2 in light green. For details concerning the applied correction procedure, the reader is referred to the Supporting Information. The experimental photocatalytic quantum efficiency for hydrogen production ranged from 1.2% to 23.4% with an optimal Ni size of 5.2 nm. Such increase with metal domain size at the lower size regime that is followed by a sharp decrease in activity as size is grown further is in good agreement with previously reported trends.6,11 This trend is more pronounced when the metal absorption is taken into account and corrected for in the quantum efficiency calculation. A rise in activity from 7 ± 0.4% for 2.3 nm tip, through 13.7 ± 0.8% for 3.1 nm tip to

the light sensitizer component into the catalytic domain. Yet, only paucity of research is focused on the effect of the catalyst size6−11 and shape5 and the role they play in affecting the efficiency. An optimal metal domain size for the hydrogen reduction catalyst was found for Pt9,10 and Ni11 cocatalysts that were grown on CdS nanostructures (rods, mesoporous particles, and sheets). In these studies, the metal size contribution was not fully separated from surface coverage effects (i.e., location and number of domains), and the size dependency was not sufficiently explained. Banin and co-workers6 confirmed the existence of an optimal metal domain size, working with Autipped CdS rods and attributed this observation to a trade-off between increasing rate for charge separation, and processes occurring at the metal−liquid interface. Here we reveal that it is the nanojunction formed between a semiconductor nanorod and metal nanoparticle that is sensitive to the size of the metal component. Careful examination of well-controlled Ni-decorated CdSe@CdS nanorods with tips of varying size enabled disentangling the size dependency from other obscuring effects, for example, coverage effects. In order to elucidate and better understand the catalyst size contribution, we correlated the activity toward photocatalytic hydrogen reduction with emission quantum yields, charge transfer dynamics, and the quantity of charge separation (measured with transient absorption spectroscopy). We further introduce a theoretical model that is corroborated with a calculation that simulates the photocatalytic results. The implications of our work go beyond photocatalysis and are of significance to the fundamental understanding of nanoscale semiconductor−metal hybrid characteristics. These nanocomposites are projected to play an important role in a wide range of technologies in applications such as solar and fuel cells, electronics, and photonics.1,16−20 Thus, the obtained insights on the size dependency of the nanojunction would be valuable for application-based design. For the light absorption and excitation unit, we employ a well-controlled nanoparticle-based artificial system,1 which consists of a cadmium selenide (CdSe) quantum dot embedded asymmetrically within a cadmium sulfide (CdS) quantum rod.21−24 These rods are decorated with metal cocatalyst. In such a structure, holes are three-dimensionally confined to the cadmium selenide,25,26 whereas the delocalized electrons are transferred to the metal tip.27−29 This architecture facilitates control over the degree of charge separation via tuning of the seed size, and rod dimensions.30−32 With a single metal tip catalyst placed at the far side of the rod, the system enables efficient long lasting charge carriers’ separation,1,33,34 and the formation of distinct and spatially segregated reaction sites for the different redox half reactions. This structure has been widely studied optically and photocatalytically and is well characterized,25,26,30,31 making it a good model system to investigate structure performance correlation. We recently developed a new and improved synthetic protocol for a well-controlled deposition of Ni nanoparticle catalysts on such a photocatalytic semiconductor system.35 This active and affordable earth abundant material has already demonstrated great potential as a substitute cocatalyst for the costly noble metals.11,36−44 The full protocol is published elsewhere and is based on the reduction of nickel precursor (nickel (II) acetylacetonate) by oleylamine and trioctyl phosphine. Control over the Ni nanoparticle size, location and number of sites on the CdS surface was attained via 358

DOI: 10.1021/acs.nanolett.7b04210 Nano Lett. 2018, 18, 357−364

Letter

Nano Letters

Figure 2. Photocatalytic quantum efficiency for the water reduction half reaction obtained with CdSe@CdS nanorod photocatalysts decorated with different-sized Ni tips. Experimental quantum efficiency in dark green bars, and quantum efficiency corrected for metal absorption in light green bars.

The efficiency for the water splitting reduction half reaction relies on the number of electrons residing on the metal as well has their energetics. It further involves numerous aspects related to the metal−liquid interface (e.g., surface area and surface reactivity of the metal catalyst). In its essence, it provides a summarized account of the final outcome for all the separate charge transfer processes that are involved in the photocatalytic reaction. In order to elucidate the mechanism and identify the source of the metal size effect, additional in depth inquiry is required. Photoluminescence with a maximum at 565 nm occurs from the lowest excitonic state localized in the CdSe seed45,46 and is quenched in the metal-tipped samples, as is reflected in the respective emission quantum yields (Table S2). Emission quantum yields, that are corrected in order to account for the metal absorption, can be found in Figure 3 and in Table S2 in

Figure 3. Emission quantum yields corrected for metal absorption contribution (triangles) and the deviation in the transient absorption signal intensity between Ni-tipped and bare rods at 471 nm (stars), which are directly correlated with the efficiency of charge separation after photoexcitation, are displayed as a function of metal domain size, alongside the measured corrected photon to hydrogen conversion efficiencies (circles), and simulated efficiencies (squares). Additionally the half-life time of the bleach decay at 458 nm with respect to the value at 1 ps determined by inspection of the data is given (pentagon).

Figure 1. TEM micrographs at two sets of magnifications presenting a set of CdSe@CdS rods with Ni tips of different size: 2.3 nm (A,B), 3.1 nm (C,D), 5.2 nm (E,F), 8.9 nm (G,H), and 10.1 nm (I,J).

50.9 ± 5.5% for the optimal 5.2 nm tip size, is followed by an decrease to 10.6 ± 1.6% for the 8.9 nm tip, and finally to 6 ± 1.5% for the 10.1 nm tip size. 359

DOI: 10.1021/acs.nanolett.7b04210 Nano Lett. 2018, 18, 357−364

Letter

Nano Letters the Supporting Information. This representation reveals that the emission quantum yields correlate inversely with the photon-to-hydrogen conversion efficiency: the lower the quantum yield of emission the higher the photocatalytic efficiency. Hence, with increasing tip size the corrected quantum yield of emission passes a minimum of 8.7% obtained for Ni tip size of 5.2 nm. The quenching of the emission in the Ni-tipped rods can be the result of the charge separation, which occurs at the interface between the semiconductor and the metal phase in the hybrid system, and prevents radiative recombination of the excitons.10,47,48 Assuming that a lower quantum yield of emission reflects improved charge separation, the observed trend in the emission quantum yields presents a first indication that the catalytic surface reaction might not be the (only) source of loss of catalytic efficiency with increasing tip size. Nevertheless, luminescence quenching in metal-tipped nanorods should not be regarded as evidence for electron transfer to the metal tip.49 Hence additional methods are necessary to interrogate the charge transfer at the semiconductor metal interface. To collect more detailed information on the occurring charge separation processes in dependence on the size of the metal tip, the tool of choice is transient absorption (TA) spectroscopy.17,27 This technique enables unambiguous investigation of charge transfer dynamics across the semiconductor−metal nanojunction. Examination of electron transfer processes into the metal (with transient absorption spectroscopy), alongside assessment of their final successful utilization on in the catalytic surface reaction (hydrogen production), will help disentangling the separate pathways from each other. This is vital for the elucidation of the function the metal size plays and its effect on the overall process mechanism. Transient absorption measurements were performed for the series of Ni-tipped rods, and a sample without Ni-tip as reference, upon excitation at 390 nm. The transient spectra of all samples (see Supporting Information Figure S8) show the characteristic negative bleach features of the CdSe and the CdS exciton transitions with maxima at 555 and 458 nm, respectively, in agreement with the static absorption spectra (see Supporting Information Figures S2 and S5). The bleach results from filling of the conduction band 1σe electron levels.17,27,50 It should be mentioned that no signature of the metal particle is observable in the transient spectra although partial direct excitation of metal-localized transitions occurs upon excitation at 390 nm. This is probably due to a missing signature of Ni nanoparticle in the detected spectral window. A comparison of the normalized kinetic traces at chosen probe wavelengths addressing the temporal development of the bleach feature of the CdS (probe 458 nm) and CdSe (probe 555 nm) localized excitonic transitions is displayed in Figure 4A,B. In general, the decay of the bleach features is accelerated in the presence of the metal tip, compared to the bare rods. This accelerated decay of the bleach is reflected in Figure 3 by the values of the half-life times with respect to the signal intensity at 1 ps. The extent of acceleration varies with the metal tip size, that is, up to 5.2 nm the bleach decays faster with increasing size, whereas for the two largest tips the decay is slowing down again. This indicates that with a tip size of 5.2 nm, a faster depopulation of the conduction band occurs, which correlates with the highest catalytic efficiency and the lowest emission quantum yield. The trend in efficiency of charge separation can be confirmed by comparison of the transient spectra at late delay times

Figure 4. Transient absorption kinetic traces at probe wavelength of (A) 458 nm (bleach feature of the CdS excitonic band edge transition) and (B) 555 nm (bleach feature of the CdSe excitonic band edge transition) and (C) transient spectra at a delay time of 1800 ps.

(Figure 4C), that is, at times beyond the typical time scale reported for charge separation processes between CdSe@CdS nanorods and metal nanoparticles and before recombination of the charge separated states.27,28 In such time frame, the charge separation process is expected to be finished, whereas recombination has not yet occurred, and the charge separated state is expected to accumulate in the sample. An indicator for the presence of a charge separated state in metal-tipped nanorods in the transient spectra is a positive absorption feature red-shifted to the excitonic absorption band edge of the CdS transition in our samples around 471 nm.5,21,23,24,44 It results from a Stark effect-induced shift of the exciton transition, induced by the electrical field which is generated by charge separation in the nanostructure. In all Ni-tipped samples, the 360

DOI: 10.1021/acs.nanolett.7b04210 Nano Lett. 2018, 18, 357−364

Letter

Nano Letters

comparison of the approaches, see Supporting Information. Noteworthy is the fact that in another previous study electron transfer rates were also found to be insensitive to the metal tip composition.28 In perfect agreement, the rates determined here for the Ni-tipped samples resemble closely the rates reported for nanorods of similar dimensions but with metal tips of different type (Pt, AuPt).28,51 Changes in the Fermi level, the work function of the metal nanoparticle and the density of acceptor states, are expected to occur upon variation of both tip size and material.10,52−57 It is thus surprising that these changes do not seem to influence the observed electron transfer rates, as predicted by theory.6,17,58 It should be stated that charge transfer through the nanojunction involves only a few nanometers of thin layer at the interface, whereas the total length of the nanorod is about 50 nm. The charge carrier has to localize in this region before the actual transfer can occur.17,59 Hence, a possible explanation for these observations is that the measured electron transfer dynamics encompasses two separate processes: (1) migration of the exciton or, if dissociation of the exciton has occurred, the electron, along the rod to the vicinity of the semiconductor− metal interface, and (2) electron transfer through the interface (interface-crossing).45 Transient absorption spectroscopy is only sensitive to the presence of electrons in the conduction band and is not able to separate the migration step from the interface-crossing step. If this exciton or charge carrier migration, which is independent of the metal nanoparticle size or composition, is significantly slower than crossing through the interface, it will dominate the measured time constants. Hence, the differences originating from these parameters will be obscured. Recent publications in the literature support the possibility of an ultrafast interface crossing step on the subpicosecond time range.60,61 On the other hand, charge migration is likely to proceed via slower diffusion. The time scale for exciton diffusion along a certain distance within the semiconductor rod can be estimated, based on exciton diffusion constants determined for CdSe@CdS nanorods, to be ∼0.5 ps for 10 nm,17,46 increasing quadratically with length and reaching ∼12 ps for 50 nm. These values are in the order of magnitude found in our measurements. Alternatively, a time scale in the order of 1 ps is estimated for electron diffusion along the complete rod length based on electron mobilities in CdSe@CdS nanorods (see Supporting Information)62 for the case when thermal dissociation of the exciton occurs first. Though this time scale is faster than the measured transfer kinetics, exciton dissociation could be the rate-limiting step in this scenario. The resulting possibility of length-dependent charge separation dynamics has thus far only been implied in the literature.4,17,48 For detailed insights and to define the mechanism, a comprehensive lengthdependent study is required. To summarize, we propose here that the charge transfer dynamics measurement is looking at two consecutive but separate processes: the first is a slow migration of the exciton or electron along the rod toward the interface (rate kT), and the second is very fast interface crossing (kIC). These two processes are illustrated in Figure 5A with wavy-gray and red arrows, correspondingly. The kinetic model in such situation follows that of a first order relation, and is determined solely by the rate-limiting step (i.e., charge migration, kT). The time constant determined by the transient absorption measurement is related to this rate via τ = 1/kT. The metal domain size does not influence the rate of charge carrier migration within the

negative bleach feature is diminished in the region between 460 and 480 nm, and the overall signal even becomes positive for the sample with a tip size of 5.2 nm, indicating the highest yield for charge separation in this sample. This is visualized in Figure 3 by plotting the deviation in the transient absorption signal intensity between Ni-tipped and bare rods at a probe wavelength of 471 nm, at the maximum of the positive feature characteristic for the charge-separated state. Again, this is in perfect agreement with the minimum in emission quantum yield observed and the maximum catalytic efficiency of this sample. To get a more quantitative picture on the charge separation processes in the series of nanorods with varying tip size, the bleach decay dynamics was modeled applying a multiexponential fitting model (for more details see Supporting Information). In order to exclusively model the kinetics of the metal-tipped rods, the presence of nontipped rods (as well as cases with only partial electron transfer) was taken in to account by including the rod intrinsic relaxation processes into to the fitting function, scaled with a factor of C. These processes were determined by an independent measurement on a sample of nonfunctionalized rods, which could be regarded as a background measurement. This approach allows extraction of time constants that are directly related to the charge transfer.28 In analogy to earlier reports for metal-tipped seeded nanorods, three time constants are found for charge separation, which can be assigned to charge transfer originating from excitons localized in different parts of the nanostructure.27,28,45,51 These time constants include 1.6 ps for transfer of electrons from a CdS region close to the metal tip, 16.6 ps for transfer from CdS excitonic states generated in a distant part of the nanorod (relative to the metal tip), and 34.5 ps for electron transfer from excitonic states localized in the CdSe seed. The kinetics of all samples independent of the metal tip size can be described by the same time constants for charge transfer but with varying amplitudes (see Supporting Information Table S5 and S6). The amplitudes, which are the pre-exponential factor of the exponents, are indicative of the relative weight of rod intrinsic relaxation versus charge separation processes. Hence, the amplitudes are reflecting on charge transfer probabilities. They cover for both the amount of charge-separated states and the acceleration/deceleration of the decay of the bleach feature observed. The amplitudes Aet for electron transfer processes from excitonic states generated in the rod vary with metal size (see Table S5 and S6 in the Supporting Information) increases to a maximum for the 5.2 nm tip size after which the weight of the charge transfer decreases again. In addition, the weight of the rod intrinsic relaxation processes, which are scaled by a weighting factor C, decreases and rises again with a minimum that is reached for the 5.2 nm tip size. This is in perfect agreement with the minimum of emission quantum yield, the maximum of the spectral feature of the charge separated state, as observed in the transient spectra, and last but not least the maximum in photon to hydrogen conversion efficiency, all obtained with 5.2 nm tip size. The observation of electron transfer constants that are nonvariant with metal tip size is in contrast to an earlier report.6 This discrepancy might be related to the fitting model in that study, which does not distinguish between rod intrinsic relaxation and charge separation processes, and as a result cannot account for the presence of a certain percentage of rods without metal tip in the ensemble. For a detailed discussion and 361

DOI: 10.1021/acs.nanolett.7b04210 Nano Lett. 2018, 18, 357−364

Letter

Nano Letters

nanojunction, by determining the probability for interface crossing. The first is the so-called “Coulomb blockade”. In a nanocrystal, the presence of one charge acts to prevent the addition of another, due to the Coulomb repulsion between them. Thus, in individual metals crystallites (or semiconductors), this “Coulomb blockade” leads to stepwise current−voltage curves, where steps for individual charging events are spaced proportional to 1/radius.65,66 Simply put, the Coulomb blockade may be regarded as an “energy penalty” which an electron must pay to enter a confined charged domain, such as the Ni tip in our system. This defines a “quasilevel” which is shifted (upward) from the Ni tip levels by an amount inversely proportional to the tip radius. This results with favorable transfer of electrons from the semiconductor into larger metal domains. This model can account for the first trend of the results for both charge transfer probabilities and hydrogen generation, which increase with metal domain size. The development in the amplitudes for larger metal tip sizes reveals decreasing charge separation efficiency. This is also reflected in the trend for emission quantum yield, where the residual photoluminescence provides an inverse measure of the efficiency of the photoinduced charge transfer across the semiconductor−metal interface. To account for this trend, we suggest that a size-dependent Schottky barrier competes with the aforementioned Coulomb blockade. Semiconductor metal−nanojunctions are particularly challenging to explore experimentally, and common characterization techniques typically involve the formation of electrical contact between the nanoparticle and bulk metal electrode/ probe, thus altering and blurring unique nanoscale phenomena.67,68 Yet, it is known that the Schottky barrier height of a nanojunction is lower than that of the planar thin-film counterparts.69−71 A gradual development of the Schottky barrier between the CdSe@CdS semiconductor nanorod and the metal domain as the Ni tip size increases is expected to result in a decrease in charge transfer probability across that interface, in good agreement with the observed results from transient absorption spectroscopy. We corroborate our theory with a calculation of quantum yield, taking into account both Coulomb blockade and sizedependent Schottky barrier height. We consider a generic model system for photocatalytic hydrogen production in Nidecorated CdSe@CdS nanorods, schematically depicted in Figure 5B. The details of the calculation, including methodology and numerical parameters, are described in the SI. Here we point that we take these two effects into account by making the Ni-tip energy level inversely proportional to the tip radius, and the time-scale to cross from the CdS to the Ni tip decrease with tip radius. These two processes compete with each other, resulting in an optimal radius of ∼5 nm. In Figure 3 we plot the calculated quantum efficiencies (orange squares) on top of the experimental data for photocatalytic photon to hydrogen conversion efficiencies, demonstrating excellent agreement between the two. As an alternate explanation, the decrease in charge transfer probability (separation efficiency) with respect to further increase in metal size could be a result of lattice strain-induced barrier that develops at the metal-semiconductor interface after the metal exceeds a certain size.72,73 Another possible change at the interface, which is difficult to confirm or exclude experimentally, relates to partial diffusion of Ni ions into the

Figure 5. (A) Illustration of the charge transfer mechanism: a slow migration of the exciton or electron along the rod, which is not affected by the metal size, is depicted with wavy-gray arrow, and the fast interface crossing is depicted with a red arrow. (B) Illustration of the metal size effect on the semiconductor−metal nanojunction: Schottky barrier develops with increasing metal size, while Coulomb blockade charging energy is decreasing. These two opposing trends result with an optimal metal size domain for the cocatalyst.

semiconductor domain, hence this model explains the size independence of the observed rates. Although the time constant for the interface crossing step, possibly through tunneling of the electron into the metal, can not be accessed by the measurement directly, its efficiency is reflected in the amplitudes in the fit and was found to be sensitive to the metal tip size. The interface crossing process competes with charge migration away from the interface region (k−T). The yield of charge separation (successful electron transfer processes into the metal) is proportional to the ratio of the interface-crossing rate (kIC), and the rate with which the charge carrier leaves again the interface region (k−T). The faster the kIC, the higher its efficiency. This is reflected in the amplitude of the charge separation process in the data. Higherfitted amplitude for this process, that is a higher yield of charge separation, is indicative for faster interface crossing. In previous reports the existence of an optimal metal domain size for the hydrogen reduction cocatalyst was attributed to a trade-off between increasing rate for charge separation and process occurring at the metal−liquid interface. In particular, the decrease in activity was assigned to a lower surface reactivity of the larger nanoparticles.6,7,63,64 This explanation however cannot account for the development observed here in the efficiency for charge transfer across the semiconductor−metal nanointerface. Hence, we propose alternative competing phenomena, as illustrated in Figure 5B, which contribute directly to the metal size effect on the semiconductor−metal 362

DOI: 10.1021/acs.nanolett.7b04210 Nano Lett. 2018, 18, 357−364

Nano Letters CdS, which might occur during deposition conditions that were adjusted for larger metal tips.28,29 To conclude, the design of effective photocatalytic systems for solar to fuel conversion will depend on the ability to correlate the photocatalyst structure, composition, and morphology with its activity. In particular, the metal cocatalyst morphology plays a key role in the optimization of the materials potential in promoting the reaction of interest. A set of Ni-decorated CdSe@CdS nanorods with different tip size were examined, and an optimal metal domain size of 5.2 nm was obtained, in good agreement with the literature. Here we reveal that the nature of a nanojunction formed between a semiconductor nanorod photosenisitizer and metal nanoparticle cocatalyst is sensitive to the size of the metal component, as reflected in the activity toward hydrogen production, emission quantum yield, and the quantity of charge separation. Remarkably, the determined charge separation time constants were found to be nonvariant with metal tip size, in contrast to earlier reports. It is proposed that electron transfer mechanism encompasses two consecutive but separate processes: slow charge migration along the rod toward the interface, followed by ultrafast interface crossing of the electron from the semiconductor into the metal phase. The first migration step dominates the time constant derived from the dynamics measurement and is not affected by the metal tip size. The quantity of charge separation on the other hand was found to be sensitive to the metal size. It is suggested that two phenomena contribute to the metal size effect on charge transfer probability across the semiconductor−metal nanojunction with opposing trends. Coulomb blockade charging energy dominates the lower metal size range, whereas size of the dependent Schottky barrier becomes more pronounced for the larger metal size range. These two opposing trends result with an optimal metal size domain for the cocatalyst. This work is of significance to any application utilizing semiconductor−metal nanocomposites and is thus expected to benefit a broad range of interdisciplinary research areas and a wide spectrum of applications such as solar and fuel cells, photocatalysis, electronics, and photonics.

■

■

ACKNOWLEDGMENTS

■

REFERENCES

We acknowledge the support by the COST Action CM1202 PERSPECT-H2O. L.A. gratefully acknowledges the support of the I-CORE Program of the Planning and Budgeting Committee, and The Israel Science Foundation (Grant 152/ 11), as well as the German-Israeli Foundation (GIF) for Scientific Research and Development (Grant 2307-2319.5/ 2011). This research was partially carried out in the framework of Russell Berrie Nanotechnology Institute (RBNI) and the Nancy and Stephen Grand Technion Energy Program (GTEP). M.W. acknowledges support by the Fonds der Chemischen Industrie (FCI).

(1) Amirav, L.; Alivisatos, A. P. J. Phys. Chem. Lett. 2010, 1, 1051− 1054. (2) Kalisman, P.; Nakibli, Y.; Amirav, L. Nano Lett. 2016, 16, 1776− 1781. (3) Nakibli, Y.; Kalisman, P.; Amirav, L. J. Phys. Chem. Lett. 2015, 6, 2265−2268. (4) Simon, T.; Carlson, M. T.; Stolarczyk, J. K.; Feldmann, J. ACS Energy Lett. 2016, 1, 1137−1142. (5) Luo, M.; Yao, W.; Huang, C.; Wu, Q.; Xu, Q. J. Mater. Chem. A 2015, 3, 13884−13891. (6) Ben-Shahar, Y.; Scotognella, F.; Kriegel, I.; Moretti, L.; Cerullo, G.; Rabani, E.; Banin, U. Nat. Commun. 2016, 7, 10413. (7) Subramanian, V.; Wolf, E. E.; Kamat, P. V. J. Am. Chem. Soc. 2004, 126, 4943−4950. (8) Han, N. S.; Kim, D.; Lee, J. W.; Kim, J.; Shim, H. S.; Lee, Y.; Lee, D.; Song, J. K. ACS Appl. Mater. Interfaces 2016, 8, 1067−1072. (9) Schweinberger, F. F.; Berr, M. J.; Doblinger, M.; Wolff, C.; Sanwald, K. E.; Crampton, A. S.; Ridge, C. J.; Jackel, F.; Feldmann, J.; Tschurl, M.; Heiz, U. J. Am. Chem. Soc. 2013, 135, 13262−13265. (10) Vamvasakis, I.; Liu, B.; Armatas, G. S. Adv. Funct. Mater. 2016, 26, 8062−8071. (11) Zhukovskyi, M.; Tongying, P.; Yashan, H.; Wang, Y.; Kuno, M. ACS Catal. 2015, 5, 6615−6623. (12) Amirav, L.; Oba, F.; Aloni, S.; Alivisatos, A. P. Angew. Chem., Int. Ed. 2015, 54, 7007−7011. (13) Razgoniaeva, N.; Moroz, P.; Lambright, S.; Zamkov, M. J. Phys. Chem. Lett. 2015, 6, 4352−4359. (14) Haruta, M. Catal. Today 1997, 36, 153−166. (15) Cuenya, B. R. Thin Solid Films 2010, 518, 3127−3150. (16) Vaneski, A.; Susha, A. S.; Rodriguez-Fernandez, J.; Berr, M.; Jackel, F.; Feldmann, J.; Rogach, A. L. Adv. Funct. Mater. 2011, 21, 1547−1556. (17) Wu, K. F.; Lian, T. Q. Chem. Soc. Rev. 2016, 45, 3781−3810. (18) Costi, R.; Saunders, A. E.; Banin, U. Angew. Chem., Int. Ed. 2010, 49, 4878−4897. (19) Talapin, D. V.; Lee, J. S.; Kovalenko, M. V.; Shevchenko, E. V. Chem. Rev. 2010, 110, 389−458. (20) Kovalenko, M. V.; Manna, L.; Cabot, A.; Hens, Z.; Talapin, D. V.; Kagan, C. R.; Klimov, V. I.; Rogach, A. L.; Reiss, P.; Milliron, D. J.; Guyot-Sionnnest, P.; Konstantatos, G.; Parak, W. J.; Hyeon, T.; Korgel, B. A.; Murray, C. B.; Heiss, W. ACS Nano 2015, 9, 1012− 1057. (21) Talapin, D. V.; Koeppe, R.; Gotzinger, S.; Kornowski, A.; Lupton, J. M.; Rogach, A. L.; Benson, O.; Feldmann, J.; Weller, H. Nano Lett. 2003, 3, 1677−1681. (22) Müller, J.; Lupton, J. M.; Lagoudakis, P. G.; Schindler, F.; Koeppe, R.; Rogach, A. L.; Feldmann, J.; Talapin, D. V.; Weller, H. Nano Lett. 2005, 5, 2044−2049. (23) Carbone, L.; Nobile, C.; De Giorgi, M.; Sala, F. D.; Morello, G.; Pompa, P.; Hytch, M.; Snoeck, E.; Fiore, A.; Franchini, I. R.; Nadasan, M.; Silvestre, A. F.; Chiodo, L.; Kudera, S.; Cingolani, R.; Krahne, R.; Manna, L. Nano Lett. 2007, 7, 2942−2950.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.7b04210. Detailed description of the experimental setup and conditions, transient spectra, and details of the fitting (PDF)

■

Letter

AUTHOR INFORMATION

Corresponding Authors

*E-mail: (L. A.) [email protected]. *E-mail: (M.W.) [email protected]. ORCID

Yonatan Dubi: 0000-0002-8988-4935 Lilac Amirav: 0000-0002-0539-0488 Notes

The authors declare no competing financial interest. 363

DOI: 10.1021/acs.nanolett.7b04210 Nano Lett. 2018, 18, 357−364

Letter

Nano Letters

(56) Yuan, H. K.; Kuang, A. L.; Tian, C. L.; Chen, H. AIP Adv. 2014, 4, 037107. (57) Leppert, L.; Albuquerque, R. Q.; Foster, A. S.; Kümmel, S. J. Phys. Chem. C 2013, 117, 17268−17273. (58) Zhu, H.; Yang, Y.; Wu, K.; Lian, T. Annu. Rev. Phys. Chem. 2016, 67, 259−281. (59) Yang, Y.; Liu, Z.; Lian, T. Nano Lett. 2013, 13, 3678−3683. (60) Dana, J.; Debnath, T.; Maity, P.; Ghosh, H. N. J. Phys. Chem. C 2015, 119, 22181−22189. (61) Dana, J.; Maity, P.; Ghosh, H. N. Nanoscale 2017, 9, 9723− 9731. (62) Kunneman, L. T.; Zanella, M.; Manna, L.; Siebbeles, L. D. A.; Schins, J. M. J. Phys. Chem. C 2013, 117, 3146−3151. (63) Kamat, P. V. J. Phys. Chem. Lett. 2012, 3, 663−672. (64) Jakob, M.; Levanon, H.; Kamat, P. V. Nano Lett. 2003, 3, 353− 358. (65) Kastner, M. A. Rev. Mod. Phys. 1992, 64, 849−858. (66) Hong, I.-P.; Brun, C.; Pivetta, M.; Patthey, F.; Schneider, W.-D. Front. Phys. 2013, 1, 13. (67) Costi, R.; Cohen, G.; Salant, A.; Rabani, E.; Banin, U. Nano Lett. 2009, 9, 2031−2039. (68) Sheldon, M. T.; Trudeau, P. E.; Mokari, T.; Wang, L. W.; Alivisatos, A. P. Nano Lett. 2009, 9, 3676−3682. (69) Savadogo, O.; Mandal, K. C. J. Electrochem. Soc. 1994, 141, 2871−2877. (70) Kwon, S.; Lee, S. J.; Kim, S. M.; Lee, Y.; Song, H.; Park, J. Y. Nanoscale 2015, 7, 12297−12301. (71) Ruffino, F.; Grimaldi, M. G.; Giannazzo, F.; Roccaforte, F.; Raineri, V. Appl. Phys. Lett. 2006, 89, 243113. (72) Nugraha, M. I.; Matsui, H.; Watanabe, S.; Kubo, T.; Hausermann, R.; Bisri, S. Z.; Sytnyk, M.; Heiss, W.; Loi, M. A.; Takeya, J. Adv. Electron. Mater. 2017, 3, 1600360. (73) Daio, T.; Staykov, A.; Guo, L.; Liu, J.; Tanaka, M.; Matthew Lyth, S.; Sasaki, K. Sci. Rep. 2015, 5, 13126.

(24) Talapin, D. V.; Nelson, J. H.; Shevchenko, E. V.; Aloni, S.; Sadtler, B.; Alivisatos, A. P. Nano Lett. 2007, 7, 2951−2959. (25) Lupo, M. G.; Della Sala, F.; Carbone, L.; Zavelani-Rossi, M.; Fiore, A.; Luer, L.; Polli, D.; Cingolani, R.; Manna, L.; Lanzani, G. Nano Lett. 2008, 8, 4582−4587. (26) Raino, G.; Stoferle, T.; Moreels, I.; Gomes, R.; Kamal, J. S.; Hens, Z.; Mahrt, R. F. ACS Nano 2011, 5, 4031−4036. (27) Wu, K. F.; Zhu, H. M.; Lian, T. Q. Acc. Chem. Res. 2015, 48, 851−859. (28) Wächtler, M.; Kalisman, P.; Amirav, L. J. Phys. Chem. C 2016, 120, 24491−24497. (29) Banin, U.; Ben-Shahar, Y.; Vinokurov, K. Chem. Mater. 2014, 26, 97−110. (30) Sitt, A.; Della Sala, F.; Menagen, G.; Banin, U. Nano Lett. 2009, 9, 3470−3476. (31) She, C. X.; Bryant, G. W.; Demortiere, A.; Shevchenko, E. V.; Pelton, M. Phys. Rev. B: Condens. Matter Mater. Phys. 2013, 87, 155427. (32) She, C. X.; Demortiere, A.; Shevchenko, E. V.; Pelton, M. J. Phys. Chem. Lett. 2011, 2, 1469−1475. (33) Amirav, L.; Alivisatos, A. P. J. Am. Chem. Soc. 2013, 135, 13049− 13053. (34) Kalisman, P.; Houben, L.; Aronovitch, E.; Kauffmann, Y.; BarSadan, M.; Amirav, L. J. Mater. Chem. A 2015, 3, 19679−19682. (35) Nakibli, Y.; Amirav, L. Chem. Mater. 2016, 28, 4524−4527. (36) Balcerski, W. C.; Ryu, S. Y.; Hoffmann, M. R. Sep. Purif. Technol. 2015, 156, 915−921. (37) Wang, J.; Li, B.; Chen, J. Z.; Li, N.; Zheng, J. F.; Zhao, J. H.; Zhu, Z. P. Appl. Surf. Sci. 2012, 259, 118−123. (38) Luo, M.; Liu, Y.; Hu, J. C.; Liu, H.; Li, J. L. ACS Appl. Mater. Interfaces 2012, 4, 1813−1821. (39) Ratcliff, E. L.; Meyer, J.; Steirer, K. X.; Garcia, A.; Berry, J. J.; Ginley, D. S.; Olson, D. C.; Kahn, A.; Armstrong, N. R. Chem. Mater. 2011, 23, 4988−5000. (40) Ran, J. R.; Yu, J. G.; Jaroniec, M. Green Chem. 2011, 13, 2708− 2713. (41) Ran, J. R.; Zhang, J.; Yu, J. G.; Qiao, S. Z. ChemSusChem 2014, 7, 3426−3434. (42) Zhang, J.; Qiao, S. Z.; Qi, L. F.; Yu, J. G. Phys. Chem. Chem. Phys. 2013, 15, 12088−12094. (43) Chen, S.; Chen, X. P.; Jiang, Q. Z.; Yuan, J.; Lin, C. F.; Shangguan, W. F. Appl. Surf. Sci. 2014, 316, 590−594. (44) Simon, T.; Bouchonville, N.; Berr, M. J.; Vaneski, A.; Adrovic, A.; Volbers, D.; Wyrwich, R.; Doblinger, M.; Susha, A. S.; Rogach, A. L.; Jackel, F.; Stolarczyk, J. K.; Feldmann, J. Nat. Mater. 2014, 13, 1013−1018. (45) Wu, K. F.; Rodriguez-Cordoba, W. E.; Liu, Z.; Zhu, H. M.; Lian, T. Q. ACS Nano 2013, 7, 7173−7185. (46) Wu, K. F.; Hill, L. J.; Chen, J. Q.; McBride, J. R.; Pavlopolous, N. G.; Richey, N. E.; Pyun, J.; Lian, T. Q. ACS Nano 2015, 9, 4591− 4599. (47) Rukenstein, P.; Teitelboim, A.; Volokh, M.; Diab, M.; Oron, D.; Mokari, T. J. Phys. Chem. C 2016, 120, 15453−15459. (48) Wu, K.; Zhu, H.; Liu, Z.; Rodríguez-Córdoba, W.; Lian, T. J. Am. Chem. Soc. 2012, 134, 10337−10340. (49) Demortiere, A.; Schaller, R. D.; Li, T.; Chattopadhyay, S.; Krylova, G.; Shibata, T.; Claro, P. C. D.; Rowland, C. E.; Miller, J. T.; Cook, R.; Lee, B.; Shevchenko, E. V. J. Am. Chem. Soc. 2014, 136, 2342−2350. (50) Gust, D.; Moore, T. A.; Moore, A. L. Acc. Chem. Res. 2009, 42, 1890−1898. (51) Wu, K. F.; Chen, Z. Y.; Lv, H. J.; Zhu, H. M.; Hill, C. L.; Lian, T. Q. J. Am. Chem. Soc. 2014, 136, 7708−7716. (52) Scanlon, M. D.; Peljo, P.; Mendez, M. A.; Smirnov, E.; Girault, H. H. Chem. Sci. 2015, 6, 2705−2720. (53) Roduner, E. Chem. Soc. Rev. 2006, 35, 583−592. (54) Hartland, G. V. Chem. Rev. 2011, 111, 3858−3887. (55) Hodak, J. H.; Henglein, A.; Hartland, G. V. J. Chem. Phys. 2001, 114, 2760−2765. 364

DOI: 10.1021/acs.nanolett.7b04210 Nano Lett. 2018, 18, 357−364