Single nanoparticle photoelectrochemistry: What is next?

The field of semiconductor photoelectrochemistry dates back to 1839 when French physicist Edmond Becquerel observed a photovoltaic effect at an illuminated silver chloride electrode.¹ Over a century later, Brattain and Garrett showed how the rates of chemical reactions under dark and light conditions were linked to the electronic properties of n- and p-type germanium electrodes.² In the early period of the 1950s and 1960s, research was focused on fundamental aspects of the semiconductor/electrolyte interface, leading to concepts such as the flatband potential^3,4 and surface states.^5–7 At the same time, the mechanisms of several important concepts were also discovered such as the current doubling effect,^8–10 photovoltaic effects at the semiconductor/electrolyte interface,^2,11 multistep reaction mechanisms,^12–14 and photocorrosion.^15–17 Through the 1960s and early 1970s, Gerischer developed a model to describe heterogeneous electron transfer reactions that included the electronic structure of the semiconductor.^18–20 At about the same time as the principles of the field were being firmly established, the 1973 oil crisis inspired an intense research effort toward practical applications of solar energy conversion to electricity or chemical fuels.^21–24 In the 1970s and 1980s, photoelectrochemical cells with >11% photovoltaic efficiency and 12% solar-to-hydrogen fuel efficiency were developed.^25–28 The cells employed expensive single crystal materials with well-defined electronic properties.

In the mid-1970s, Gerischer and Heller discussed the possibility of using inexpensive, polycrystalline semiconductors for photoelectrochemical energy conversion applications.²² The authors discussed how the liquid electrolyte provided an easy and rapid rectifying junction to separate charge carriers, which is not easily achieved in a solid-state p-n junction device. Indeed, several studies showed that photoelectrochemical cells made from polycrystalline materials approached the energy conversion efficiency of single crystalline materials (e.g., n-CdS photoelectrochemical solar cells).^29–34 These impressive results spurred interest in constructing efficient photoelectrochemical energy conversion systems based on inexpensive materials instead of electronics-grade semiconductors.

The invention of the dye sensitized solar cell (DSSC) in 1991 ignited a nanotechnology revolution within the field of photoelectrochemistry.³⁵ The discovery of the DSSC signaled that inexpensive, solution-processable nanoparticle electrodes could efficiently transport electrons and produce large photocurrents, despite the enormous number of particle-particle interfaces in the nanoparticle film. The DSSC inspired researchers in the field to develop nanostructured photoelectrochemical cell designs for a wide variety of reactions; Maldonado and co-workers tabulated an exhaustive list of cell efficiencies.³⁶ One common theme in the tabulated data is that nanostructured electrodes with the same composition but slightly different morphology (e.g., nanowire length or width) exhibit large variations in cell performance. These findings motivate fundamental studies that focus on understanding how the chemical, electronic, and physical properties of nanoscale materials influence their photoelectrochemical properties. In this Perspective article, we review the basic principles of bulk semiconductor photoelectrochemical cells, point out the differences between nanostructured and bulk electrodes, and discuss how single nanoparticle photoelectrochemical measurements can reveal the underlying causes of performance heterogeneity in nanostructured electrodes.

Here, we review the basic principles of semiconductor (photo)electrochemistry that were established for bulk semiconductors. This section introduces the language of semiconductor electrochemistry (e.g., flatband potential, depletion region, Gärtner-Butler model) and describes how and why the photocurrent changes with an applied potential. Rigorous treatments of the physical chemistry of semiconductor/electrolyte interfaces under dark and light conditions can be found in the literature.^18,37–48

First, we consider the key principle in the field of semiconductor electrochemistry: a strong electric field forms in the space charge region at the semiconductor/redox electrolyte interface. Figure 1(a), left, shows an n-type semiconductor (e.g., TiO₂, BiVO₄, Fe₂O₃) that is not yet in contact with a redox electrolyte. In this scenario, the Fermi level of the semiconductor (E_F) is more negative on the electrochemical scale than the redox potential of the solution (E^0/+). Figure 1(a), right, shows the n-type semiconductor/electrolyte interface after contact, where E_F has moved to the same energy level position as E^0/+. The solid phase reaches equilibrium with the liquid phase through a charge equilibration process. Electrons move from the solid phase to the liquid phase until the Fermi level of the semiconductor is equal to the solution potential. The positively charged region in the solid and the negatively charged region at the solid/liquid interface are referred to as a space charge region. If we assume that all the electrons transfer from dopant atoms in the solid, then the charge equilibration process leaves behind a region of fixed positively charged dopant atoms at the semiconductor/electrolyte. The dopants at the semiconductor/electrolyte interface are depleted first [i.e., x = 0 in Fig. 1(b)], followed by dopants in the semiconductor interior (x > 0). Since we have assumed that all dopant atoms in the positively charged region of the solid have been completely depleted of electrons and the region has an abrupt boundary [purple region in Fig. 1(a)], we refer to this region as the depletion region and the key assumptions make up the depletion approximation. We note that the discussion above also applies to p-type semiconductor/electrolyte interfaces.

Poisson’s equation relates the charge density to the potential in the depletion region, which leads to a quadratic dependence of the potential energy of electrons with distance from the semiconductor/electrolyte interface [i.e., the band bending effect in Fig. 1(b)].⁴⁹ The total potential drop in the band-bending or depletion region is V_bi [see Fig. 1(a), right], which is equal to (E_F − E^0/+). The thickness of the depletion region depends on the applied potential (E) and is given by $W=2εε0(E−Efb)/qNd$ (in units of cm), where ε is the relative dielectric constant of the semiconductor, ε₀ is the vacuum permittivity (in units of C V^–1 cm^–1), E is the applied potential relative to the reference electrode (V), E_fb is the flatband potential relative to the reference electrode (V), and N_d is the doping density of the semiconductor (cm^–3). E_fb is the applied potential that “flattens” the bands or returns the semiconductor/electrolyte interface to the noncontact condition shown in Fig. 1(a), left.⁵⁰ The maximum electric field strength at the solid/liquid interface⁵¹ is given by $Emax=WqNdεε0$⁠. Thus, the magnitude and thickness of the electric field are determined by the semiconductor and redox electrolyte properties (e.g., ε, N_d, E_F, and E^0/+). In summary, a strong electric field spontaneously forms at the semiconductor/electrolyte interface. The field rapidly separates and transports carriers to the solid/liquid interface where they participate in redox reactions with species in solution.

The underlying processes that contribute to the rate (i.e., photocurrent) of photoelectrochemical reactions are (1) carrier generation in the semiconductor interior via absorption of light, (2) charge separation and transport to the solid/liquid interface, and (3) interfacial charge transfer reactions between the semiconductor and redox species in the electrolyte. A general expression for the photocurrent density (J_PEC) of a photoelectrochemical cell is given by J_PEC = J_abs × ϕ_sep × ϕ_int,^49,52–55 where J_abs is the absorbed photon flux (s^–1 cm^–2), ϕ_sep is the charge separation and transport efficiency within the semiconductor (i.e., J_abs × ϕ_sep is the fraction of photogenerated holes that reach the n-type semiconductor/electrolyte interface), and ϕ_int is the interfacial charge transfer efficiency. The Gärtner-Butler model^56,57 quantitatively describes J_PEC vs E by assuming that (1) all changes in E are entirely manifested as a potential drop within the depletion region of the semiconductor and (2) the carrier generation rate as a function of position into the semiconductor follows the Beer-Lambert law [purple line in Fig. 1(b)]. The model considers two contributions to J_PEC: (1) a drift current due to charge carriers generated in the depletion region [W in Fig. 1(b)] and (2) a diffusion current due to charge carriers generated up to a distance, L, from the edge of the depletion region, where L is the minority carrier diffusion length in units of cm in Fig. 1(b). L is the hole diffusion length for n-type semiconductors. In other words, the Gärtner-Butler model assumes that ϕ_sep = 1 and ϕ_int = 1 for all photogenerated carriers within W and L. The assumptions are rationalized by the rapid charge transport time across the depletion region. For example, a TiO₂ electrode with ε = 100,⁵⁸ N_d = 10¹⁷ cm^–3, and E − E_fb = 1 V exhibits a W = 330 nm and $Emax$ = 6 × 10⁴ V/cm. Given that the hole mobility in TiO₂ is 1 cm²/V s,⁵⁹ a photogenerated hole will traverse across the depletion region, driven by a strong electric field, in about 0.5 ns.

The above assumptions yield the Gärtner-Butler equation $JPEC=−qϕ01−exp(−αW)1+αL$⁠, where ϕ₀ is the incident light power density (s^–1 cm^–2) and α is the monochromatic linear absorption coefficient (cm^–1). For an n-type semiconductor, J_PEC increases with increasing anodic potentials because W increases quadratically with E. The Gärtner-Butler model has been used to describe the J-E behavior of photoelectrochemical cells,^57,60,61 especially in the large band bending region (e.g., E − E_fb > 0.5 V). The model often fails when E approaches E_fb because the electric field strength decreases and the assumption that ϕ_sep = 1 may not be valid.^59,62–66 Nonetheless, the Gärtner-Butler model provides insight into photoelectrochemical cell design: tune the electronic properties of the semiconductor to match W and L with the light penetration depth (1/α). The electronic properties of the semiconductor are defined by the doping concentration in the solid, which must be controlled at the part per million-level or less. Fine control over the crystal purity often requires expensive or energy-intensive solid-state synthesis methods. Indeed, the highest efficiency photoelectrochemical cells are based on high quality semiconducting materials (e.g., GaAs, GaP, InP, WSe₂, and MoSe₂);^22,67,68 the highest efficiency cells for solar hydrogen production are made from multijunction GaInP/GaAs^69,70 or GaInAs/GaInP⁷¹ absorbers. However, the high cost of these materials has motivated researchers to explore inexpensive nanomaterials as light absorbers for practical photoelectrochemical energy conversion applications.

Nanoscale materials are attractive candidates for large area, low-cost photoelectrodes. For example, inexpensive, solution-processed nanoparticle inks could be deposited on flexible substrates via high throughput deposition processes such as roll-to-roll manufacturing.⁷² This “solar paint” concept⁷³ represents an exciting and massively scalable strategy to mount photoelectrochemical cells on building walls and roofs. Several articles reviewed the materials chemistry challenges associated with the development of practical photoelectrochemical systems.^46,74–77 Those challenges include the discovery of earth-abundant and photochemically stable materials with the optimal band gap and redox properties to capture solar photons and drive desirable chemical transformations such as the hydrogen evolution reaction (HER)⁷⁸ or carbon dioxide reduction.⁷⁹ Here, we discuss the fundamental differences between nanoscale and bulk semiconductors, potential advantages of and challenges associated with nanostructured electrodes, and how single particle photoelectrochemical measurements have the potential to address important open questions in the field.

Nanoparticle electrodes behave differently than bulk electrodes. One major factor that contributes to the different photoelectrochemical behavior is the space charge region at the nanoparticle/electrolyte interface. In Sec. I B, we discussed the underlying assumptions of the depletion approximation and calculated that the space charge or depletion region thickness is 330 nm for a bulk n-type TiO₂ electrode with N_d = 10¹⁷ cm^–3 and E_F − E^0/+ = 1 V. This length scale is much larger than a typical nanomaterial, which is generally defined as an object with one critical dimension that is less than 100 nm. For example, a 20 nm-diameter TiO₂ particle with a typical N_d value of 10¹⁷ cm^–3 is much smaller than the expected depletion region thicknesses for a planar TiO₂ electrode with the same N_d. When W exceeds the physical dimensions of the nanoparticle, the entire potential drop (e.g., E_F − E^0/+ = 1 V) does not occur within the space charge region of the nanoparticle. Instead, the potential drop occurs across the nanoparticle and the Helmholtz double layer at the solid/liquid interface.^49,80 A major consequence of this effect is that the depletion region thickness of nanoparticles does not increase quadratically with E as it does for bulk materials. Several studies have considered the electric potential profile of 0D spherical particles,^81–83 1D nanowires and nanorods,^36,80 and 2D nanosheets.^84,85 The models typically assume a perfect geometry and a uniform doping density within the structure. However, many nanostructured photoelectrodes have irregular geometries, such as nanowire arrays that are not completely vertically aligned [see Fig. 1(c)] and nanosheet films made from randomly oriented sheets that vary in width and layer thicknesses. The exact potential profile of irregular nanostructured electrodes in the dark and under working photoelectrochemical conditions is not entirely understood. Regardless of the exact potential profile, the nanoparticle dimensions limit the space charge region thickness and total potential drop within the solid.

Since electric fields in fully depleted nanoparticles may not play a major role in the charge separation and transport processes, an alternative strategy to achieve high photocurrent efficiencies using nanostructured electrodes is to minimize the transport distance for charge carriers while also maintaining high light absorption. The key concept of nanostructured photoelectrochemical cells is to design the semiconductor dimensions such that all photogenerated carriers are produced within a diffusion length from the solid/liquid interface. For example, Fig. 1(c) shows a 1D nanowire array photoelectrode for solar fuels production where electrons in these p-type semiconductors are transported to the electrolyte where they participate in, for example, the hydrogen evolution reaction or the CO₂ reduction reaction. In this scheme, charge carrier generation occurs along the micrometer-long nanowire length, while charge carrier transport occurs over the nanometer-wide diameter [Fig. 1(d)]. Vanmaekelbergh et al. first demonstrated that nanostructuring low-grade GaP wafers via photoanodic etching improved the photocurrent efficiency.⁸⁶ Lewis and co-workers demonstrated that the orthogonalization approach shown in Fig. 1(d) improved the photocurrent efficiency of Cd (Se, Te) nanowire arrays compared to planar electrodes.⁸⁷ Another intriguing possibility is to use ultrathin two-dimensional semiconductors such as monolayer MoS₂ or WSe₂ as light harvesting materials [Fig. 1(e)].^88–91 Semiconductor monolayers have the potential to convert solar energy to electricity or fuels with high efficiency because all photogenerated carriers are produced at a charge-collecting interface [Fig. 1(f)]. Since the ultrathin materials only absorb a small fraction of incident light, thin film architectures or plasmonic electrode architectures are still needed to enhance overall light absorption.^92,93

Our understanding of the photoelectrochemical properties of nanostructured electrodes is derived mainly from ensemble-average experimental measurements. The problem with ensemble-average measurements is that they represent the particle-averaged response of individual objects that vary in size and shape. For example, the electric field profile in a small particle likely differs from a large particle, and these differences are averaged in conventional photoelectrochemical current measurements. Ensemble-level studies have clearly demonstrated that nanoscale morphology influences electron transport efficiency and dynamics in nanostructured electrodes.^{81,82,94–100} However, conventional measurement tools do not reveal directly how heterogeneity among nanoparticles, particle-particle interactions, and the irregular electrode morphology contribute to the photoelectrochemical response. For example, van de Lagemaat, Frank, and co-workers showed that “rare” nanoparticles likely limit the electron transport efficiency through TiO₂ nanoparticle photoelectrodes.^101,102 Identifying rare particles requires new experimental tools with single particle-level spatial resolution. Single particle-level measurements enable structure/function correlations between particle size, shape, composition, and photoelectrochemical properties that remain hidden in ensemble level measurements. It is critical to develop single particle-level measurement approaches because the ensemble-averaged photoelectrochemical response may be dominated by a small population of highly active “champion”¹⁰³ or inactive “spectator” particles.¹⁰⁴ However, measuring photocurrents from single semiconductor nanoparticles is difficult.

The photocurrent response from a single nanoparticle is expectedly small. The photocurrent (i) from a single 3 nm-diameter nanocrystal (NC) can be calculated according to i = qI_abs/η, where q is the electronic charge (C), I_abs is the number of absorbed photons per NC (s⁻¹), and η is the overall absorbed photon to current efficiency. One can estimate I_abs from I₀σ, where I₀ is the incident laser power density (s⁻¹ cm⁻²) and σ is the NC absorption cross section (cm²). If 25 µW of 532 nm green laser light excites a 500 nm-diameter spot on the working electrode (I₀ = 3 × 10²² s^–1 cm^–2) that contains a single 3 nm-diameter NC (with σ = 1 × 10⁻¹⁴ cm²),¹⁰⁵ then I_abs = 3 × 10⁸ s^–1. Assuming η = 1 or 100%, then i = 50 picoamperes for a single 3 nm NC under the above illumination conditions. Thus, a major challenge in single particle photoelectrochemical measurements is to detect the small light-induced current response.

The rise of single particle photoelectrochemistry began in 2016. Several research groups developed experimental approaches to measure the small photoelectrochemical response of single nanoparticle electrodes. Here, we define a single nanoparticle electrode as a semiconducting material with at least one critical dimension below 100 nm. This definition includes <100 nm-wide 1D semiconductor nanowires and <100 nm-thick 2D semiconductor nanoflakes. In this perspective, we focus on single particle-level investigations of photoeffects at single nanoparticle electrodes and do not consider spatially resolved measurements of single particle electrocatalysis.^106,107 We consider studies that directly link photoelectrochemical properties with the nanoparticle structure and do not highlight interesting single particle photoelectrochemistry collision experiments.^108–110 Single particle-level studies have revealed underlying performance heterogeneity in nanoscale semiconducting systems, and they strive to predict new architectures to improve ensemble-level performance.

One approach to measure single nanoparticle photoelectrochemical behavior is to construct an electrochemical cell where a single nanoparticle working electrode is immersed in a liquid electrolyte. This design confines current flow through the single nanoparticle/electrical contact interface and ensures that the electrochemical current under illumination stems from a single nanoparticle. In 2016, Su et al.¹⁰⁴ developed the single nanoparticle photoelectrochemical cell approach to study HER on single Si nanowire photocathodes. The authors grew vertically aligned silicon nanowires onto a patterned silicon substrate so that electrical micropads contact each individual wire [Figs. 2(a)–2(b)]. To confine the electrochemistry to a single nanowire, a microchamber filled with liquid electrolyte was positioned over a set of nanowires [Fig. 2(c)] and the electrochemical response from a single nanowire was measured via its individual contact pad. A Pt wire electrode was inserted through the chamber sidewall for two-electrode measurements; a third microreference electrode could presumably be employed in a similar manner for three-electrode measurements. Simulated sunlight from a xenon arc lamp illuminated the entire sample from above [Fig. 2(c)]. Although the light source excites many nanorods on the silicon substrate, the electrochemical current stems from the single nanowire in the microchamber electrochemical cell. The experiment geometry confines hole transport from the nanowire interior to the nanowire/pad interface and electron flow to the nanowire/solution interface.

Su et al. studied the influence of single particle-level behavior on the ensemble-level response of a nanowire array electrode. The authors synthesized two types of nanowires: (1) p-type Si nanowires (or p-Si) and (2) p-Si nanowires with a surface n-type layer that forms a built-in p-n junction (n⁺p-Si nanowires). The vapor-liquid-solid (VLS) growth process allowed for precise control over nanowires’ electronic properties. Pt nanoparticle HER catalysts were deposited onto the nanorod surfaces by soaking the entire nanowire-coated silicon chip in K₂PtCl₆ for 3 min. Figure 2(d) shows current-voltage (I-V) characteristics of p-Si and n⁺p-Si nanowire devices under dark and illumination conditions. The dark currents for both devices were less than 1 pA. Upon illumination of the n⁺p-Si device with 100 mW/cm² AM 1.5G sunlight [green trace in Fig. 2(d)], the authors observed that the cathodic current due to hydrogen evolution onsets at 0.5 V vs reversible hydrogen electrode (RHE), increases approximately exponential with increasingly cathodic potentials, and saturates at −0.2 V. The incident photon flux limits the photocurrent magnitude at cathodic potentials. The onset potential and plateau region are shifted to more positive potential for the p-Si device [red trace in Fig. 2(d)]. The more favorable (i.e., negative) onset potential for the n⁺p-Si device was attributed to increased charge separation due to increased band bending at the surface-doped semiconductor nanowire/solution interface.

Single nanowire measurements revealed an intriguing correlation between the photovoltage output of a large area nanowire array electrode and that of single nanowires. Figure 2(e) shows the distribution of photocurrent onset potentials for single nanowire devices (parameterized as V_ΔI); the onset potential is an indicator of the device photovoltage (V_OC). On average, the V_ΔI values for n⁺p-Si devices were larger than the p-Si devices [530 mV vs 280 mV, as indicated by the top and bottom of the pink rectangular band in Fig. 2(e)]. Interestingly, the V_OC values measured from nanowire array devices were lower than the average V_OC values of single nanowire devices. In fact, the V_OC values of nanowire array devices closely matched that of single nanowire devices with the lowest V_ΔI values. This correlation suggests that the worst-performing (smallest V_OC) nanowires in a nanowire array limit the photovoltage of the entire nanowire array device. While the exact cause of this behavior is not entirely understood, the authors attributed the correlation to nanowire-dependent materials properties that are presumably introduced via the wire growth and fabrication procedures. The intriguing photovoltage correlation, which could only be obtained via single particle-level measurements, emphasizes that material quality must be carefully controlled and optimized to obtain large area nanowire array photoelectrodes with the maximum possible power output.

Another approach to single particle electrochemistry is to spatially control charge flow at the nanoparticle/electrolyte interface. In 2016, Velický et al. used a microdroplet electrochemical approach to study the electron transfer kinetics and photoelectrochemistry of pristine monolayer and few-layer MoS₂.¹¹¹ Figure 3(a) shows the experimental setup. MoS₂ nanoflakes were exfoliated from bulk crystals and deposited on an insulating polymer-coated Si substrate. The MoS₂ flake was electrically connected to a copper wire using conductive paint, thereby forming a single nanoentity working electrode [Ag paint appears in the top of Fig. 3(a)]. Atomic force microscopy (AFM), Raman microspectroscopy, and photoluminescence (PL) microspectroscopy were used to characterize the layer thicknesses within different regions of a single nanoflake [Figs. 3(c)–3(d)]. Following detailed sample characterization, a micropipette delivered 20-50 μm-sized electrolyte droplets to different regions on the MoS₂ nanoflake. Each droplet was positioned onto specific regions of individual flakes to obtain local (photo)electrochemical measurements as a function of MoS₂ layer thickness. The authors measured scan rate-dependent cyclic voltammetry (CV) data using a reversible redox couple, [Ru(NH₃)₆]^3+/2+, under dark and illumination conditions [Fig. 3(e)]. The heterogeneous electron transfer rate constant (k⁰) was determined from the peak separation vs scan rate data. This approach was developed for metal electrodes¹¹² and does not take into account the potential-dependent surface electron concentration in the semiconducting MoS₂ electrode.¹¹³ The authors reported thickness-dependent k⁰ and capacitance values for these exfoliated monolayer, few-layer, and bulklike MoS₂ nanoflakes [Figs. 3(f)–3(g)]; these observations remain hidden in ensemble-level measurements. Bulk MoS₂ exhibited faster electron transfer kinetics than monolayer MoS₂. In addition, k⁰ increased linearly and nonlinearly with light intensity for bulklike and monolayer MoS₂, respectively. The authors attributed this observation to differences in light absorption and carrier diffusion in monolayer vs bulk samples. The intensity-dependent k⁰ behavior for the bulk sample is distinct from ideal bulk semiconductor/electrolyte interfaces that exhibit a light intensity-independent k⁰.^114,115

Hill and co-workers developed a scanning electrochemical cell microscopy (SECCM) approach to study the photoelectrochemical properties of individual WSe₂ nanosheets.⁸⁵ The authors used a pipette that was mounted on an XYZ stage to scan nanoscale electrolyte droplet across a single nanosheet. The liquid electrolyte droplet defines the spatial resolution of the technique (200–300 nm) and ensures that the electrochemical current response stems from only one nanosheet; ∼30–100 nm droplets are routinely used, and 10 nm droplets are possible.^116,117 The authors deposited p-type WSe₂ nanosheets on conductive ITO electrodes and scanned a 200–300 nm electrolyte droplet across the nanosheet surface using a motorized XYZ stage [Fig. 4(a)]. In this SECCM approach, a large light spot illuminates the entire nanosheet uniformly, while photoelectrochemical reactions take place at the confined nanosheet/droplet region. The conductive substrate provides electrical contact to many nanosheets simultaneously instead of the single sheet electrical contacts employed in Fig. 3(a). Figure 4(b) shows the experimental time sequence during a single SECCM measurement: (1) the pipette-sample distance decreases until the electrolyte makes contact with the nanosheet, (2) the current spikes when the droplet contacts the nanosheet, and (3) the electrochemical current-potential response is measured under chopped light conditions. The pipette retracts from the surface and hops to a new location after each measurement. A fast (>1 V/s) linear potential sweep waveform was chosen to minimize data collection time at each pixel. The z-movement of the pipette enables simultaneous topographic and photocurrent measurements of a single nanosheet [Figs. 4(c)–4(e)]. Each pixel contains dark and light current-potential information [Fig. 4(f)].

Hill and co-workers studied the photoelectrochemical behavior of p-WSe₂ nanosheets in electrolytes that contained outer sphere [Ru(NH₃)₆³⁺ reduction] and inner sphere (H⁺ reduction) redox species. The inner sphere reaction is likely more surface sensitive than the outer sphere reaction. Therefore, the inner sphere reaction could indicate variations in catalytic properties due to surface structural heterogeneities (e.g., defects). Indeed, the photocurrent maps of the Ru(NH₃)₆³⁺ reduction reaction were much more uniform across the basal planes and at edge sites than for the inner sphere HER reaction. The authors analyzed the spatial distribution of photocurrents and discovered structure/function relationships on these illuminated nanosheets. On average, the photocurrent decreases at large steps (>50 nm) and increases at small steps. Thus, edge sites on these WSe₂ nanosheets serve as both catalytic sites for HER and charge recombination centers; this behavior depends on the step height. In addition, the onset for the HER reaction shifts to negative potentials at all steps, indicating that a larger driving force is needed to drive the photoelectrochemical HER reaction at step edges vs basal planes. These results, coupled with charge transport simulations, led the authors to propose that WSe₂ nanosheets with a large density of short, steplike features could serve as efficient photoelectrodes for solar fuels production. These predictions, which could only be gleaned from single particle-level measurements, illustrate how nanoscale measurements can guide the design of nanostructured photoelectrode architectures.

The single nanoparticle electrochemical cell and SECCM approaches both spatially control current flow in the electrochemical cell. The experimental geometries employ small area electrical or solution contacts that minimize background currents and enable sensitive current-voltage curves under both dark and light conditions. For example, Su et al. measured electrochemical currents with picoampere sensitivity. The removable sample chamber allows for rapid testing of multiple electrolyte solutions as well as presample and postsample characterization via SEM imaging and associated spatially resolved materials characterization techniques such as AFM, STM, X-Ray photoelectron spectroscopy (XPS), Raman, and PL microspectroscopy. In addition, SECCM provides sample topography information at the same time as the photoelectrochemical measurement. Simultaneous photocurrent and topography measurements are particularly important and useful for layer thickness-dependent studies of 2D materials. SECCM is a high-throughput method despite the point-by-point nature of the measurements.

One limitation of the single nanoparticle cell approach is the need to make electrical contacts to single nanoparticles. The serial nature of the measurements leads to lower throughput. In addition, because the electrolyte surrounds the entire working electrode surface, it is not possible to obtain subparticle resolution HER activity along the vertically aligned nanowire without additional structured illumination methods. SECCM is a high-throughput method, but it is unclear to what extent electrolyte species adsorb to the electrode surface and influence the photoelectrochemical response upon repetitive scans of the same nanoparticle electrode.

In 2016, Sambur et al. developed a single nanoparticle illumination approach to measure the electrochemical response of a single TiO₂ nanorod.¹¹⁸ In this approach, a tightly focused laser beam excites a near diffraction-limited spot (390 nm) on a single nanorod that is supported by an ITO electrode. The light-induced current response stems from a single nanoparticle even though hundreds or thousands of particles may also be deposited on the same ITO electrode. The total electrochemical current in a typical single particle illumination experiment stems from (1) the non-Faradaic double layer charging current of the entire electrochemical cell, including the nanoparticles and the exposed electrode substrate, (2) the Faradaic dark current due to redox reactions at the nanoparticle and/or bare substrate, and (3) the Faradaic photocurrent due to redox reactions at the single nanoparticle interface. The major challenge in single particle illumination experiments is to measure the photocurrent response from a single nanoparticle amongst the double layer charging and dark catalysis currents. Double layer charging currents may be minimized by employing slow scan rates in potential sweep measurements (<1 mV/s) or waiting for long times in chronoamperometry measurements. Faradaic dark currents may be minimized by working in an applied potential window where the electrochemical reaction in the dark is insignificant [see dark current data in Fig. 2(d)]. To overcome this detection challenge, Sambur et al. used chopped laser excitation and measured the nanoampere-level current response using a lock-in amplifier.

Hesari et al. extended this single particle illumination approach to study the role of particle-particle interfaces on charge carrier collection efficiency.¹¹⁹ Particle-particle interfaces are ubiquitous in nanoparticle film photoelectrochemical cells such as dye sensitized solar cells. Charge carrier transport in nanoparticle films occurs over micrometer length scales and across multiple particle-particle interfaces. The authors designed an elegant photoelectrode configuration to study current flow through a single particle-particle interface [Fig. 5(a)]. In this design, micrometer-long TiO₂ nanorods were spin-coated on interdigitated ITO electrodes. The spin-coating conditions were optimized so that the nanorods deposited in two configurations: (1) a single nanorod attached to an edge of an ITO strip or (2) two nanorods linked in a cross structure and only one nanorod touched the ITO edge. The TiO₂ nanorod coated-ITO electrode served as the working electrode in a three-electrode microfluidic photoelectrochemical cell that was mounted on the stage of an inverted optical microscope. A 375 nm laser was directed through a microscope objective, and the light was focused to a 380 nm diameter excitation spot on the sample. The authors excited two types of nanorod regions: type-A spots located before the particle-particle interface and type B spots located after the particle-particle interface [see red and green circles, respectively, in Fig. 5(b)]. Photogenerated electrons within type B spots must traverse a single particle-particle interface to reach the ITO electrode. Photogenerated holes move to the solid/electrolyte interface and participate in either the water oxidation or sulfite oxidation reaction. The hole diffusion length in these TiO₂ nanorods was about 100 nm.⁵⁸ Since the excitation spot size is much larger than the diffusion length, the photogenerated holes likely react within the focused laser spot region and therefore the photocurrent reports on local hole-induced surface reactivity. The authors studied the influence of the particle-particle interface on ϕ_sep by choosing a photoelectrochemical reaction (sulfite oxidation), where ϕ_int is near-unity.

Figure 5(f) shows the average photocurrent-potential curves from >150 type-A and type-B spots (red and green circles, respectively). On average, the type-B spots exhibit a ∼31% smaller photocurrent response at +0.4 V than type-A spots. The significant photocurrent decrease can be attributed to the single particle-particle interface. Next, to understand the role of charge carrier generation distance (d) to the ITO electrode for type-A and type-B spots, the authors measured the steady state photocurrent i_ss as a function of d for both spots on 59 different cross structures. The binned and average single-spot data in Fig. 5(g) shows that i_ss decreases approximately exponentially with increasing distance from the electrode edge; the magnitude and decay depend on the spot type. The authors measured i_ss vs d data as a function of applied potential, quantitatively analyzed the i_ss-d decay behavior, and ultimately quantified the average photocurrent loss due to a single particle-particle interface [parameterized as Δi_ss,0 in Fig. 5(h)]. The relative photocurrent loss due to a single particle interface (∆i_ss,0) varies from 11% to 27% over the potential range of +0.2–0.4 V, corresponding to an average loss of 20% for a single interface. Extrapolating this result to 11 interfaces would cause the photocurrent to drop below ∼10% of the original response in the absence of particle interfaces. This study demonstrated that sub-particle-level photoelectrochemical measurements can be applied to study charge transport in multiple nanoparticles which are the relevant electrode architectures for practical photoelectrochemical applications.

Our group has recently extended this focused laser illumination approach to study 2D MoSe₂ nanoflakes^120,121 and monolayer MoS₂, WSe₂, and MoSe₂ nanosheets.^122,123 In a typical experiment, a nanosheet-coated ITO electrode is assembled into a three-electrode electrochemical cell that is filled with 1M NaI electrolyte and mounted on a confocal-type Raman and PL microspectroscopy setup. A focused laser beam excites the samples in a point-by-point fashion by moving the sample stage in nanometer or micrometer-scale increments. Upon illumination of these n-type materials with focused laser excitation, photogenerated holes travel to the nanosheet/electrolyte interface and oxidize I^– to I₂. The I₂ product rapidly reacts with I^– to form I₃^–. Photogenerated electrons travel to the nanosheet/ITO contact. The potentiostat measures electrochemical current as a function of the excitation position. Contrast in scanning light spot photocurrent maps report on the net collection efficiency of carriers as a function of carrier excitation location; the photocurrent maps do not report on the location of interfacial charge transfer reactions at the semiconductor electrolyte/interface.

Figure 6(b) shows a map of the external quantum efficiency (EQE) of the single MoSe₂ nanoflake in Fig. 6(a), where EQE = i/qI₀, i is the photocurrent (in units of A), and I₀ is the incident laser power (in units of s^–1). To explore the role of charge carrier recombination at perimeter edge sites [indicated by red pixels in Fig. 6(b)], we analyzed how the EQE varied with distance to the nearest perimeter edge [r in Fig. 6(c)]. Figure 6(c) shows that the EQE increases with r for the nanoflake shown in Figs. 6(a)–6(b). This trend, which holds for 59 other nanoflakes that were analyzed in the experiment, indicates that photogenerated electrons near perimeter edges recombine rather than being collected at the ITO/MoSe₂ contact, in agreement with the previous literature that reports exposed Mo atoms at edge sites serve as recombination centers.^124–126 We also quantified the photocurrent efficiency of illuminated perimeter vs interior edge sites. Our analysis revealed that illuminated interior edges produce more photocurrent than illuminated perimeter sites [Fig. 6(d) inset] and 4% of all perimeter edges are “hot” edge sites that exceed the ensemble-average EQE of the parent bulk crystal that was used for mechanical exfoliation [Fig. 6(d)].

In 2019, our group developed a correlated scanning laser reflection and photocurrent microscopy approach to study structure/photoelectrochemical property relationships in monolayer, bilayer, and few-layer-thick 2D materials.^122,123 In this approach, a Raman microspectroscopy setup measures the laser reflection intensity signal as the focused light spot moves across the sample. The reflection intensity increases when the focused laser spot excites 2D materials and decreases when the spot excites the ITO electrode. This technique acquires a structural image of the sample (laser reflection map) at the same time as the functional image (photocurrent map).

We applied this correlated laser reflection/photocurrent mapping approach to study the influence of high intensity laser illumination on the photoelectrochemical properties of MoSe₂ photoelectrodes. Several studies have shown that high intensity light irradiation or laser annealing can tune the optical and physical properties of pristine 2D materials.^127–133 Those studies suggested that laser annealing could be beneficial for photoelectrocatalytic applications, but the effect had not been studied at the ensemble or single nanosheet-levels. To do so, we measured photoelectrochemical maps of ultrathin MoSe₂ nanosheet-coated ITO electrodes before and after a high-power laser annealing procedure. Figure 6(f) shows a photocurrent map of the pristine MoSe₂ sample region in Fig. 6(e). The sample contained mostly 10–15 µm wide monolayer-MoSe₂ and bilayer-MoSe₂ triangles. The 3–10 µm wide dark black objects represent multilayer-thick or bulk MoSe₂. The bright and dark regions in Fig. 6(f) qualitatively correlate with monolayer and bilayer regions of MoSe₂, respectively, which suggested that pristine bilayer regions did not produce a measurable photocurrent response. The photocurrent map after the first laser treatment showed an abrupt enhancement for both monolayer and bilayer MoSe₂ [Fig. 6(h)]. Interestingly, we observed that bilayer MoSe₂ apparently activates from edge sites because the inner-most areas remain inactive following the first laser treatment [Fig. 6(h)]. The photocurrent map after the second focused laser treatment [Fig. 6(i)] shows completely active bilayer MoSe₂ regions; the bilayer photocurrent response exceeded that of the monolayer [Fig. 6(j)]. The photocurrent enhancement effect was attributed to a vacancy healing effect where ambient O fills Se vacancy sites as evidenced by ensemble-level XPS measurements. This vacancy healing process removes trap states and increases the overall photogenerated carrier collection efficiency. An important technical aspect of the study is that we developed an image analysis procedure, based on the correlated laser reflection microscopy technique, to quantitatively overlay the structural information contained in transmission images onto the photocurrent maps [Fig. 6(g)]. The overlay procedure enabled us to extract photocurrents from monolayer and bilayer regions and determine structure/function properties as a function of layer thickness and laser annealing steps [Fig. 6(j)].

Like SECCM, the scanning light spot approach is a high-throughput method to study 0D nanoparticles, 1D nanorods, and 2D nanosheets. The focused light spot serves as not only a charge carrier generation source but also a probe for optical spectroscopy techniques such as PL and Raman microspectroscopy; PL and Raman spectra may be acquired at every pixel in the photocurrent map.^120,122 The technique provides detailed information on the relationship between carrier generation and recombination, but it unfortunately does not report directly on the location of photoelectrochemical reactions at the solid/liquid interface. Unlike SECCM, the scanning light spot approach does not enable direct measurements of the dark current-potential relationship of single nanoparticles. The spatial resolution of the technique is typically limited by the diffraction-limited focused light spot (∼250 nm) but could be improved with near field optics.^134–136 Since the laser power density is typically high in focused light spot measurements (>kW/cm²), the chemical and structural integrity of the sample should be monitored as a function of illumination time.

In this perspective, the field of single particle photoelectrochemistry focuses on two major goals. The first goal is to develop fundamental knowledge of the nanoparticle/electrolyte interface. The physical chemistry of nanoscale semiconductor/electrolyte interfaces differ from bulk/semiconductor interfaces. We do not fully understand how to model the semiconductor/electrolyte interface when the material dimensions are smaller than the space charge region thickness, as is the case for nanostructured electrodes (see Secs. I B and I C). There are two important points to consider: (1) a small change in the spatial distribution of dopants likely has a large change on the electric field profile in the particle, and as a result, (2) one must take into account the arrangement of dopants and/or ionized impurities in the particle when modeling the photocurrent-potential response. Some models have considered the potential distribution of fully depleted nanoscale semiconductors,^82,83,137 but the possibility of incorporating atomistic-level details into the photoelectrochemical measurements has not been tested experimentally at the single nanoparticle-level.

The second goal is to offer design strategies for photoelectrochemical applications based on nanostructured architectures. Spatially resolved photoelectrochemical measurements often reveal inactive (or highly active) regions within a single nanowire or nanosheet. If the inactive (or highly active) regions share similar chemical and structural features across many individual objects (e.g., terraces width, step edges, or perimeter edge sites), then the activity information may be used to predict an optimum nanoparticle morphology for a desired photoelectrochemical reaction. Another possibility is to use the spatially resolved measurement platform to not only characterize activity but also modify the pristine nanoparticle electrode. For example, Sambur et al. used focused laser illumination to measure the photocurrent due to water oxidation on a single nanorod as well as locally deposit a water oxidation cocatalyst on desired spots on the nanoparticle surface.¹¹⁸ Thus, single particle photoelectrochemical platforms hold tremendous promise for creating multicomponent semiconductor nanoparticle electrodes with tailored photoelectrocatalytic properties. Sections III A and III B provide our perspective on experimental challenges that must be overcome to achieve the aforementioned goals.

The atomic-structure of the nanoparticle could influence charge separation/transport and interfacial charge transfer in the following ways. The exact positions of impurity dopant atoms in the nanoparticle core can heavily influence the magnitude and direction of the internal electric field that drives charge separation and transport. Aside from the interior impurities/defects, surface impurities and defects can engage directly with solution-phase reactant molecules. Since the defects/impurities have electronic energy levels that are different from the allowed energy levels in the semiconductor, the levels are often referred to as surface defect states. Surface defect states have a variable effect on energy conversion efficiencies. Defect states can either increase overall efficiency by catalyzing interface charge transfer reactions or decrease efficiency by promoting surface charge recombination.¹³⁸ Yet, the total number, spatial distribution, and energy levels are often unknown. No studies exist that use the 3D coordinates of dopant atoms to compute the electric potential distribution in a semiconductor nanoparticle. We envision that the field of single particle photoelectrochemistry is poised to address this issue.

The field must improve the spatial resolution of existing single particle photoelectrochemical microscopy methods to develop atomic-level structure/function relationships. More importantly, the high-resolution photoelectrochemical imaging approaches should be compatible with atomic-scale and nanoscale materials characterization tools such as atom probe tomography¹³⁹ and electron,¹⁴⁰ X-ray,¹⁴¹ and optical¹⁴² spectroscopies. Ideally, the (photo)electrochemical properties of single nanoparticles could be linked with the three-dimensional atomic structure. If we can identify how atomic-level structure contributes to each underlying process in solar energy conversion (charge separation, transport, and interfacial charge transfer), then we can develop design principles for high efficiency photoelectrodes.

While significant attention has been paid to photoeffects at the nanoparticle electrodes, quantitative analysis of the dark current-potential relationship as a function of nanowire electronic and physical properties can yield critical information regarding surface recombination at nanowire/electrolyte interfaces. For example, Maldonado and co-workers showed how dark currents can be used to quantify (1) the surface recombination velocity of electrons and holes and (2) the rate constants for interfacial charge transfer for nanowires with cylindrical or tapered geometries.³⁶ Quantifying the dark current-potential response separately from the light current-potential response allows one to focus on the role of the interior electric field on charge carrier separation and transport steps.

In conclusion, the emerging field of single particle photoelectrochemistry has opened up a new frontier in solar energy conversion research; it is now possible to probe the photoelectrochemical activity of single nanoparticles. We reviewed state-of-the-art experimental approaches and provided our perspective on new frontiers and grand challenges in the field. Single particle photoelectrochemistry has tremendous potential to reveal the underlying physical chemistry of nanoparticle/electrolyte interfaces and offer design strategies for photoelectrochemical applications based on nanostructured architectures.

This material is based upon the work supported by the Air Force Office of Scientific Research (AFOSR) under award number FA9550-17-1-0255.