Hybrid Plasmonic Nanomaterials for Hydrogen Generation and Carbon Dioxide Reduction

The successful development of artificial photosynthesis requires finding new materials able to efficiently harvest sunlight and catalyze hydrogen generation and carbon dioxide reduction reactions. Plasmonic nanoparticles are promising candidates for these tasks, due to their ability to confine solar energy into molecular regions. Here, we review recent developments in hybrid plasmonic photocatalysis, including the combination of plasmonic nanomaterials with catalytic metals, semiconductors, perovskites, 2D materials, metal–organic frameworks, and electrochemical cells. We perform a quantitative comparison of the demonstrated activity and selectivity of these materials for solar fuel generation in the liquid phase. In this way, we critically assess the state-of-the-art of hybrid plasmonic photocatalysts for solar fuel production, allowing its benchmarking against other existing heterogeneous catalysts. Our analysis allows the identification of the best performing plasmonic systems, useful to design a new generation of plasmonic catalysts.

O ver the past two centuries, the rapid and continuing increase in energy consumption has led to a global energy and climate crisis. In the past few decades, almost 75% of human-caused carbon dioxide (CO 2 ) emissions were a result of burning fossil fuels. 1 Yet, to this date their share makes up more than 80% of the energy mix. 2 It is evident that a transformation toward renewable energy sources is urgently needed. For millions of years, nature has used the sun as its primary energy source by photosynthesis during which sunlight is used to split water and produce energy-rich chemical compounds from CO 2 . Taking inspiration from this, artificial photosynthesis has become one of the most important research fields in the past decade. It consists mainly of two types of processes: one is water-splitting into hydrogen (H 2 ) and oxygen (O 2 ) or the H 2 production from H 2containing compounds, and the other is the reduction of CO 2 into fuels and chemicals, such as carbon monoxide, hydrocarbons, and oxygenates. These products are generally referred to as solar fuels.
In artificial photosynthesis, the first step is to harvest and concentrate the solar energy in molecular-sized regions. Plasmonic nanoparticles (PNPs) are promising materials to perform these tasks, due to their exceptional absorption and their ability to control light and heat at the nanoscale. 3−6 Therefore, the inclusion of a plasmonic material into a multicomponent nanostructure can greatly enhance its photocatalytic performance compared to the individual components. In recent years, an increasing number of these so-called hybrid plasmonic nanostructures have been investigated for chemical reactions, giving rise to the field of plasmon-assisted catalysis. 7−12 Despite being a young field, there are already several reviews, 13−15 perspectives, 16−18 and even a book 19 covering different topics of the area. Some focus on the theoretical aspects and the working mechanisms of plasmonic photocatalysis. 20−22 Others discuss reactions driven by a particular plasmon-induced phenomenon such as hot-carrier generation, 23 photothermal effects 24 or both. 25 Furthermore, Devasia et al. focus on the possibility of controlling reaction pathways with plasmonic systems. 26 Finally, some reviews focus on particular reactions, such as the applications of plasmonic catalysts to organic transformations 27 or the CO 2 reduction reaction (CO 2 RR). 28 Among the reviews on plasmon-assisted catalysis there are few on hybrid systems, so far mostly limited to bimetallic or plasmonic metal−semiconductor systems. 29−33 After around 10 years of exponential growth of the field, it is necessary to benchmark plasmonic materials for catalysis, not only among them but also against other existing heterogeneous catalysts for solar fuel production. This could give a real dimension of the improvements and achievements so far while showing the potential of this new type of heterogeneous (photo)catalysis in the quest of sustainable catalysis and clean energy production. Here, we will center on the considerable advances in the development of hybrid nanomaterials toward the production of solar fuels via plasmon excitation in the liquid phase. Many systems with great potential have been investigated for plasmonic catalysis in the gas phase. 9,11,25,34−40 However, in this Review, we are limiting the examples to reactions under mild reaction conditions (i.e., heterogeneous catalysis in the liquid phase). In addition, we will narrow the scope of this Review to the reactions that are relevant for solar fuel production, namely H 2 generation and CO 2 reduction, but we will be as broad as possible regarding the evaluation of different types of novel hybrid systems, i.e., not only limited to metals and semiconductors but also including plasmonic hybrids with perovskites, 2D materials, and metal−organic frameworks for both pure photocatalysis and photoelectrocatalysis. With the increasing complexity and diversity of hybrid plasmonic nanostructures comes the necessity of a quantitative comparison among them. Here, we propose a series of standard metrics that should be reported for future plasmonic catalysts. Identifying the best performers is essential to design future plasmonic platforms for solar fuel production. That is the scope of this Review.
We first briefly elaborate on the motivation to employ plasmonic materials for photo(electro)catalysis and enumerate the enhanced properties that can be obtained in hybrid nanostructures. Second, we describe the reactions to produce solar fuels that are the focus of interest of this Review. These are the reactions toward CO 2 reduction and H 2 generation. Third, we present the performance metrics that will be used to compare different materials/systems. In the main section, recent developments in each of the hybrid systems are reviewed. Last, in the Discussion section, the systems are summarized and compared for each reaction in a concluding meta-analysis.
■ PROPERTIES OF HYBRID PLASMONIC NANOMATERIALS PNPs interact strongly with light through the excitation of localized surface plasmon resonances (LSPRs), which correspond to the coherent oscillations of their free electrons coupled to the electromagnetic field. 3 Plasmonic materials are typically metals with negative real permittivity (i.e., with high concentration of free electrons) such as silver, gold, copper, etc. The plasmon resonances are highly tunable, and their frequency can cover the ultraviolet (UV), visible, and nearinfrared (NIR) regions of the electromagnetic spectrum (i.e., the totality of solar emission). 5 When resonantly illuminated, the absorption cross sections of PNPs reach tens of times their geometrical size, making them efficient harvesters of light. In addition, PNPs can act as solar light antennas. They can concentrate light in small volumes of typically a few nm 3 near their surface, where the intensity can reach values as high as 10 5 the solar one. 4 That volume is on the order of molecular sizes, transforming the PNPs into nanoreactors within molecular scales.
In artificial photosynthesis, the first step is to harvest and concentrate the solar energy in molecular-sized regions. There is a variety of plasmon-driven processes that can affect the activity and selectivity of target molecules. 20,41 Plasmon resonances are of mixed optoelectronic nature, and as such, the energy from incoming light is partly stored in the kinetic energy of charge carriers and in the electromagnetic field. These resonances can decay radiatively or transfer their energy to a charge carrier. This highly energetic charge carrier then relaxes, during which secondary and lower energy charge carriers are excited, and ultimately heat is released to the surrounding of the PNPs. This means that a resonantly illuminated PNP can supply a broad range of energy quanta, from highly energetic charge carriers with the energy of the incoming photons (∼2−3 eV) to thermal phonons (∼0.025 eV). A nearby molecule or material can use these energy quanta to perform a chemical transformation in thermal or non-thermal ways. The versatility of plasmon-driven processes offers the opportunity to obtain alternative reaction pathways toward a desired but kinetically or thermodynamically unfavorable product. For example, they can operate in a particular intermediate step of a reaction, altering the pathway toward a different product. 42,43 However, the catalytic activity of typical plasmonic metals (Au, Ag, Cu, etc.) for solar fuel production is comparatively low. 44,45 In recent years, plasmonic components have been interfaced with other nanomaterials, searching for superior hybrid plasmonic nanocatalysts. These hybrid plasmonic catalysts are summarized in the top half of Figure 1. The promising properties of PNPs are combined with additional enhanced photocatalytic properties from the second material that are in many cases not present in the single-component counterparts. The bottom half of Figure 1 schematizes some of the properties that are desirable (and potentially obtainable) in a hybrid plasmonic catalyst. While the plasmonic component harvests light and supplies energy through charge carriers, heat, or electromagnetic fields, a second (non-plasmonic) component can, for example, improve the interaction with target molecules, enhance the stability of the system or enlarge the lifetime of excited carriers. In the following, we present the particular opportunities that each material/interface opens for plasmonic catalysis.
Plasmonic Metal−Metal. The catalytic activity of plasmonic metals can be improved by combining plasmonic metals with common catalytically active metals such as palladium, platinum, rhodium, etc. The inclusion of such metals can help lower the activation energy of a reaction step or enhance the adsorption/desorption dynamics of reactants/ products on the surface of the nanocatalyst. In addition, combining the two metals modifies the electronic energy bands close to their interface, potentially facilitating charge flow, reducing the chemical potential of the charge carriers, and enhancing their extraction. 46 In bimetallic systems, the decay of plasmon resonances gets altered, favoring non-radiative (excitation of carriers, heat generation) versus radiative channels (scattering). 47−49 Interestingly, the spatial distribution of absorption can be shifted toward the catalytic metal generating carriers exactly where they are more reactive. 38,50,51 Plasmonic Metal−Semiconductor. The lifetime of photoexcited charge carriers in plasmonic metals is typically less than 1 ps. 52,53 This time is usually too short for the carriers to act on an adsorbate, a detrimental fact for the overall reactivity of the system. Longer charge carrier lifetimes can be attained by interfacing the metal with a semiconductor. The decay of the plasmon can generate charge carriers in the semiconductor through three main mechanisms. First, photoexcited electrons in the metal with sufficient energy can be transferred across the Schottky barrier 54 to the conduction band of the semiconductor, in a process called indirect electron transfer. 24,55 Second, a plasmon can decay by directly exciting an electron into the conduction band of the semiconductor and a hole in the metal, in a process called plasmon-induced charge-transfer transition 55 or direct electron transfer. 13,56 The third mechanism is a plasmon-induced resonant energy transfer, 57,58 where an electron−hole pair is excited in the semiconductor. The presence of a Schottky barrier at a metal/ semiconductor interface 54 traps electrons and prevents or delays them from coming back to the metal. 55 The lifetime, just like the indirect electron injection across the barrier, depends on the potential difference and a trade-off has to be made between the harvesting and recombination of carriers. 59,60 Furthermore, structural defects and/or vacancies in the semiconductor can also act as electron traps, enhancing their lifetime, and are usually the most reactive sites. 61 These processes allow electrons to become available in the semiconductor to drive reduction reactions while holes remain in the metal for oxidation reactions. 13 Inversely, there have also been demonstrations of semiconductors harvesting the holes. 40,62 Without the photoexcited electrons in the plasmonic metal, wide bandgap semiconductors like titanium dioxide (TiO 2 ) are photocatalytically active only in the UV. The energy transfer mechanisms from the plasmonic material to the semiconductor extend their operational wavelength range, accommodating a larger fraction of the solar spectrum and potentially increasing the solar-to-fuel conversion efficiency. 63,64 Plasmonic Metal−Perovskite. Perovskites are a versatile subclass of semiconductors with attractive optoelectronic and catalytic properties. They are characterized by their ABX 3 crystal structure, where A and B represent cations and X an anion (see Figure 4a, below). As semiconductors, the physical and chemical properties of perovskites leading to a catalytic enhancement are generally as described above. This is especially true for ABO 3 -type perovskite oxides, which are cost efficient and can accommodate a wide range of substituting and doping elements to modulate their electronic and catalytic properties but are generally photocatalytically active only under UV irradiation owing to their wide bandgap. Their combination with plasmonic materials allows the absorption of visible light, making irradiation of wavelengths with photon energies below the bandgap of the perovskite accessible. 65−68 Both charge or energy transfer from the plasmonic material to the perovskite can occur. 69,70 The perovskite oxides are in many ways very similar to semiconductors included in the general semiconductor section; However, the perovskite category is included in this Review due to a second type of perovskite, namely ABX 3 -type perovskite halides (X = F, Cl, Br, I). They have a tunable absorption across the entire visible spectrum 71 as well as long carrier lifetimes. 72 In addition to the benefits discussed above, by combining perovskite halides and plasmonic materials, the characteristic absorption spectra of both components can be matched such that the strong near-field enhancement of the PNP can significantly enhance the absorption of light by the halide perovskite, potentially boosting their catalytic performance.
Plasmonic Metal−2D. A 2D material is the co-catalyst with the largest possible available surface. Of particular interest are materials with outstanding electrical conductivity and mobility, such as carbon-based 2D and transitional metal-based 2D materials. The photogenerated electrons in the PNPs can transfer to the 2D material, where they benefit from a larger lifetime fostering substrate reactivity. 73 Furthermore, electrons can accumulate in the 2D material or promote the formation of charged excitons (with more than one electron coupled to a hole). 74,75 This accumulation facilitates reactions that require multiple electrons. Finally, the presence of defects, dopants, or oxygenated functional groups in the 2D material can act as catalytic sites for specific reactions. 76 Plasmonic Metal−MOFs. Metal−organic frameworks (MOFs) are ordered periodic structures formed by nodes of metal ions linked by organic ligands. 77−79 Plasmonic materials can gain interesting properties relevant to catalysis when combined with this versatile class of porous materials. 80−84 First, MOFs can act as substrates, providing support and increasing stability. 85 Second, they are highly versatile in composition and pore size, allowing structural tailoring according to the application. For example, their organic linker can be adjusted to catalyze a specific reaction or co-reaction. 86 This property makes them promising systems to achieve high selectivity of product formation. Third, they can be structurally designed so that their catalytic centers lie in the near-field enhancement region of the plasmonic component, enhancing their absorption. 85 Fourth, their high porosity can allow an extraordinarily large surface area for a reaction to occur 87 or the preconcentration of reactants next to the PNPs. 88 Finally, MOFs can also act as a charge carrier mediator to enhance the electron−hole pair separation efficiency. 89 Plasmon-Assisted Electrochemistry. Electrocatalysis, like photocatalysis, has been studied extensively in recent years. 90 While promising, electrochemical reactions often require high overpotentials. Plasmonic materials can enhance the performance of electrocatalytic systems by utilizing the solar energy. This can result in lower electrical energy consumption and higher product formation, which manifest as a reduced overpotential or increased current density. 91−93 The introduction of a plasmonic material can also modify the chemical landscape, which can affect selectivity by triggering or enhancing otherwise thermodynamically and/or kinetically hindered reactions. 94 ■ PLASMON-ASSISTED REACTIONS TOWARD SOLAR FUEL GENERATION In this Review, we discuss hybrid photocatalysts with at least one plasmonic component that is resonantly illuminated. This excludes, for example, systems where the presence of a plasmonic metal can have a positive impact on the catalytic activity of the system but is either not illuminated or illuminated in a region with negligible absorption. 95,96 It should be noted that we consider both photocatalytic processes (typically thermodynamically favorable, ΔG < 0, with G the Gibbs free energy) and photosynthetic processes (thermodynamically unfavorable, ΔG > 0) as plasmon-assisted reactions. For simplicity, they will from here on be referred to as photocatalytic. Among the large list of recently demonstrated plasmon-assisted reactions, we focus on a few particularly relevant reactions toward the generation of solar fuels: H 2 generation reactions and CO 2 RRs, as described below and summarized in Table 1.
Hydrogen Generation. Because hydrogen can be burned with no greenhouse emissions, it is a promising candidate to replace fossil fuels. Hydrogen has a high gravimetric energy density of 33.3 kW h kg −1 , which is almost 3 times higher than that of gasoline (12.4 kW h kg −1 ). 97,98 It is stable and can be liquefied, stored, or transported through pipelines. However, the majority of hydrogen is still produced from fossil fuels, resulting in excessive carbon emissions. 99 For hydrogen to become a sustainable fuel, it has to be produced from renewable energy sources.
Photocatalytic hydrogen sources can be categorized as follows: (1) Water. Solar hydrogen production via watersplitting is especially attractive, as both water and sunlight are abundant. Materials for photocatalytic water-splitting are either investigated toward their performance of full water-splitting, where H 2 and O 2 evolution should occur in a stoichiometric 2:1 ratio, or by using a sacrificial hole acceptor, for only the hydrogen evolution reaction (HER). This simplifies research and prevents charge recombination. However, this sacrificial agent should also be cheap and abundant, as otherwise this could become a new bottleneck for the reaction to be scaled up.
(2) Other hydrogen-containing compounds. While hydrogen has a high energy content per mass, its energy content per volume is poor. This makes storing enough hydrogen, e.g., onboard of means of transport, difficult. A possible solution is the chemical storage in compounds with a higher storage density. This has resulted in compounds with a high hydrogen content like formic acid (HCOOH), 1 0 0 ammonia borane (NH 3 BH 3 ), 101 or sodium borohydride (NaBH 4 ) 102 gaining attention, especially for mobile and portable applications.
Carbon Dioxide Reduction. Carbon dioxide (CO 2 ) is one of the main greenhouse gases contributing to climate change. 103 CO 2 conversion can have the two-fold benefit of decreasing its atmospheric concentration while also producing fuels. Compared to hydrogen production, the reactions for CO 2 conversion are much more complex, as a variety of Among the recently demonstrated plasmon-assisted reactions, we focus on a few particularly relevant reactions toward the generation of solar fuels: H 2 generation reactions and CO 2 reduction reactions.
reaction pathways coexist. Possible products include carbon monoxide (CO), formic acid (HCOOH), methane (CH 4 ), methanol (CH 3 OH), ethane (C 2 H 6 ), propane (C 3 H 8 ), and ethanol (CH 3 CH 2 OH), among others. These products require a different number of charge transfers, as shown in Table 1, and contain different numbers of carbon. Products with two or more carbon atoms, so-called C2+ products, are desirable but difficult to produce. Several steps can be involved, leading to the possibility of branching reaction pathways and various products being produced simultaneously. Therefore, a suitable large-scale photocatalyst for CO 2 reduction should not only present a high efficiency but also a high selectivity to one product. In this Review, we will quantitatively evaluate the prospect of using hybrid plasmonics toward these goals.

■ PERFORMANCE METRICS
The main challenge when comparing different hybrid nanocatalysts is to find quantitative and relevant performance metrics. Different aspects of a catalyst performance may be evaluated, such as its activity, selectivity, stability, or temperature of operation. However, it is not straightforward to determine metrics depending only on the nanocatalyst's intrinsic properties because its performance may be strongly altered by reaction conditions such as the illumination characteristics or reactant concentration. The typically reported performance metrics used in the field are discussed below. Table 2 summarizes the metrics that will be extracted from the original publications and written in tables for comparison.
Activity is quantified by the formation rate of a product (amount of generated product per time per mass of hybrid catalyst, eq 1). It must be noted that this metric assumes a linear relation of product formation with respect to time and catalyst concentration, which is often not the case. 104  (1) In this Review, we will use the formation rate (eq 1) as the main metric for activity, as it is most commonly reported. If it is not possible to extract the formation rate, the turnover frequency (TOF) or turnover number (TON = TOF × time) is reported in the tables. Across the literature, the TOF is used with conflicting definitions. In hybrid nanomaterials where the reaction occurs in only one of the components, the number of catalytic sites might be used instead of the total mass of the catalyst. This value is rarely reported as it is difficult to determine because the information on the number of active catalytic sites is not always available or might fluctuate in time.
Instead, TOF is more commonly referred to as the mole of product molecules per mole of catalyst per time (eq 2). The key role of illumination in plasmon-assisted catalysis is not reflected in the activity metrics discussed so far. Therefore, the quantum efficiency (QE) should be used as a complementary metric (eq 3). 105 It is defined as the number of reacted electrons per absorbed photon and determined for a specific wavelength. Whenever a quantum efficiency is reported, it is usually the apparent quantum efficiency (AQE), defined as the number of reacted electrons per incident photon (eq 4). This is the case because the number of absorbed photons is often difficult to determine. The AQE is a lower bound for the QE and will be used in this Review whenever it is reported.
In photoelectrocatalysis, the metrics reported are different from the ones for photocatalysis. A metric for the activity is the overpotential (mV) at a given current density (mA cm −2 ) (or vice versa). The onset potential (mV), the potential at which the reaction starts to occur, describes the reaction energy barrier. Finally, the Tafel slope (mV dec −1 ), obtained by plotting the overpotential as a function of the logarithm of the current density, is used to estimate the reaction kinetics of the system. Selectivity is a major concern for CO 2 conversion. For photocatalytic systems, it is generally reported as the ratio between the formation rate of the major product and all products in percent (eq 5).
For electrocatalytic systems, the Faradaic efficiency (FE) is more commonly reported. It describes the proportion of reacted electrons facilitating the formation of a certain product. The stability of the catalyst is referred to as the longest investigated time under operando conditions in which its activity did not significantly decrease. Whenever the stability was not examined, the experiment duration is reported in the table instead. The main challenge when comparing different hybrid nanocatalysts is to find quantitative and relevant performance metrics.
Temperature is an important parameter that modulates the rate of reactions through the Arrhenius law. Illuminated plasmonic systems offer the opportunity of driving reactions at mild conditions, without the need for high external temperatures. Here, whenever possible, the operating temperature of the reactor will be used as a metric. It must be noted that this value might strongly differ from the surface temperature of the nanocatalysts. 106 The light enhancement (LE) aims to quantify the contribution of light to a given reaction. When doing this, one should also compensate for the temperature increase upon illumination; 107 however, this is not frequently done. Here, it will therefore simply be estimated as the ratio between the activity (i.e., formation rate or TOF) of a nanocatalyst in the presence and in the absence of light.
Similarly, the plasmonic enhancement (PE) is the activity ratio of the illuminated system in the presence and absence of the plasmonic component. In photoelectrocatalysis, the enhance-ments are evaluated as a shift in the overpotential at a current density of 10 mA cm −2 of the reactions.
In the following, we will review the latest demonstrations of hybrid plasmonic nanomaterials for solar fuel production.

■ PLASMONIC METAL−METAL
In recent years, several configurations of hybrid plasmonic bimetallic nanoparticles ranging from alloys, core−shell, partially coated structures to self-assembled catalysts, have been employed to boost the H 2 production and CO 2 reduction through plasmon excitation (illustration in Figure 2a). A summary of these publications can be found in Table 3 and  Table 4, including the performance metrics defined in the introduction, as well as their respective experimental conditions. In the following, we will describe selected examples on these topics.
Plasmonic Metal−Metal Hybrids for H 2 Generation. Toward more efficient photocatalytic H 2 generation, plasmonic materials can benefit from the addition of late transition metals. Engineering the materials to exploit both lightharvesting and catalytic properties is one of the utmost challenges for these so-called plasmonic bimetallic systems. Next, we outline recent relevant reports in which different configurations were investigated for H 2 production.
An important study on alloys, carried out by Kang et al., showed that the efficiency of the Au-based alloyed catalyst was dependent on both the chemical composition of the hybrid and the hydrogen source. 108 Specifically, in their article Au was alloyed with different metals (Rh, Ru, Pt, Pd, Co, and Ni), and the resulting catalysts were tested toward the hydrolytic dehydrogenation of both ammonia borane (NH 3 BH 3 ) and sodium borohydride (NaBH 4 ). The best performers for both reactions are illustrated in Figure 2b. For the photocatalytic hydrolysis of NH 3 BH 3 , Au 1 Rh 1 , 1:1 atomic ratio alloys, showed the highest H 2 production (TOF = 240 min −1 ), followed by its Pt-analogue (Au 1 Pt 1 , TOF = 200 min −1 ). For both systems, the influence of light was evidenced by the almost 3-fold increase compared to dark conditions. When compared to their respective monometallic counterparts, Rh and Pt, the hybrids showed a plasmonic H 2 production enhancement of 2.4 and 1.8, respectively, indicating the benefits of such systems. However, for NaBH 4 dehydrogenation, Au 1 Ru 1 performed better than Au 1 Rh 1 , Au 1 Pt 1 , and the other alloys, achieving a production of TOF = 178 min −1 . In addition to its superior H 2 production, the Au 1 Ru 1 hybrid was the catalyst that maximized the use of the incoming light. In comparison to the experiments in dark, the activity increased by a factor of 3.8 under illumination. Again, compared to the monometallic RuNPs, the inclusion of the plasmonic Au led to an improvement of 2.2-fold. This improved H 2 production was a result of hot carrier injection from Au into the catalytically active Ru atoms, weakening the B−H bonds in the borohydride.
Unlike alloys, in core−shell structures, the adsorbates can only interact with the outermost metal, which typically is the catalytic metal. This introduces an important difference between the systems. Instead of isolated active centers, the entire surface of the core−shell structure becomes active for catalytic processes. Xu et al. fabricated a hybrid plasmonic photocatalyst by coating an Ag nanocube with a thin layer of Pd (Ag@Pd NC). 109 Panels i and ii of Figure 2c illustrate the well-defined spatial arrangement of the metals, showing that Pd was placed only at the surface, encapsulating the Ag optical antenna. The Ag@Pd NCs were tested toward the hydrolytic dehydrogenation of NH 3 BH 3 , and changes in the activation barrier under visible light illumination were studied. The Ag@ Pd NCs showed a H 2 production rate of 102 × 10 6 μmol g −1 h −1 at 40°C, which is one of the highest rates in Table 3. This production under illumination was 1.2 times larger compared to dark conditions. In spite of this outstanding production, the most interesting finding is that the activation energy (E a ) was lowered from 50 kJ mol −1 to 31.8 kJ mol −1 under simulated sunlight illumination, as shown at the bottom of Figure 2b. The authors attributed the lowered E a barrier to the fact that excited electrons on Ag accumulated within the thin Pd shell. Those excited carriers would then populate the lowest unoccupied molecular orbital (LUMO) of the Pd−NH 3 BH 3 complex at the interface, thus facilitating the H 2 production.
By coating a plasmonic antenna with a catalytic material, the optical properties of the overall system are damped. Instead, partially coating the plasmonic components allows us to preserve the plasmon and still exploit the interface between both. Tong et al. 110 compared the performance of hybrid systems in which the catalytic metal was either homogeneously or heterogeneously grown on a plasmonic antenna. Both Pdcovered and Pd-dotted Ag@Au hexagonal nanoparticles were fabricated and tested toward the generation of H 2 out of formic acid (HCOOH) (Figure 2d). The Pd-covered Ag@Au showed a H 2 production of 15 × 10 3 μmol g −1 h −1 when illuminated with visible light at 50°C with a light enhancement factor of 1.15. However, when the Pd was shaped into dots instead of a continuous shell, the performance was further improved up to 113 × 10 3 μmol g −1 h −1 . The control of the morphology of the active metal not only enabled a boost in H 2 generation, but also resulted in a larger light enhancement. Indeed, the improvement of 4.5-fold in its activity under visible light illumination is the largest light enhancement factor among the examples reported in Table 3. The anisotropic distribution resulted in a lower damping of the LSPR in comparison to the full Pd shell and therefore in higher spatially inhomogeneous electric field at the surface of the Pd dots. According to the authors, the enhanced adsorption of HCOOH on Pd due to the local fields and an increased hot carrier injection rate from the plasmonic to the catalytic metal were crucial to improve the H 2 production.
Until now, most of the reports utilized Au or Ag as plasmonic metals. However, there is a substantial need to move toward earth abundant metals for sustainable plasmon-driven catalysis. While copper-based systems are the most promising choice due to its inherent resonance within the visible region, the chemical stability of these systems still remains challenging. Luo et al. recently employed Cu−Zn alloys to perform methanol reforming under simulated sunlight illumination. 113 It was found that the chemical composition of the alloyed nanoparticles influenced the H 2 production. The largest production (328 × 10 3 μmol g −1 h −1 ) was observed for the Zn 1.3 Cu 98.7 sample. Both the large productionsecond best in Table 3and its promising energy conversion ability reinforce the shift toward earth abundant plasmonic metals.
All these examples shed light on the design of bimetallic hybrid nanocatalysts. The optimization of catalysts is specific to their application, as an excellent catalyst for H 2 generation from one H 2 source might not be suitable for H 2 generation from other H 2 -containing compounds. It has been demonstrated that tailoring the chemical composition and the spatial arrangement of the materials can be beneficial to enhance the interaction between the adsorbate and the catalyst, leading to better catalytic performances. From the reviewed articles included in Table 3, partially coated hybrid bimetallic catalysts seem to be the most promising configurations for sunlightdriven photocatalysis. Compared to fully coated catalysts, these structures only slightly damp the LSPR while still exploiting the interface between the metals for efficient hot carrier transfer.
Plasmonic Metal−Metal Hybrids for CO 2 Reduction. In recent years, the ability of PNP to generate energetic carriers when resonantly illuminated has been exploited in multielectron processes such as CO 2 reduction. However, to our knowledge, to date there appears to be no evidence of plasmonic hybrid bimetallic photocatalysts employed for this reaction. Thus, in this section we will briefly discuss and summarize some relevant monometallic examples in this field (see Table 4).
Yu et al. reported one of the first examples. 111 Here, a colloidal suspension of 12 nm AuNPs coated with polyvinylpyrrolidone (AuNP-PVP), was employed to reduce CO 2 . This was possible due to the stability provided by the surfactant when illuminated in long-hour experiments. One of the most striking features of this report was the tunability of the reaction pathway as it was dependent on both the energy of Ag laser (@514.5 nm, 10 mW) Au-TiO 2 (O) 300 W Xe lamp (420 > λ > 320 nm) the photons and the photon flux. As can be seen in Figure 2e, when illuminated with visible light (λ > 400 nm, 300 mW cm −2 ), the AuNPs promoted the formation of both CH 4 and ethane (C 2 H 6 ). To understand the mechanism, the researchers further studied the role of the plasmon and interband transitions in the reaction pathway by varying both the wavelength and power. When using green photons matching the LSPR (λ = 532 nm), the AuNPs only produced CH 4 irrespective of the power. However, when the incident photons were more energetic (λ = 488 nm), AuNPs not only were able to produce CH 4 , but also promoted C−C coupling at high powers (>300 mW cm −1 ), resulting in C 2 H 6 production. The possibility to conduct a 14 e − (C 2 H 6 ) instead a 8 e − (CH 4 ) reduction was attributed to longer lifetimes of energetic charge carriers when exciting the interband transitions compared to the hot carriers created by means of LSPR. Hybrid plasmonic systems can potentially decrease charge recombination and therefore could allow the production of C2+ products by plasmon excitation of bimetallic systems in the future. Dhiman et al. investigated the influence of hotspots on the methanation of CO 2 using dendritic plasmonic colloidosomes (DPC) (see Figure 2f). 112 Clustering the NPs allowed the utilization of not only the visible, but also the NIR region of the solar spectrum. Interestingly, they found that the performance of the system was dependent on the Au load and the strength of the hotspots created within the interparticle gaps. Specifically, only the sample with ∼8 nm AuNPs (DPC-C4, four growing steps) enabled the reduction of CO 2 to generate CH 4 , as shown in Figure 2f. The CH 4 production of 1.50 μmol g −1 h −1 is one of the highest in Table 4, although it is worth mentioning that the experiments were conducted at 85°C. Conversely, colloidosomes with either smaller or larger nanoparticles were not able to reduce the CO 2 neither to CO nor CH 4 . The performance of the 8 nm AuNPs colloidosomes was attributed to the synergistic effects of the optimized electromagnetic and thermal hotspot coupled to the hot carriers transfer for Au surfaces to adsorbed CO 2 . It has been demonstrated that multielectron processes can be driven by PNPs, although there is still plenty of room to explore toward the fabrication of plasmonic bimetallic photocatalysts for CO 2 reduction. Proper engineering of the hybrid should focus on ways to increase the pumping of hot carriers toward the active material. Also, the strong adsorption of CO on metallic surfaces must be taken into account for proper materials selection. This phenomenon should not be underestimated as it leads to the inactivation of the active sites.
An emerging way to engineer the materials for C−C coupling is to create nanometer gaps between the plasmonic and catalytic metal surfaces. In this type of catalyst, the reduction of CO 2 at the surface of the PNP can create intermediateenriched environments nearby the catalytic metal surface. This is usually referred to as in-tandem reactors.

■ PLASMONIC METAL−SEMICONDUCTOR
The charge transfer from a plasmonically excited metal nanoparticle to a semiconductor, schemed in Figure 3a, has been beneficial for increasing the photocatalytic production of solar fuels. The pursuit of an optimized hybrid photocatalyst Experiment duration. b Negligible product detected without plasmonic material. c Neglibible product detected without light. d QY of hot electron injection from transient absorption spectra. has led to the combination of PNPs with several semiconductors such as TiO 2 , CdS, and Cu 2 O, among others. Recent examples of these types of hybrid materials are summarized in Table 4 and Table 5, as well as the performance metrics defined in the introduction. In the following, selected examples for both H 2 generation and CO 2 reduction will be discussed.
Plasmonic Metal−Semiconductor Hybrids for H 2 Generation. As shown in Table 5, more than 60% of the hybrids for plasmon assisted photocatalytic H 2 production utilized TiO 2 as the semiconductor component, owing to its unique photochemical and photoelectrochemical properties. Herein, the best photocatalyst for hydrogen production was realized by Qin et al. 130 They designed and employed an efficient photocatalyst for plasmon-enhanced visible light based on Pt nanoparticles (PtNPs) supported on TiO 2 photocatalysts (Pt/TiO 2 ). While anatase-structured TiO 2 is only active in the UV, the inclusion of 100 nm plasmonic PtNPs allowed to extend the photoactivity of the hybrid to the visible. The improved H 2 formation rate of 356 × 10 3 μmol g −1 h −1 , the highest among all the reviewed metal−semiconductor examples, was a result of transferring hot electrons from Pt to the conduction band of TiO 2 . To understand the underlying mechanism, the PtNPs were replaced by AuNPs (50 nm). Unexpectedly, this system did not produce as much H 2 as its Pt analogue. While the excitation of the plasmon in both metals enabled the injection of energetic electrons into the conduction band of TiO 2 , their different Schottky barrier strongly affected the photocatalyic performance. In short, the higher Schottky barrier of the Pt-TiO 2 heterojunction of ϕ Pt = 1.7 eV (compared to ϕ Au = 0.9−1.0 eV with Au) inhibited the recombination of the injected electrons in TiO 2 with the holes in the metal. This led to an extended lifetime and larger extraction toward H 2 production. Remarkably, in addition to its high production, it is also one of the most stable hybrids of this section (56 h).
Negrin-Montecelo et al. published an important study on the critical influence of morphology and composition of the plasmonic component in PNP−TiO 2 nanostructures. 131 Herein, the authors interfaced both Au nanorods (NR) and core− shell Au@Ag NR with TiO 2 and tested them toward the dehydrogenation of HCOOH (Figure 3b). In their experiments, it was found that by coating a AuNR with a Ag shell, the H 2 production was increased from roughly 36.5 × 10 3 μmol g −1 h −1 to 64.9 × 10 3 μmol g −1 h −1 . As expected, TiO 2 on silica beads was not able to trigger the reactions when illuminated with photons below the energy of the bandgap. This reinforces the role of the plasmonic component as the photosensitizer material. Theoretical calculations revealed that the Au@AgNRs were able to create both more and higher energy electrons than the bare AuNRs, facilitating the injection of hot electrons into the conduction band of the semiconductor.
Recently, CdS has emerged as a promising semiconductor to be exploited in the generation of H 2 , as seen in the first five examples listed in Table 5. Yang et al. synthesized hybrid dual plasmonic antenna consisting of partially CdS-coated Au satellites assembled on SiO 2 -coated Ag nanoparticles (Ag@ SiO 2 /Au@CdS with pinholes). 132 Panels i and ii of Figure 3c show the spatial arrangement of the metals. When tested for the HER, this structure yielded one of the highest performances for this type of PNP−SC hybrids (191 × 10 3 μmol g −1 h −1 , Figure 3c). It was possible to improve the formation rate by 7346 times, as evidenced by the plasmonic enhancement factor. The combination of these three materials resulted in an optical response extending from the blue to the red of the visible region. The inherent absorption of the CdS was enhanced due to the overlap with the Ag resonance, and hot carriers were pumped into the conduction band of the semiconductor by the Au satellite, leading to a notable improvement in the performance. Finally, when the Ag antenna was replaced by a Au optical antenna in the hybrid, the performance was slightly inferior (143 × 10 3 μmol g −1 h −1 ). The authors attributed this result to the poor overlap between the Au LSPR and the CdS absorption band in the blue region, as shown in the extinction spectra in Figure 3c.
In conclusion, tremendous advances have been made in the combination of PNP with semiconductors toward the generation of H 2 . Due to its high photoactivity, low cost as well as excellent chemical stability, TiO 2 is the most used semiconductor material in hybrid plasmonic semiconductor systems so far. Nevertheless, novel hybrids with promising performances are arising. Challenges to be addressed in the near future are the susceptibility of CdS to corrosion under light illumination 150 and the high recombination rate of carriers of ZnO, 151 limiting the chemical stability and catalytic activity, respectively.
Plasmonic Metal−Semiconductor Hybrids for CO 2 Reduction. The creation of an heterojunction between metals and semiconductor is a promising strategy to suppress the HER thereby enhancing the photocatalytic CO 2 RR. Both the activity and selectivity can be tuned to favor a desired product of the reaction. Recent reports on the application of these hybrids are summarized in Table 4.
One of the most remarkable examples in terms of selectivity and production of carbon-based products was recently reported by Wang et al. 123 The authors synthesized Au/ CdSe-Cu 2 O dumbbell nanorods (see Figure 3d, panels iii, iv, v) which were utilized for CO 2 RR under visible illumination (λ > 420 nm) at different H 2 O %wt in the reactant solution. Throughout the tests, H 2 was never detected, and CO as well as more electron-demanding products such as CH 4 were produced instead. As shown in Figure 3d, the production reached maximum values of 80.1 and 42.9 μmol g −1 h −1 for CO and CH 4 , respectively. The water provided the H atoms for the 8e − /8H + reduction of CO 2 to CH 4 . By combining Au with these semiconductors, a plasmonic enhancement of 1.92 was achieved. The results obtained for this hybrid (Au/CdSe-Cu 2 O dumbbell nanorods) were explained by the charge transport in the materials. In short, all three components can be excited simultaneously under visible illumination, and as a result, the photogenerated electrons accumulate in Cu 2 O due to the recombination of its respective holes with electrons injected from Au and CdSe.
Chen et al. 126 reported a systematic study where the reactivity and selectivity of TiO 2 with (101) facets were found to increase with the photoreduction of Au−Pd nanoalloys of different ratio compositions. By anchoring the Au−Pd cocatalyst on TiO 2 (101), the Au−Pd alloys with the stoichiometric ratio of 6:1 provided a maximum selectivity toward hydrocarbons of 85% in the CO 2 RR under visible light illumination (see Figure 3e). In addition to the detection of CH 4 , which was the main product (12.9 μmol g −1 h −1 ), ethane (C 2 H 6 ) and ethylene (C 2 H 4 ) were detected, evidencing the capability to perform C−C coupling. While the production of CH 4 was still observed when only Pd was reduced on top of TiO 2 (101), the inclusion of Au led to a notable 5.2-fold increase.
The application of hybrids toward the reduction of CO 2 is still under development, and so far only a few combinations of metals and semiconductors have been explored. The majority of the examples listed in Table 4 employed both Au and TiO 2 NPs, and only one out of seven used Ag as the plasmonic component. It is also worth mentioning that in only one report the photocatalyst was tested under simulated solar illumination conditions. It is necessary to aim for sustainable ways to carry out these processes. In this regard, cheap semiconductors such as TiO 2 and plasmonic metals (e.g., Al or Cu) could emerge as candidates. Cu-based systems are promising especially for C2+ products and because of the abundancy of Cu. However, to our knowledge, they have not been explored yet. At last, the design of novel heterostructures should be carefully addressed to maximize the hot carriers injection in the active centers, as well as the lifetime of hot carriers created by means of plasmon excitations. This will lead to higher efficiencies and will bring the hybrids one step closer to industrial applications.

■ PLASMONIC METAL−PEROVSKITE
The synergistic effects of the combination of the lightharvesting properties of PNPs with the optoelectric and catalytic properties of perovskite materials have resulted in the investigation of various geometrical and material arrangements (see Figure 4b). In Table 6, the hybrids that have been tested toward H 2 generation and CO 2 reduction are summarized using the metrics specified in the introduction. In the following, examples with noteworthy properties in terms of formation rate, material composition, or operational wavelength range will be discussed. To our knowledge, most systems studied to date for solar fuels production are based on perovskite oxides. Halide perovskites are another promising set of materials due to the fact that their tunable absorption across the visible can be enhanced by the local field enhancement of PNPs. This has been shown to be useful in photovoltaic applications. 152−154 However, the current bottleneck for photocatalytic applications is their instability. Under external stimuli such as moisture, halide perovskites are easily decomposed, which is detrimental for many catalytic applications like water-splitting. This may be one of the reasons why there is no demonstration of hydrogen production on halide perovskites yet, and that the first demonstration of CO 2 reduction in such a system was published in 2020. 155 Plasmonic Metal−Perovskite Hybrids for H 2 Production. Plasmon−perovskite hybrids investigated toward H 2 production recently are shown in Table 6. Out of the eight different pervoskite oxide materials shown in Table 6, seven 70,156−163 are wide band gap semiconductors which require UV light for optical excitation. Extending the light absorption into the visible region with plasmonic hybrids is a promising strategy to increase the solar-to-fuel conversion efficiency (illustrated in Figure 4c). To probe whether hybridizing such a wide bang gap perovskite with a plasmonic material allows for photocatalytic H 2 O reduction using only visible light, Wang et al. synthesized SrTiO 3 nanocubes both with and without AgNPs. 157 When illuminated with a 300 W Xe lamp filtered to wavelengths longer than 420 nm, the SrTiO 3 nanocubes alone predictably did not produce any hydrogen. The Ag/SrTiO 3 nanocomposite, however, was able to catalyze H 2 production and reached a formation rate of 264.5 μmol g −1 h −1 . Only one other perovskite hybrid system was reported to reach a comparable high formation rate at wavelengths longer than 420 nm (see Table 6). This hybrid photocatalyst investigated by Cai et al. not only allowed the utilization of visible light but extended the usable range into the infrared (IR). 156 This was achieved by a hybrid system consisting of Au nanorods (AuNRs) on lanthanum titanium oxide (La 2 Ti 2 O 7 , LTO) nanosteps, depicted in panel ii of Figure 4b. Figure 4c shows how the absorption peak shifts when AuNRs of different plasmon resonances are incorporated in the system. It illustrates how the geometry of the PNP can be used to tune the absorption of a plasmon-perovskite hybrid system across the solar spectrum. The system reached its maximum apparent quantum efficiency (AQE) of 1.4% at 920 nm, the wavelength coinciding with the longitudinal plasmon resonance of the employed AuNRs. While neither the AuNRs nor LTO individually produced a significant amount of H 2 when illuminated with visible light, the combined system did, as can be seen in Figure 4d. The synergistic effect arose because the low energy photons could not photoexcite the system without the AuNRs due to the large bandgap of LTO and, on the other hand, the hot electrons generated in the AuNRs under resonant illumination could only be extracted effectively through the interface with LTO.
Most plasmonic systems are restricted to noble metals, making large-scale industrial applications expensive. This has resulted in a search for cheaper plasmonic alternatives. As far as we know, the only example of a plasmon−perovskite hybrid for H 2 production utilizing non-precious materials was published by Guo et al. 160 By combining barium titanate (BaTiO 3 ) with aluminum, a plasmonic H 2 -production enhancement factor of 6.8 was achieved, showing the potential of non-noble metals. Additionally, the use of BaTiO 3 allows for the combination of plasmonics with piezoelectronics, where mechanical stimuli significantly enhance the photogenerated carrier separation and transfer, further increasing the H 2 -production rate by a factor of ∼2.
The Al/BaTiO 3 hybrid as well as eight 70,159,164,165 out of the 15 systems shown in Table 6 were illuminated using UV− visible−IR light, optically exciting both components. One effect that should not be neglected when discussing plasmonic hybrid systems is the effect the metal has on the system, even when the plasmon is not excited. When a semiconductor is excited above the band gap, the metal can act as a charge trap, increasing carrier lifetimes. Xu et al. 158 probed this by testing the H 2 production on KTaO 3 and NaTaO 3 decorated with AgNPs for both visible−NIR and UV−visible−NIR excitation. In both cases, the addition of AgNPs increased the formation rates. However, the H 2 formation rates of the composites under UV−visible−NIR light were higher than under purely visible−NIR light. The enhancement in H 2 production by the addition of AgNPs is therefore described as a combination of AgNPs acting as co-catalysts by increasing the charge separation of electron−hole pairs generated in the perovskite by excitation with UV light, as well as absorption in the visible and subsequent hot electron transfer from the AgNPs. The goal is to eventually use the full solar spectrum spanning from the UV to the IR for the photocatalytic production of H 2 , and hence all synergistic effects can and should be exploited. However, at the current stage, still far from industrial applications, systematic wavelength-dependent studies reporting not only on production rates but also the quantum efficiencies, like the work of Cai et al. 156 and Zhong et al., 164 are needed to understand the underlying mechanisms and design catalysts accordingly.
The studies conducted to date and presented above as well as in Table 6 show that plasmon−perovskite hybrids can improve H 2 production rates compared to the pristine perovskites by enhancing the absorption across the solar spectrum and improving charge carrier dynamics. While they are limited to perovskite oxides and H 2 generation from water, these findings show the potential of exploring also other H 2 generating reactions and eventually perovskite types like perovskite halides.
Plasmonic Metal−Perovskite Hybrids for CO 2 Reduction. The combination of a plasmonic material with a perovskite for photocatalysis has shown to be beneficial not only for H 2 production but can also enhance other photocatalytic processes like dye or isopropanol degradation, 168−171 or the removal of algae. 172 However, as far as we know, there are only three publications on photocatalytic CO 2 reduction using plasmon-perovskite hybrids. They are shown in Table 6. The first, by Wan et al. was briefly discussed above, as the same catalyst was also investigated toward H 2 production. 157 As expected, no CO 2 reduction occurred under visible−NIR light for the SrTiO 3 nanocubes. For the Ag/SrTiO 3 hybrid, three products were detected: CO, H 2 , and CH 4 . As commonly reported for plasmonic photocatalysis with perovskite oxides, the enhanced photocatalytic reduction activity is attributed to improved visible light absorption and efficient charge separation. The competing HER increases with increasing Ag loading, which is explained by an increase in the recombination Liao et al. published the first halide perovskite-based plasmonic photocatalyst in 2020. In their study, the experiments were not carried out in an aqueous solution but in a mixture of acetonitrile and isopropanol. CsPbBr 3 nanocubes were decorated with AuNPs (see panel i in Figure 4b) and tested toward their CO 2 reduction performance with both visible−NIR light (λ > 420 nm) and light below the bandgap of CsPbBr 3 (λ > 580 nm). 155 The results are in agreement with the findings of Xu et al. 158 (discussed above for H 2 production), where the increase in activity due to the presence of the AuNPs could be attributed to both charge transfer from the perovskite to Au and charge transfer from Au to the perovskite, dependent on which entity was optically excited.
For full visible−NIR light illumination, the introduction of AuNPs yields a 3.2-fold enhancement. The results for below bandgap excitation with wavelengths longer than 580 nm can be seen in Figure 4e. As expected, no CO 2 reduction can be detected for the perovskite alone. The Au/CsPbBr 3 extends the usable wavelength range further into the visible, as indicated by the CO 2 conversion rate. Since then, these findings have been supported by Tang et al., who used a macroporous CsPbBr 3 framework combined with AuNPs 167 instead of Au/CsPbBr 3 nanocubes. With the first application of plasmonic perovskite hybrids toward the reduction of CO 2 published in 2019, the current systems lack in efficiency but are proof that there is a lot of room to explore. Thus far, the main product is CO. Whether a higher selectivity toward C2+ products can be achieved remains to be answered.   The recent publications on hybrid plasmon perovskite structures reviewed here show that the combination of plasmonics and perovskites is a promising strategy for the photocatalytic production of solar fuels. The systems explored so far are mostly limited to perovskite oxides. Their abundance makes them attractive materials for large-scale photocatalysis. To enhance their solar catalytic performance while keeping costs low, the combination with abundant plasmonic materials like aluminum will be an interesting and potentially fruitful route to explore. Apart from the opportunities for perovskite oxides, halide perovskites for plasmon-assisted photocatalytic applications are another promising field. Their tunable absorption across the visible can be enhanced by the local field enhancement of PNPs, which has been shown in photovoltaic applications. 173 The current bottleneck for photocatalytic applications is their instability. Many efforts are currently made to improve their stability 174 and decrease their environmental toxicity, 175 which will allow further exploration of their combination with plasmonic materials for the production of solar fuels.

■ PLASMONIC METAL−2D
This section covers metal−2D hybrid-based solar H 2 production and CO 2 reduction (see Figure 5a). Conventionally, 2D materials refer to single-crystalline materials that consist of single or several layers of atoms. Strictly speaking, the electron movement within the material should be only confined along the perpendicular direction of the 2D plane. 176 These atomically thin crystalline 2D materials are commonly obtained by mechanical or chemical exfoliation, and due to this nature, their reproducibility is highly limited. Some argue that nanosheets and -plates can also be referred to as 2D materials as long as a certain degree of quantum confinement along the thickness direction is maintained. 177 These types of materials are studied intensively in recent years due to their mass producibility. 178,179 Therefore, in the following, nanosheets and nanoflakes will also be categorized as 2D materials. Tables  7, 8, and 9 summarize some of the recent advances on metal− 2D hybrid systems. This section aims to cast insight into tuning and controlling the properties of plasmonic metal−2D hybrids for photocatalytic applications. As can be seen from the tables, overall, by combining plasmonic metal nanostructures and 2D materials, the photocatalytic performances of the hybrid systems have improved compared to their 2D singlecomponent counterparts.
Plasmonic Metal−2D Hybrids for Hydrogen Production. By tuning the shape and size of the metal nanoparticles, the solar light absorption profiles of the hybrid system can be modified to ensure the optimum synergy in between metal and 2D materials. 180−184 To study the effect of the metal nanoparticle shape, Zhang et al. compared the photocatalytic H 2 generation performances of 2D Molybdenum disulfide (MoS 2 ) decorated with Au nanospheres (AuNSs), nanorods (AuNRs), and triangles (AuNTs). 181 To eliminate the band gap excitation of MoS 2 , visible light was used, and negligible amounts of H 2 were detected using either pristine Au nanostructures or MoS 2 . However, as shown in Figure 5c, all three hybrid structures were able to catalyze H 2 generation. Among the three types of Au nanostructures, AuNR-MoS 2 exhibited the best performance. The H 2 generation rate of AuNR/MoS 2 was 1.6 and 3.7 times higher than those of AuNT/MoS 2 and AuNS/MoS 2 . This superior hydrogen generation of AuNR-MoS 2 suggests that efficient plasmon-induced hot-electron transfer at the metal−2D interface is promoted by the stronger electric field from AuNRs (see Figure 5b). The authors also found that the slower recombination of electrons and holes in AuNRs contributed to charge separation in both AuNR and MoS 2 across the Schottky barrier.
The mass ratio between the metal particles and the 2D materials is another determining factor of the photocatalytic reaction kinetics. 182,212,218,220 Guo et al. examined the role of the metal particle loading amount. They loaded 18 nm AuNPs with varying mass percentages ranging from 0.5 to 4.0% onto graphitic carbon nitride (g-C 3 N 4 ) nanosheets and investigated them toward photocatalytic H 2 generation. As can be seen from Figure 5d, the H 2 evolution rate increased until the loading amount reached an optimum and then decreased with continuing increase of the mass ratio of AuNPs. The number of the particles per area increases as the loading amount increases. After a certain loading amount, the close vicinity of the metal particles promotes electron−hole recombinations. 226 This finding also implies that the optimum loading mass increases with increased size of the metal nanoparticles. This is because for larger particles to achieve the same number density, the required total mass of the loaded particles will also increase. The effect of metal nanoparticle size is described in more detail in the CO 2 reduction section.
Co-loading or alloying metal−metal nanoparticles can further influence the photocatalytic activity of the hybrid system. 187,200,203,207,227 Cheng et al. demonstrated that the H 2 generation could be benefited from co-loading Au clusters and AuNPs onto g-C 3 N 4 . 228 Au cluster here refers to an "ultrasmall nanoparticle" that consists of tens to hundreds of gold atoms and has quantized energy levels. The authors argued that this configuration allows Au clusters to act as electron acceptors to reduce the charge recombination within AuNPs; hence, increasing the hot-electron lifetime within the particles. The H 2 generation rate of the AuNP cluster-NP/g-C 3 N 4 increased by a factor of 6 compared to AuNP/g-C 3 N 4 and by 20 times compared to Au cluster/g-C 3 N 4 . This was attributed to a combined effect of a lowered Schottky barrier between Au and g-C 3 N 4 via strong sp 2 hybridization and the prolonged hotcarrier lifetime. This enhancement mechanism also applies to bimetallic system such as AuPd loaded g-C 3 N 4 . 229 For instance, Han et al. demonstrated that by alloying Au and Pd, the H 2 production rate of the alloy−g-C 3 N 4 hybrid nanocomposite was increased by 3.5 and 1.1 times as compared to Au−g-C 3 N 4 and Pd−g-C 3 N 4 , respectively. Their investigations on the nanocomposites indicated that the lifetime of the photogenerated hot carriers in the AuPd alloy was longer than those in Au and Pd and therefore inhibited the recombination of the hot carriers and facilitated greater photocatalytic H 2 generation.
Due to the high cost of noble metals, researchers also started to explore non-noble plasmonic metals. 186,212 In Tables 7 and  8, less than 5 out of more than 30 hybrid systems studied toward H 2 production are non-noble-metal-based. Xu et al. demonstrated that Cu as a non-noble metal is also capable of enhancing photocatalytic water-splitting when combined with 2D tungsten disulfide (WS 2 ). 186 Their Cu-WS 2 heterostructure improved the H 2 generation rate by 40-and 2.2-fold as compared to the WS 2 and Cu single-component system under simulated 1 sun irradiation (Figure 5f). This study not only showed the potential of non-noble metals as a proof of principle, it also reached one of the highest H 2 production rates (64 000 μmol g −1 h −1 ) and exhibited one of the highest plasmonic activity enhancements (40) in Table 9.
To conclude, upon combining metal nanostructures and 2D materials, the hybrid systems exhibited improved HER performances. Plasmonic materials, due to their superior light absorption and unique optical properties, can enhance the light-harvesting of 2D materials and extend their usable wavelength range. Through proper engineering of plasmonic metal−2D hybrid systems, the recombination rate of the photogenerated carriers can also be reduced significantly, which results in improved H 2 production.
Plasmonic Metal−2D Hybrids for CO 2 Reduction. In the H 2 generation section, it was briefly mentioned that the size of the metal nanoparticles can also influence the photocatalytic activity of plasmonic metal−2D hybrid systems. To examine this, Li et al. grew AuNPs of various sizes on g-C 3 N 4 nanosheets for CO 2 reduction. 183 They showed that AuNPs with 4−6 nm diameter exhibited the best CO 2 reduction performance under visible light irradiation ( Figure  5e). The Fermi level of the smaller particles is lower. This means the Schottky barrier between AuNP and g-C 3 N 4 is smaller and the photogenerated electrons are more readily transferred from g-C 3 N 4 to AuNPs.
As concluded in multiple of the aforementioned studies, the band structure at the Schottky junction is crucial to the photocatalytic activity of the system. A material-dependent study by Sun et al. comparing the CO 2 photoreduction performance of Au−MoS 2 and Ag−MoS 2 further demonstrated the role of the band structures at the heterojunctions. 74 This study showed that the band structure at the metal−2D interface can impact the formation rates of specific reduction products and their selectivity (Figure 5g). At the Au−MoS 2 interface the band structure prevented the electron transfer from MoS 2 to Au, whereas the hot-electrons from Au could inject to MoS 2 . This resulted in electron accumulation on MoS 2 upon light irradiation. On the other hand, at the Ag− MoS 2 junction, the electron transfer from MoS 2 to Ag was favored resulting in less electrons being accumulated at MoS 2 . Subsequently, in the Ag-based system, mainly CO was generated, whereas in the Au-based system CH 4 was the major product. This result shows that the selectivity in CO 2 reduction processes can be carefully controlled by tuning the composition of plasmonic metal−2D nanocomposites.
Optimization of the Schottky junction at metal−2D interfaces can prolong the hot carrier lifetime and improve the charge transfer to promote the photocatalytic performance of the hybrid systems. Moreover, particularly for CO 2 photoreduction, the band structure at the interface plays a significant role in selectivity. This provides us with immense opportunities in photocatalytic system development.
Even though the plasmonic metal−2D hybrid systems have exhibited excellent potential for synergetic photocatalytic performance, the large-scale reproducibility of conventional 2D materials still remains as a bottleneck. Both "top-down" exfoliation and the "bottom-up" synthesis are not easily scalable yet. Although tremendous efforts have been devoted to the mass production of 2D materials, there remain substantial challenges before they can make a real impact in industrial applications. 235 Chemically synthesized nanosheets and nanoplates have been gaining popularity as alternative 2D materials due to their producibility and photocatalytic performance. Therefore, chemically synthesized 2D materials with good crystallinity and dispersity and their plasmonic hybrid systems deserve further exploration for photocatalytic applications.  Figure 6a). 81,82,236 In the following, a unified nomenclature is used to increase readability. Particles assembled on other compounds are positioned after a "/"-sign while incorporated components are written before an "@"-sign. From left to right, the components are arranged from core to shell. Furthermore, only the common names and important features of the respective MOFs are mentioned in the text as another measure to increase readability. For example, MIL-125, a framework built up from 1,4-benzenedicarboxylate (organic linker) and TiO 2 (metal node), will be described as a semiconductor-like MOF.
Plasmonic Metal−MOF Hybrids for H 2 Production. Recently explored hybrids that exploit the combination of PNPs and MOFs to boost the performance of photocatalytic H 2 production are summarized in Table 10. One of the highest H 2 production rates was reported for Pt@MIL-125/Au. The hybrid catalyzed the water-splitting reaction with a H 2 formation rate of 1743 μmol g −1 h −1 . 89 In the optimized hybrid, PtNPs were incorporated into a porous semiconductorlike MOF (MIL-125), and the structure was subsequently decorated with rod-shaped AuNPs (see Figure 6b). The porosity of the MOF allows the reactants and products to access and leave the catalytic sites, respectively. Under visible light radiation, hot electrons generated at the AuNPs were steered through the lowest unoccupied molecular orbital (LUMO) of the MOF toward the PtNPs. Due to the difference in work functions of the MOF (semiconductor-like) and the Pt (metal), electrons left the MOF in favor of the Pt surface and were trapped there (Schottky junction; see Figure 6b). This electron transport pathway resulted in an accelerated unidirectional charge transfer and an amplified e − −h + separation. The extent of e − −h + separation was greatly influenced by the spatial separation of AuNPs and PtNPs. Figure 6b shows the impact of incorporating the PtNPs into the MOF compared to assembling both the PtNPs and the AuNPs on the MOF surface (MIL-125/Pt/Au), where the formation rate is about 10 times lower. Figure 6b also illustrates the role of the plasmonic component in harvesting the light as no catalytic activity was reported without the AuNPs (0 μmol g −1 h −1 for Pt@MIL-125).
Like the previous hybrid, UiO-66(Zr 85 Ti 15 )/Au@Pd was designed to enrich the surface of the catalytic compound (here Pd) with electrons. 86 Here, a MOF (UiO-66(Zr 85 Ti 15 )) with photoactive linkers and doped metal nodes was decorated with plasmonic core−shell Au@Pd nanoparticles resulting in the second highest formation rate reported (see Table 10). This combination resulted in several processes upon visible light radiation that boosted the hydrogen formation from formic acid.

ACS Energy Letters
http://pubs.acs.org/journal/aelccp Review hybrid has a 12.6 times higher production rate with Au compared to without. The resulting formation rate of 3315 μmol g −1 h −1 is the second highest reported in Table 10. More importantly, the hybrid shows how the building blocks of MOFs can be smartly designed to, for example, promote certain steps in the synthesis, facilitate charge transfer, or enhance absorption properties in combination with PNPs. While the concepts used in the previously discussed hybrids are promising, by far the highest formation rate with 25 000 μmol g −1 h −1 was reported for MIL-101/Au@CdS catalyzing the water-splitting reaction. 237 Here, Au@CdS core−shell particles were assembled on a MOF structure (MIL-101). The hybrid exploited the already highly reactive catalyst CdS (7400 μmol g −1 h −1 ; semiconductor) for achieving high formation rates. By assembling CdS on the MOF-surface, the formation rate was enhanced 2.7-fold. The MOF acts as a support, resulting in well-dispersed CdS particles and therefore possibly more active catalytic sites. By absorbing visible light, the MOF generated excited electrons that enriched the surface of CdS, therefore facilitating the reaction. Introducing AuNPs in the form of core−shell Au@CdS particles enhanced the reactivity by another 30%. The authors propose that the AuNPs accelerated the charge transfer and additionally extended the response spectrum of CdS by the LSPR.
When comparing the values in Table 10, two approaches for H 2 production utilizing PNP−MOF hybrids stand out. In absolute numbers, introducing an additional material that is already highly catalytically active (e.g., the semiconductor CdS) 237 that can be further enhanced by PNPs and MOFs can result in very high formation rates. In relative numbers (enhancement), targeting single, crucial steps in reactions by introducing functional groups to the MOF seems especially promising. 86 In all systems reported for H 2 production, the hybrid consisted of three components, with the PNPs assembled on the MOF-surface (see Table 10). Different, so far unexplored geometries and structural assemblies could result in new features that improve the catalytic performance.
Plasmonic Metal−MOF Hybrids for CO 2 Production. One of the most recognized works and one of the first PNP− MOF hybrids used as photocatalyst for CO 2 conversion was Ag@Re 3 -MOF (see Figure 6c). 85 Here, a molecular CO 2 -to-CO conversion photocatalyst (Re I (CO) 3 (BPYDC)Cl) was covalently incorporated into a MOF (UiO-67) which was grown around cubic AgNPs. This combination yielded two main advantages. On the one hand, the incorporation of the molecular compound into the rigid MOF structure prevented its dimerization and increased the stability of the photocatalyst 48-fold to 48 h. On the other hand, coating the AgNPs with the MOF spatially confined the catalytic Re centers close to the metal surface. Under visible light illumination, the AgNPs created an intensified near-field that could be absorbed by the incorporated Re photocatalyst. As a result, the CO 2 -to-CO conversion was enhanced 7-fold due to the incorporation of the PNP.
Another interesting system worth mentioning, although not in the liquid phase, was studied by Halas et al. 88 For this, they combined aluminum nanocubes (AlNCs) with a MOF known for a high CO 2 gas uptake. The hybrid system was used to catalyze the reverse water gas shift reaction (CO 2 + H 2 → CO + H 2 O) in the gaseous phase. The hybrid exhibited a ∼4.5-fold increased CO 2 uptake compared to the pristine AlNCs. The CO 2 preconcentration near the surface of the plasmonic material resulted in a ∼3-fold increase of CO production compared to the pristine AlNCs.
By far the highest product formation rate for CO 2 conversion in Table 10 was reported for a hybrid containing a zeolitic imidazolate framework (ZIF), a subclass of MOFs. This subclass of MOFs has promising features for CO 2 reduction and is currently explored for CO 2 conversion using various approaches and materials. 243−245 Two of the main limiting factors of current ZIF materials are sufficient solar light absorption and e − −h + separation. 246 Combining ZIFs with PNPs may be a perfect synergistic fit, as PNPs can bring both of these features to the system. In the reported hybrid, ZIF-67 decorated with AuNPs generated methanol (MeOH) with a rate of 2380 μmol g −1 h −1 and a selectivity of 97%, while also being stable for at least 12 h. 87 It was proposed that hot electrons generated in the AuNP upon illumination could overcome the Schottky barrier formed at the PNP−MOF interface. As a result of the electrons being irreversibly injected into the LUMO of the MOF, the lifetime of the carriers was extended resulting in an electron-enriched MOF surface where the reaction subsequently occurs. Both components of the hybrid were crucial for the catalytic performance. While the MOF provided the catalytic sites, no product was detected without the plasmonic material.
In the majority of the reported hybrids, the PNP loading amount (wt%) was optimized. However, the size of the MOF crystals was not. Guo et al. showed for MIL-101(Cr)/Ag that controlling the MOF size can have a high impact on the catalytic performance of a PNP−MOF hybrid. 241 When reducing the MOF crystal size in the hybrid from 800 to 80 nm, the CO 2 -to-CO conversion rate was increased ∼20-fold. The authors attributed this to the increased fraction of highly catalytic edge and corner sites of the MOF. This shows that all components should be carefully tuned to optimize the catalytic performance.
The examples for both H 2 production as well as CO 2 reduction show that PNP−MOF hybrids have various potential pathways to increase activity and selectivity. Interestingly, the highest formation rates for both H 2 production as well as CO 2 reduction were achieved by NPdecorated hybrids. For the further development of PNP− MOFs, hybrids could be designed that facilitate single, crucial steps in reactions by introducing functional groups to the MOF. 86 This could open energetic bottlenecks of reactions and show new paths for catalysis. Furthermore, both the spatial arrangement of the hybrid components 89 and the size of the MOF crystals 241 could be implemented as common screening parameters. Generally, data concerning the long-term stability of MOF-PNP hybrids is currently not available, and only two hybrids have been reported with a stability of at least 24 h (see Table 10). This shows that long-term stability is an issue to be addressed for potential industrial applications.

■ PLASMON-ASSISTED ELECTROCATALYSIS
Despite the great progress in photocatalysis, the solar-to-fuel conversion efficiency is still low. Compared to plasmonic photocatalysis, electrocatalysis shows a much higher hydrogen generation and CO 2 reduction activity. However, the high overpotential and low current density are the current bottlenecks hindering energy efficient industrial applications. Consequently, combining plasmonics with electrocatalysis is a promising strategy to improve the energy conversion efficiency.
Plasmon-Assisted Electrochemistry for H 2 Production. All different types of hybrid systems that were discussed above for photocatalytic H 2 generation may also be useful for plasmon-enhanced electrocatalytic applications. Therefore, the broad spectrum of materials investigated is summarized in Table 11. It shows plasmonic metal and semiconductor, 2D and MOF hybrid systems. In this table, the observables Overpotential and Tafel slope are given both under illumination and in dark (the values for dark conditions are given in brackets). The comparison of values in dark and under illumination allows for an evaluation of the enhancement that plasmonics provides to the hybrid system. This evaluation can only be qualitative as the changing observables may not exclusively be attributed to the plasmonic enhancement. 248 Both in the table and in the discussion below, all overpotentials are reported at a current density of 10 mA cm −2 .
A study conducted by Wei et al. shows the effect different configurations of the same metal−metal hybrids can have on the performance. 247 They probed the plasmon-enhanced electrocatalytic activity of Au nanostars modified with AgPt alloys in three different arrangements: AgPt-covered, -tipped, and -edged Au nanostars (see Figure 7c). The Au nanostars offer a wide range of tunability. By tuning the size of both the nanostar core and the nanostar branches, different resonant wavelengths can be achieved. Additionally, the position of the AgPt NPs allows further tuning, as, for example, the electromagnetic field has its highest enhancement in so-called "hot spots" at the nanostar tips. 269 All three systems showed reduced overpotentials and Tafel slopes under visible−NIR irradiation. However, when comparing the configurations, the AgPt-tipped Au nanostars showed a superior performance with an overpotential of 58 mV RHE . This was 17 mV lower than for the AgPt-edged and 33 mV lower than for the AgPt-covered Au nanostars (see Figure 7d). Together with the reduced Tafel slope of the AgPt-tipped Au nanostars (35 mV/dec) this implied a more efficient charge transfer and faster reaction kinetics of this modified nanostar type. The superior activity was reported to be a result of the best synergy between the Au nanostars acting as light absorbers, thereby generating heat and exciting electrons, and the AgPt alloys as electron acceptors that provided the reactive sites. Another interesting example highlights the synergistic effect that can arise between PNPs and conventional semiconductor electrocatalysts for solar-light-enhanced water-splitting. Du et al. investigated a catalyst consisting of AuNPs and NiCo layered double hydroxides (Au@NiCo LDH) toward both reactions involved in water-splitting, the HER and the oxygen evolution reaction (OER). 258 To reach a current density of 10 mA cm −2 , the Au@NiCo LDH required an overpotential of 119 mV for the HER and 203 mV for the OER under illumination. The comparison with the overpotentials needed in dark (160 mV for HER and 234 mV for OER) suggested that the plasmonic excitation of Au led to a higher charge transport efficiency. This shows the potential of incorporating plasmonic materials into conventional electrochemical systems to significantly reduce the electrical energy consumption.
Seven out of 11 metal−2D hybrids in Table 11 contain MoS 2 . While multiple configurations of this 2D material with different plasmonic metals have been explored, 260,262,263,265,297 to our knowledge, one of the first metal−MoS 2 hybrids for plasmon-enhanced electrocatalytic H 2 generation was reported by Shi et al. in 2015. 263 It consisted of Au nanorods (AuNRs) and MoS 2 nanosheets. When illuminated with a 808 nm laser, matching the plasmon resonance of the AuNRs, the onset potential decreased by 60 mV compared to the dark condition. Under the same laser illumination, the activation energy of H 2 evolution on the AuNR-MoS 2 was calculated to be 75.65 kJ mol −1 , much lower than that of MoS 2 alone (94.82 kJ mol −1 ). Moreover, the plasmon excitation resulted in the Tafel slope of AuNR-MoS 2 decreasing by 15 mV/dec. This indicates a higher hydrogen coverage on the surface of the catalysts and shows that plasmon excitation can enhance the electrochemical hydrogen evolution not only thermodynamically but also kinetically.
To our knowledge, so far only two PNP−MOF hybrid systems 268,298 have been investigated toward plasmon-assisted electrochemical H 2 generation. Currently, PNP−MOF systems seem to be significantly weaker in performance than the majority of the metal, semiconductor, and 2D hybrids, with the lowest overpotential reported thus far being 292 mV. In comparison, the best PNP−metal, 215 PNP−semiconductor, 257 and PNP−2D 250 hybrids operated at overpotentials as low as 18, 45, and 28 mV. To conclude, plasmon excitation can improve the electrocatalytic hydrogen evolution activity both thermodynamically and kinetically. With Au being the plasmonic component in 20 out of the 23 listed configurations, it is still the most utilized plasmonic material in all plasmonic electrocatalysts for H 2 generation reported in Table 11. One emerging non-noble plasmonic material is titanium nitride (TiN) which has shown superior performance for photoelectrochemical water-splitting. 60 The remarkable hot carrier collection of TiN@TiO 2 was attributed to its broadband absorption as well as the forming of an ohmic junction for efficient electron collection. 60 Plasmon-Assisted Electrochemistry for CO 2 Reduction. Similar to electrochemical H 2 production, the electrochemical reduction of carbon dioxide typically suffers from low reaction rates that necessitate high overpotentials. Additionally,  the low selectivity, especially toward multicarbon products, impedes the electrochemical production of carbon-based fuels. Recent efforts to overcome these limitations by plasmonassisted electrochemical (CO 2 ) reduction are summarized in Table 12. Figure 7a shows the schematic setup and mechanism for plasmon-assisted electrocatalytic CO 2 reduction. A study conducted by Kim et al. showed that plasmonic excitation can be used to increase the reaction rate of CO 2 reduction. 271 In this study, the electrodes consisted of Ag nanopyramids patterned on a Ag surface. The dark current density was −0.8 mA cm −2 under −1.1 V RHE . Under illumination the current density increased by 31%, up to −1.0 mA cm −2 . The observed photocurrent was attributed to a resonant transfer of photogenerated plasmonic hot electrons to the lowest unoccupied molecular orbital (LUMO) acceptor energy levels of adsorbed CO 2 molecules or reductive intermediates. This shows the potential of plasmon-assisted electrochemistry to overcome kinetic barriers. The ability of plasmonic electrocatalytic systems to modify chemical reaction landscapes was also studied by Creel et al. on plasmonically active Ag thin film electrodes. 94 They found that under illumination, the FE for CO increased to up to 95%, while in pure electrocatalysis it was only 17%. At the same time, the production of H 2 was suppressed (from FE of 75% in dark to 10% under illumination), as shown in the left panel of Figure 7b. 94 Moreover, methanol production was observed only under illumination (right panel of Figure 7b). The FE reached up to 2% (at −1.1 V RHE ), which is reportedly an unprecedented value for silver cathodes. This shows that plasmonics provides a promising approach to promote complex electrochemical CO 2 RRs over other competing reactions.
Apart from the metallic electrodes discussed above, the application of metal−semiconductor hybrid materials in plasmon-assisted electrochemistry can potentially further enhance CO 2 reduction. This was probed by Landaeta et al. with a plasmonic−metal semiconductor system, consisting of Ag and Cu 2 O/CuO (Ag/Cu x O). 91 The illuminated hybrid system allowed the production of acetate with a FE of 54% at a potential of −0.4 V Ag/AgCl . The formation of acetate at such a low potential had not been reported before. Moreover, The current density reached 0.55 mA cm −2 upon illumination, which was 3.4 times as high as the value in dark. The enhanced performance of the hybrid originated from synergistic effects between the metal and the semiconductoras no photo-electrocatalytic activity was observed on either just the Ag nor the Cu x O alone at such low potentials. The synergistic effect between Ag and Cu x O was attributed to reduced charge recombination processes thereby decreasing the needed overpotential. Additionally, Ag stabilized the Cu 2 O/CuO within the applied potential range, further improving the catalytic performance.
Cu is widely studied for electrochemical CO 2 reduction because of its ability to catalyze the electrochemical conversion of CO 2 to valuable fuels and chemicals beyond CO. 274 To improve selectivity, Corson et al. coated Cu nanostructures with silver to create a plasmonically active cathode. 270 Illumination allowed the selective enhancement of 5 of the 14 typically observed copper CO 2 reduction products while simultaneously suppressing hydrogen evolution. 270 Their Faradaic efficiencies can be found in Table 12. Under illumination, the CO production was enhanced at low overpotentials, whereas at high overpotentials the production of ethylene, methane, formate, and allyl alcohol was increased.
In conclusion, introducing plasmonics into electrocatalytic systems is a promising method to improve the activity of both HER and CO 2 RR and increase the selectivity of CO 2 RR. These properties allow the utilization of light to decrease the energy consumption in producing renewable and clean fuels. However, most plasmonic metals investigated in plasmonassisted electrocatalytic systems are noble metals, which are costly and therefore hinder industrial applications. Additionally, in CO 2 RR the Faradaic efficiencies of hydrocarbons and multicarbon products are still low, which need to be improved in the future.

■ DISCUSSION
In this section, we will summarize and compare the performance of the hybrid plasmonic photocatalysts reviewed so far. We will organize the comparison according to the following aspects. First, we make a general review on the most frequently utilized plasmonic components and light sources in plasmonic catalysis. Second, we quantitatively compare photocatalysts according to their activity and selectivity, breaking down the different reactions of interest for hydrogen production and CO 2 reduction. Third, we compare the performance of the different plasmonic electrocatalysts for HER and for CO 2 reduction.
In the vast majority of the systems explored until now, the plasmonic component was either Au or Ag. Figure 8a shows a histogram of the main plasmonic component among the hybrid systems in this Review: 68% use Au, 24% use Ag, and only 8% use a different material. Reasons for this preference may include the high photostability of Au and the availability of versatile protocols to shape Au into different nanomorphologies with tailored resonances along the visible spectrum. Earth abundant plasmonic materials have been less explored.
Basing hybrid plasmonic catalysts on expensive, rare metals is a bottleneck in the development of sustainable large-scale photocatalytic methods. New abundant plasmonic materials like Al, 160,275,276 Cu, 35,62,113,186,199 Mg, 277 metal oxides, 278−280 metal nitrides 281 or even 2D materials, 93 are emerging candidates in plasmonic catalysis. However, they have been relatively unexplored until now. Moreover, hybrid combinations of these materials are a very interesting and emerging field to explore further. Research in these directions and with these materials could merge plasmonic catalysis with sustainable and large-scale solar-to-chemical energy conversion routes, disclosing the real potential for the industrial application of these concepts.
Another key parameter in (electro)photocatalysis is the light source. Figure 8b shows the histogram of illumination sources among the publications reviewed here. The most used is the xenon lamp, because its emission spectrum resembles the solar one in the wavelength range 290−700 nm. They account for 68% of the examples, followed by monochromatic laser light with 9%. Lasers are chosen to match the plasmonic absorption peak of the used photocatalyst and are a useful tool to study the wavelength dependence of their properties. In addition, they can provide higher irradiances. The third most frequent illumination source is light-emitting diodes (LEDs), 8%. 365 nm LEDs are particularly popular for electrochemical CO 2 reduction, as they can induce the generation of allyl alcohol on Ag/Cu hybrids. Only 7% use AM 1.5 G sources, which are the standard for solar simulators.
Quantitative comparison among different photocatalysts is challenging due to the existent discrepancies in reported metrics and the high number of relevant experimental parameters involved, usually not reported. 104,105 We believe that the field can benefit from standardization of reported activity metrics and measurement conditions. The adopted metric should incorporate information on the amount of catalyst and the intensity of the employed light. For this reason, we propose the implementation of the following activity metric: where n product is the amount of generated product, t is the time interval, m catalyst is this mass of illuminated catalyst, and I is the Basing hybrid plasmonic catalysts on expensive, rare metals is a bottleneck in the development of sustainable large-scale photocatalytic methods. incident light irradiance. This quantity will only be meaningful if measured far from saturation, i.e., with a linear dependence on time and catalyst concentration, in a large excess of reactants. Plasmonic systems have resonances, and therefore their activities are expected to be wavelength dependent. For this reason, the spectral composition of the illumination source must always be reported. Standardization of the employed light sources would allow a proper comparison among systems. The reasonable choice for the field of artificial photosynthesis would be AM 1.5 global light sources. Additionally, to investigate the role of the spectral composition of light on the efficiency, measuring single-wavelength yields using monochromatic irradiation may be useful. It is important to compare plasmonic catalysts operating at similar bulk temperatures of the photoreactor. However, less than half (47%) of the publications reviewed here have reported a controlled reactor temperature. In addition, variations in the geometry of the photoreactors could also lead to discrepancies in the reported performances, even with the same employed catalyst. 282 To overcome this issue, the field could benefit from the establishment of a standard photocatalyst, to be used as an internal reference at each laboratory, in a similar fashion as it was proposed for semiconducting photocatalytic materials. 283 In many of the examples reviewed here, the information on reaction conditions is incomplete. In addition, even when reported, conditions vary among the different examples. Nevertheless, we believe that insightful conclusions can be obtained from the comparison of formation rates. Although the comparison between two photocatalysts is not exact under different conditions, we will focus here on evaluations of orders of magnitude in the produced formation rates. While doing so, we aim at identifying the best performers and the ones with the highest potential among the multiple hybrid plasmonic catalysts demonstrated in the past years. Figure 9 shows the formation rates extracted from all tables in this Review for photocatalytic H 2 generation ( Figure 9a) and CO 2 reduction (Figure 9b). Exploration of plasmonic systems for H 2 has been the most extensive, accounting for 71% of the publications reviewed here. Among the systems used for H 2 production, the most popular were plasmonic metal−2D catalysts (38%), followed by metal−electrochemistry (25%), metal−semiconductor (18%), metal−perovskite (7%), metal−metal (7%), and metal−MOF (5%). For CO 2 reduction, the most popular were also plasmonic metal−2D catalysts (25%), followed by metal−electrochemistry (20%), metal−semiconductor (18%), metal−metal (17%), metal− MOF (13%), and metal−perovskite (7%). The range of formation rates reported and the most efficient systems depend on the reaction under study. We will now break down the analysis into the different reactions of interest.
Hybrid Plasmonic Nanomaterials for Photocatalytic Hydrogen Production. The performance of different plasmonic catalysts will be evaluated for the following three hydrogen production reactions: ammonia borane dehydrogenation, formic acid dehydrogenation, and water-splitting.
The hydrolytic dehydrogenation of ammonia borane was found to be the hydrogen generation reaction with the highest formation rate. An outstandingly high formation rate of 10 8 μmol g −1 h −1 at 40°C was observed employing core−shell Ag@Pd nanostructures. 109 To put this number into context, one of the highest reported performances of heterogeneous catalytic systems for hydrolysis of ammonia borane at room temperature is 69.8 mol/min using porous Pt 25 Pd 75 NPs 284 This translates to 3 × 10 7 μmol g −1 h −1 if a molar ratio of 25% Pt and 75% Pd is assumed. Therefore, the core−shell Ag@Pd nanostructures offer a comparable performance with just a 1.3 nm layer of Pd and a relatively mild illumination source of 50 mW cm −2 (on the order of half the average solar irradiance).
The hydrogen generation from formic acid dehydrogenation was possible with formation rates of the order between 10 2 and 10 6 μmol g −1 h −1 , as shown in Figure 9a. The metal−metal AgPd catalyst loaded on 2D graphitic carbon nitride nanosheets achieved the highest formation rate (2 × 10 6 μmol g −1 h −1 ). 187 Furthermore, AuNPs interfaced with a metal (Pd) or semiconductor (CdS) were also capable of photocatalyzing formic acid dehydrogenation with formation rates in the order of 10 5 μmol g −1 h −1 . 114,132 However, metal− semiconductor systems were able to reach these formation rates at room temperatures, while the highest activities in metal−metal systems required temperatures between 40 and 50°C. In addition, semiconductors are typically cheaper and more abundant than catalytic metals, indicating that metal− semiconductors systems could be preferable for this reaction over metal−metal systems. Overall, the best activities for photocatalytic hydrogen generation from formic acid with plasmonic systems are comparable to the activity of heterogeneous catalysis at room temperature and without light, where formation rates on the order of 10 6 μmol g −1 h −1 were also achieved, e.g., using 2.2 nm Ag 42 Pd 58 NPs. 285 Reviewing H 2 generation from water-splitting reactions, we observed formation rates of the order between 10 0 and 10 5 μmol g −1 h −1 . The highest activity was 3.5 × 10 5 μmol g −1 h −1 , achieved using plasmonic PtNPs incorporated in nanostructured TiO 2 , 130 closely followed by a dual metal− semiconductor system including both Au@CdS and Ag@ SiO 2 and achieving 2 × 10 5 μmol g −1 h −1 at a temperature of 15°C. 132 In addition, using metal−metal ZnCu plasmonic alloys, formation rates on the order of 10 5 μmol g −1 h −1 were also achieved, illustrating an interesting demonstration that does not utilize Au or Ag. 113 . Another example that does not use Au or Ag as the plasmonic component are CuNPs interfaced with WS 2 nanosheets that produced 10 4 μmol g −1 h −1 of hydrogen under 1 solar irradiation. 186 Finally, it has been shown that it is possible to drive water-splitting reactions using perovskites. However, up to date, most examples still show limited efficiencies in the order of 10 3 μmol g −1 h −1 . Remarkably, the highest plasmonic activities for water-splitting are 1 order of magnitude larger than the best activities achieved by other non-plasmonic graphitic carbon nitride-based photocatalysts. 286,287 This demonstrates that plasmonic nanomaterials have a substantial potential for water-splitting reactions.
Hybrid Plasmonic Nanomaterials for Photocatalytic CO 2 Reduction. As shown in Figure 9b, the formation rates for CO 2 reduction are in general lower than for hydrogen production, in the range of 10 −1 −10 3 μmol g −1 h −1 . Moreover, we can observe that the main product of many catalysts for CO 2 reduction is CO (56%), followed by CH 4 (32%). Figure  9c shows formation rate of each reactions main product versus its selectivity. The plot has less points than Figure 9b because selectivity is less commonly reported. Interestingly, in the cases where it is reported, it is higher than 50%. It must be noted that there is a lack of examples where products with more than one carbon (such as C 2 H 6 , C 2 H 4 , and C 3 H 6 ) are produced with high selectivity. However, these chemicals have been obtained as minor products in numerous publications. Figure  9d shows the reported minor products for CO 2 reduction, showing that C−C coupling reactions are indeed possible to achieve with plasmonic catalysts. This is a very interesting but poorly explored route so far.
To date, the highest reported performance for producing CO was obtained using AgNPs deposited onto a MOF. Specifically, MIL−101(Cr)/Ag showed a CO production of 8 × 10 2 μmol g −1 h −1 at 50°C, with a selectivity of 61.3%. 241 The hybrids with the next best production rates are plasmonic−2D Ag/rGO achieving 10 2 μmol g −1 h −1 at 45°C , with a high selectivity close to 100%. 184 Furthermore, plasmonic-semiconductor systems formed by Au/CdSe-Cu 2 O obtained a CO formation rate of 80.2 μmol g −1 h −1 . 123 It should be noted that to date, there have been no reports of bimetallic PNPs for this reaction. The highest activity obtained with plasmonic hybrids is comparable to what can be achieved by a non-plasmonic g-C 3 N 4 nanosheet with B−H bonding decorated MOF (∼10 3 μmol g −1 h −1 ). 288 The second main product obtained from the reduction of CO 2 was CH 4 . This reaction requires multiple electron transfers, and therefore CH 4 is more difficult to produce than CO. The best catalysts for producing CH 4 were metal− 2D materials. All examples exhibited formation rates on the order of 10 2 . A system consisting of AgNPs combined with Ndoped TiO 2 nanosheets achieved a formation rate of 79 μmol g −1 h −1 with almost 100% selectivity. Another plasmonic-2D material Au/MoS 2 , also showed CH 4 production rate on the same order of magnitude, achieving 20 μmol g −1 h −1 with a selectivity of 98%. From the metal-semiconductor category there was one example achieving that order of magnitude with 13 μmol g −1 h −1 . It must be noted that CH 4 production from CO 2 was also demonstrated with metal and metal−MOF systems but with lower activities. The formation rates for CH 4 production with plasmonic systems are still lower than the ones recently achieved by non-plasmonic photocatalytic systems such as g-C 3 N 4 nanosheet with B−H bonding decorated MOF (1.5 × 10 4 μmol g −1 h −1 ) or titanium carbide compounds (786 μmol g −1 h −1 ). 288,289 Finally, there are two examples worth mentioning where methanol and formic acid were obtained as main products with close to 100% selectivity using plasmonic-MOF systems. First, AuNPs incorporated in a zeolitic imidazolate framework (ZIF-67) produced 2 × 10 3 μmol g −1 h −1 of methanol under irradiation from a solar simulator. 87 Second, hybrids of porphyrin frameworks loaded with AuNPs produced 40 μmol g −1 h −1 of formic acid at room temperature. 240 These two examples show the great potential of plasmonic−MOF systems for selective reduction of CO 2 into solar fuels.
Hybrid Plasmonic Nanomaterials for Photoelectrocatalytic H 2 O Splitting and CO 2 Reduction. Illuminated plasmonic nanomaterials can successfully decrease the overpotentials required for electrocatalytic hydrogen evolution. Figure 10a summarizes the required overpotentials of all the plasmonic system reviewed here, at a current density of 10 mA cm −2 . Notably, there are several examples of metal−metal hybrids requiring overpotentials lower than 100 mV. The best performer plasmonic system was metal−metal Pt/Fe@Au nanorods that were able to reduce the overpotential to 18 mV under irradiation. 215 In addition, this system showed a Tafel slope of 29.3 mV/dec, which is close to the theoretical minimum of the Tafel slope in hydrogen evolution from aqueous solution. 290 Plasmonic metals in combination with semiconductors and 2D materials were also capable of delivering low overpotentials. Upon illumination, the overpotential was 45 mV for nanoperforated Au films on a Ti/TiO 2 film, due to the facilitated electron transfer at the Au and TiO 2 Schottky barrier. 257 By employing Au nanorods on a graphene/MoS 2 substrate it was possible to lower the required overpotential to 28 mV. 259 On the other hand, MOFs systems presented larger overpotentials above 292 mV. Although MOFs with large specific surface area could provide more catalytic sites for plasmonic electrocatalysts, their usually high resistivity retards the hydrogen evolution activity.
One way to evaluate the role of light in photoelectrocatalysis is to measure the Tafel slope decrease under illumination. Figure 10b shows the Tafel slope reduction of every system in this Review. Decreases are most significant in metal−2D hybrids, highlighting the beneficial contribution of plasmonics to promote photoexcited charge separation in these materials.
Overall, it can be concluded that metal−metal and metal− 2D hybrid nanostructures have shown the greatest photo- The main product of many catalysts for CO 2 reduction is CO, followed by CH 4 . Figure 10. (a) Overpotential for plasmon-assisted electrocatalytic HER for all hybrid systems reviewed here. (b) Tafel slope decrease upon illumination of hybrid plasmonic electrocatalysts. electrocatalytic performances and are the most promising materials for catalyzing the HER.
Plasmonic systems are also beneficial to enhance CO 2 electroreduction. However, up to date, we were able to find only eight published examples. Among these, some remarkable performances can be highlighted. For example, by using a metal−metal Ag/Cu hybrid plasmonic system, it was possible to enhance the CO partial current density by 200% when illuminated. 270 In another example, using a Cu/p-NiO metal− semiconductor hybrid, the partial current density values of CO and HCOOH increased by 260% and 100% under illumination. 62 Regarding the selectivity of CO 2 reduction, most reviewed plasmonic electrocatalysts displayed an improved FE toward specific products. For example, the FE of carbon monoxide and formate on Cu/p-NiO with illumination increased by 217% and 267%, respectively. 62 Another interesting example of improved FE was the achieved 95% CO FE of Ag/Ti film under 365 nm LED illumination at −0.6 V RHE . 94 This performance is better than that of one of the top pure electrocatalysts: nanoporous Ag achieved 92% CO FE at −0.6 V. 291 These results indicate that metal−semiconductor hybrids are promising systems to improve CO 2 electroreduction activity and selectivity by plasmon excitation.
In spite of the advances made in the design of hybrid plasmonic nanomaterials for photoelectrocatalytic CO 2 reduction and hydrogen evolution, the role of the plasmon is still unclear. In pure electrocatalytic CO 2 reduction, the interface between metal and oxide semiconductor is usually considered as the active catalytic center as it facilitates the stabilization of reaction intermediates. 292 Plasmon excitation is expected to be beneficial for CO 2 reduction at the interface, because the photothermal effect and electromagnetic field enhancement could facilitate the CO 2 activation by lowering the reaction energy barrier. 293 However, the exact role of plasmon excitation on the stabilization of CO 2 intermediates is still elusive. More detailed investigation should be focused on unraveling the origin of the plasmonic enhancement. For example, electrochemically in situ Raman 190 and Fourier transform infrared spectroscopy (FTIR) 270 can be employed to unveil the plasmon-enhanced photoelectrocatalytic reaction mechanisms. These methods employed with and without light irradiation could provide some insight into how the plasmon excitation changes the reaction pathway of electrocatalytic processes on plasmonic electrodes. Additionally, a significant enhancement of photocurrent is crucial for the improvement of plasmon-assisted photoelectrocatalytic performance. However, the short lifetime of hot carriers in plasmonic nanomaterials results in a low incident photon-to-electron conversion efficiency (IPCE), which limits the activity enhancement in plasmon-assisted photoelectrocatalytic hydrogen evolution and CO 2 reduction. To this end, the interface between components in hybrid plasmonic nanostructures that can enhance the hot electron−hole separation efficiency is a key factor to achieve a high plasmonic enhancement. 13,115,117 Therefore, it is pivotal to optimize the interface between plasmonic metals and hybrid components to prolong the lifetime of hot electrons generated from the plasmon dephasing. Consequently, more efforts should be made to explore the materials with high electron mobility or long electron diffusion length as the candidate to couple with plasmonic metals.
Outlook. We have critically reviewed and benchmarked a large set of recently reported hybrid plasmonic materials for solar fuel production. Remarkably, the hydrogen formation rates achieved by plasmonic catalysts are comparable to state of the art technologies, with the advantage of generally employing milder reaction conditions or less expensive materials, such as metal−semiconductor or metal−2D hybrids. However, most of these examples were performed with irradiances larger than expected for sunlight-driven catalysis. To reach or surpass these formation values with irradiances comparable to sunlight, further development of sunlight concentrators or focusing devices is still needed as well as novel reactor designs that can maximize light penetration and absorption. Scalability of these technologies seems possible but is scarcely explored so far. On the other hand, carbon dioxide reduction with plasmonic catalysts has not yet shown such appealing results as hydrogen generation. In general, the selectivity and efficiency for this reaction when employing hybrid plasmonic materials are still far from those of the excellent electrocatalysts reported in the literature. In this sense, the combination of plasmonic materials with metal organic frameworks appears as a promising route to explore, with few but remarkable results so far. Additionally, the combination of more than two types of materials into higher order plasmonic hybrid systems may not only allow the exploitation of the properties of the respective two-component hybrid systems but also introduce new synergies. 85,180 Finally, even though we have only focused on solar fuels, plasmonic hybrids might present interesting opportunities for the production of highly valuable chemicals that in most cases require the use of expensive precious metal (electro)catalysts. 294 In this context, earth-abundant plasmonic materials have been rarely explored for catalysis; Al, TiN, some metal oxides, among others are materials that deserve further attention. Interestingly, plasmonic materials as support of single catalytic atoms could also become a promising route for the future development of the field. 295 From the analysis of the systems reported here, the combination of plasmonic materials with also earth-abundant hybrid materials, such as nitrides (TiN, g-C 3 N 4 ) and MOFs, may lead to new routes toward efficient and selective sustainable catalysis. We thank Vera Hiendl for support in graphics design.