1.

Introduction

1.2.

Slow Photons and Photonic Crystals

To clarify how PCs can enhance photocatalytic processes, including ${\mathrm{H}}_2$ generation, some background information on PCs and how they interact with light is necessary. Whether in pure or sensitized photocatalyst materials, the phenomenon characteristic of PCs that has been most commonly exploited to enhance light absorption in semiconductor photocatalysts is the “slow photon effect”—a slowing of light propagation in photonic materials at wavelengths near the photonic bandgap. PCs are ordered, periodic structures with alternating regions of high and low refractive index that satisfy Bragg diffraction conditions.^8–11 A “forbidden band” of radiation with wavelengths on the same order as the periodicity of the PC structure cannot propagate through the PC, leading to a strong reflectance. The center of this band of radiation is known as the stop band or photonic bandgap. At the high- (blue) and low-energy (red) regions of the stop band, photons are multiply scattered in the lower refractive index and higher refractive index portion of the PC, respectively (Fig. 1),¹² thus effectively lowering the group velocity of light at the edges of the stop band.¹³ The propagation of slow photons increases their optical path length within the semiconductor or sensitizer and enhances the probability of generating charge carriers.

Fig. 1

Simplified optical band structure of a PC. At photon energies approaching a full bandgap or a stop band from the red side, the group velocity of light decreases and light can be increasingly described as a sinusoidal standing wave that has its highest amplitude in the high-refractive index part of the structure. At energies above the band gap or stop band, the standing wave is predominantly localized in the low-index part of the photonic crystal. Reproduced with permission from Ref. 12. Copyright 2003 American Chemical Society.

Although more exotic structures exist, inverse opals (IOs) are the most common PC structure utilized in photocatalysis. These structures are typically prepared by templating monodisperse polystyrene spheres in an opal arrangement and infiltrating the voids of the face-centered cubic (fcc) array with oxide precursors before removing the template by calcination^12,14–18 (Fig. 2). The photonic bandgap (${\lambda }_{\mathrm{p}}$) for the IO films obeys a modified version of Bragg’s law

Eq. (1)

$${\lambda }_{\mathrm{p}}=1.6D\sqrt{n_{\mathrm{p}}^{2}f+n_{\mathrm{s}}^2(1-f)}\sin \theta ,$$

where $D$ is the diameter of the spherical voids in the fcc structure, $f$ is the volume fraction of the spherical pore volume, $n_{\mathrm{p}}$ is the refractive index of the pore-filling fluid, $n_{\mathrm{s}}$ is the refractive index of the solid framework, and $\theta$ is the angle between incident light and (111) plane of the fcc structure.¹⁵ The exact frequency of the photonic bandgap is highly sensitive to the refractive indices of the solid and voids and will redshift as the refractive index of the pore-filling fluid increases (Fig. 3). The critical spatial dimension, $D$, is controlled directly during IO synthesis; the stop band is synthetically tuned as a function of the spherical template diameter (Fig. 4).¹⁹

Fig. 2

Scanning electron micrograph (SEM) of (a) colloidal crystal template of polystyrene spheres (200 nm) and (b) three-dimensional ordered macromesoporous architecture of $\mathrm{Mo}:{\mathrm{BiVO}}_4$. Reproduced with permission from Ref. 18. Copyright 2014 American Chemical Society.

Fig. 3

Transmission UV–visible spectra of a ${\mathrm{TiO}}_2$ IO structure with 169-nm voids and filled with solvents of different refractive index and insert showing band position as a function of refractive index of the void-filling solvent. The photonic stop band redshifts as the refractive index of the pore-filling fluid increases. The refractive index of the ${\mathrm{TiO}}_2$ solid framework was estimated to be 2.40. Reproduced with permission from Ref. 15. Copyright 1999 American Institute of Physics.

Fig. 4

Transmission spectra measured at normal incidence of opal structures deriving from three different polystyrene sphere diameters. Reproduced with permission from Ref. 19. Copyright 2011 American Chemical Society.

2.

Coupling Photonic Crystals to Photocatalysis—Tuning the Red Edge of the Stop Band to the Photocatalyst Absorption Maximum

3.

Hydrogen Generation Photocatalysis at Photonic Crystals

Using the design principles described above, many researchers have exploited the slow photon effect to enhance ${\mathrm{H}}_2$ generation by photochemical or photoelectrochemical water splitting in photocatalysts including ${\mathrm{TiO}}_2$,^27–30 ${\mathrm{WO}}_3$,¹⁹ ${\mathrm{Fe}}_2{\mathrm{O}}_3$,³¹ ${\mathrm{Bi}}_2{\mathrm{WO}}_6$,³² and ${\mathrm{BiVO}}_4$^18,33 (results and references are summarized in Table 1). At Pt-modified ${\mathrm{TiO}}_2$ PCs,²⁷ the photocatalytic efficiency of ${\mathrm{H}}_2$ generation as a function of the stop band position obeyed the same trends as those observed for dye degradation reactions by Chen et al.²⁰ The largest enhancement factor under broadband excitation of 2.5 occurred when the stop band was tuned to 342 nm—where the red edge overlapped the ${\mathrm{TiO}}_2$ band edge.²⁷

Table 1

Diverse examples of photonic enhancement of photocatalytic and photoelectrochemical water splitting/H2 generation reactions. Note that conditions that may impact the enhancement factor, such as excitation conditions, use of hole scavengers in reaction slurry/electrolyte, and photoanode bias in photoelectrochemical measurements, are omitted for the sake of brevity.

While narrower bandgap alternatives to ${\mathrm{TiO}}_2$ for photocatalytic water splitting^2–4 may absorb more visible light, none features 100% photon-to-carrier conversion efficiency, and thus all can still benefit from improved light utilization. Alternatives to ${\mathrm{TiO}}_2$ also suffer from drawbacks involving other parts of the photocatalytic process, such as valence/conduction band energies that are inefficient for water-splitting chemistry, and from fast electron–hole recombination kinetics. In particular, ${\mathrm{Bi}}_2{\mathrm{WO}}_6$ and ${\mathrm{BiVO}}_4$ suffer from high bulk recombination rates and thus benefit from the shortened carrier path lengths to the semiconductor surface featured in the IO architecture.^18,32–34 Nan et al.³⁴ show particularly strong enhancement factors of both surface-normalized photocurrents and photocatalytic oxygen and ${\mathrm{H}}_2$ generation at the IO form of ${\mathrm{BiVO}}_4$ (Fig. 5). While the red edge of the stop band in their ${\mathrm{BiVO}}_4$ IO structures ($\sim 450\mathrm{nm}$) overlaps well with the band edge of ${\mathrm{BiVO}}_4$, their observed enhancement is probably due to more than the slow light effect alone. The simultaneous improvement of light management and electron–hole separation with IO forms of ${\mathrm{Bi}}_2{\mathrm{WO}}_6$ and ${\mathrm{BiVO}}_4$ enhances photoelectrochemical water splitting relative to non-PC forms of the semiconductors to a significantly greater extent than that achieved when expressing ${\mathrm{TiO}}_2$ in PC form (about $3\times$ to $8\times$ for ${\mathrm{Bi}}_2{\mathrm{WO}}_6$ and ${\mathrm{BiVO}}_4$ compared to about $2\times$ to $2.5\times$ for ${\mathrm{TiO}}_2$, Table 1). Note that while the improvement is more substantial in the PC forms of ${\mathrm{Bi}}_2{\mathrm{WO}}_6$ and ${\mathrm{BiVO}}_4$, the overall photoefficiencies are still typically lower than that achievable with ${\mathrm{TiO}}_2$.

Fig. 5

Chronoamperometry measurements for (1) $\mathrm{I}\mathrm{O}\text{-}{\mathrm{BiVO}}_4$ films modified with carbon quantum dots; (2) unmodified $\text{IO}\text{-}{\mathrm{BiVO}}_4$ films; (3) Mo-doped ${\mathrm{BiVO}}_4$ films; undoped ${\mathrm{BiVO}}_4$ films. (a) Chronoamperometry measurements at an external potential of $0.9V$ versus the reversible ${\mathrm{H}}_2$ electrode (RHE) with repeated on/off cycles under UV/visible irradiation; (b) linear sweep voltammetry measurements at $50\mathrm{mV}{\mathrm{s}}^{-1}$ under UV/visible irradiation—the dashed lined represents the dark current of the undoped ${\mathrm{BiVO}}_4$ film; (c) photoconversion efficiency as a function potential versus RHE; (d) surface-area-normalized ${\mathrm{H}}_2$ and ${\mathrm{O}}_2$ generation rates under UV/visible irradiation. Reproduced with permission from Ref. 34. Copyright 2015 American Institute of Physics.

Zhang et al.³³ compared the photoelectrochemical water-splitting activity of IO and planar $\mathrm{Mo}:{\mathrm{BiVO}}_4$ photoanodes. When the stop band was tuned to 513 nm, near the $\sim 520\text{-}\mathrm{nm}$ electronic bandgap of the oxide, photoelectrochemical activity was enhanced by about $8\times$. Photoelectrochemical activity was still enhanced—but to a much lesser extent—when the stop band was tuned to a frequency at which slow photons did not overlap the electronic bandgap excitation of ${\mathrm{BiVO}}_4$. The enhanced activity in the absence of the slow photon effect is attributed to improved charge separation imparted by the PC architectures. Zhou et al.¹⁸ used transient photocurrent decay to measure the extent to which IO architectures enhanced charge migration during photoelectrochemical water splitting in ${\mathrm{BiVO}}_4$ PCs. Carrier lifetimes were extended in ${\mathrm{BiVO}}_4$ photoanodes with a macroporous IO structure compared to those observed in disordered films. The longest lifetimes were measured in IO structures containing both macro and mesoporosity; photoelectrochemical water-splitting activity was enhanced by roughly $3\times$ to $7\times$ in such films.

Mitchell et al.³⁵ applied a somewhat different approach to PC-enhanced photocatalysis, in which CdS nanocrystals (NCs) supported in a photocatalytically inactive zirconium oxide (${\mathrm{ZrO}}_2$) PC performed photocatalysis directly, instead of acting as a sensitizer. They observed a maximum $\sim 5\times$ enhancement for ${\mathrm{H}}_2$ generation under visible light illumination relative to nonphotonic $\mathrm{CdS}/{\mathrm{ZrO}}_2$ at the “blue edge” of the PC stop band, suggesting that the supported CdS nanoparticles exploit slow photons in the lower-index voids in the PC—in contrast to the case in which the PC support itself is photocatalytic. Their results—combined with the myriad results that demonstrate the opposite case in which the PC itself acts as the photocatalyst material and enhancement of photocatalysis occurs via localization of light coincident with the red edge of the stop band in the PC—convincingly demonstrate that the localization of the different band edges in PC-derived photocatalysts can be effectively tailored to application via materials design.

4.

Combining Photonic Crystal Architectures with Plasmonics

Plasmonic metal nanoparticles, in particular Au and Ag, have recently received substantial attention as sensitizers for solar water-splitting catalysts due to their chemical stability and surface plasmon–resonant frequencies that overlap the solar spectrum.^40–43 Whereas PCs improve carrier generation at or near the electronic band edge, plasmonic sensitizers can extend the useful range of photocatalysts much farther into the red (i.e., to lower energy, longer-wavelength radiation). Properly applied, the combination of PCs and plasmonic sensitizers enhances photochemical/photoelectrochemical activity to a greater extent than either strategy alone.^{33,36,37,44,45} For example, Zhang et al.³³ measured a roughly 26% enhancement in photoelectrochemical water splitting when they modified planar ${\mathrm{BiVO}}_4$ films with $\sim 20\text{-}\mathrm{nm}$ plasmonic Au nanoparticles, but measured nearly 55% enhancement when they modified PC ${\mathrm{BiVO}}_4$ with the same plasmonic Au nanoparticles. When plasmonic Au was present, the photocurrent response was extended beyond the electronic bandgap of ${\mathrm{BiVO}}_4$ at 530 to 560 nm.

Analogous to other cases of combining photonics with sensitizers, the optimum enhancements from PCs + plasmonic sensitizers are achieved when the slow photon regions are positioned to overlap the surface plasmon resonance (SPR) frequency maximum while simultaneously avoiding reflective losses in electronic bandgap excitation.^36,37,44,45 A large increase in optical extinction occurs when the slow photon region of a PC is well-matched to the SPR.^36,37 Zhang et al.³⁶ coupled ${\mathrm{TiO}}_2$ PC structures to ${\mathrm{TiO}}_2$ nanotube (NT) arrays, creating ${\mathrm{TiO}}_2$ NT/PC (NTPC) assemblies and subsequently deposited SPR-active Au NCs of different sizes onto the ${\mathrm{TiO}}_2$ NTPCs. The intensity of the SPR response of Au NCs with SPR maxima well-coupled to the PC photonic bandgap was more strongly enhanced than those with more poorly matched SPR maxima (Fig. 6). The SPR-driven photoelectrochemical response (Fig. 7) and ${\mathrm{H}}_2$ generation were similarly more strongly enhanced at the well-matched $\mathrm{Au}/{\mathrm{TiO}}_2$ NTPCs. Zhang et al.³⁷ achieved a maximum $2\times$ enhancement in photoelectrochemical water splitting when the red edge of the stop band of PC ${\mathrm{TiO}}_2$ was positioned at 518 nm, close to the SPR frequency maximum of the $\sim 10\text{-}\mathrm{nm}$ Au particles, but at a position where the photonic stop band (at 450 nm) would not interfere with bandgap excitation of the ${\mathrm{TiO}}_2$. When the Bragg reflection of the PC overlapped the Au SPR, the plasmonic enhancement was significantly suppressed.

Fig. 6

Diffuse refelectance UV–visible absorption spectrum of Au NCs on ${\mathrm{TiO}}_2$ NTs and the ${\mathrm{TiO}}_2$ NTPC assembly (with a stop band centered $\sim 552\mathrm{nm}$) with the spectrum of ${\mathrm{TiO}}_2$ subtracted for clarity of the Au SPR features. Reproduced with permission from Ref. 36. Copyright 2013 American Chemical Society.

Fig. 7

Photoelectrochemical properties of the ${\mathrm{TiO}}_2$ NTPC and $\mathrm{Au}/{\mathrm{TiO}}_2$ NTPC. (a) Amperometric $I-t$ curves at an applied potential of $1.23V$ versus RHE under illumination of visible light with wavelength $\ge 420\mathrm{nm}$ with 60 s light on/off cycles; (b) incident photon-to-current conversion efficiency plots in the range of 400 to 700 nm at $1.23V$ versus RHE. Reproduced with permission from Ref. 36. Copyright 2013 American Chemical Society.

5.

Butterflies and Leaves: Alternate, Bioinspired, Photonic Crystal Geometries for Photocatalysis

Nature provides examples of structures that have unique interactions with light—whether they are naturally occurring PCs or disordered structures that multiply scatter light—that have inspired structural photocatalyst mimics. Butterfly wings^38,46,47 and artificial leaves³⁹ have both recently been used as structural templates for photocatalysts. Butterfly wing-templated ${\mathrm{TiO}}_2$ materials that feature regular, bicontinuous, gyroid structures with clear photonic bandgaps have been fabricated.⁴⁶

Although not a PC, the Papilio helenus Linnaeus butterfly wing contains structural elements that improve light utilization, including honeycomb-like hole arrays that multiply scatter light, “ridges” that guide light into the hole array, and a bottom reflective layer that recaptures light escaping from the honeycomb arrays. Platinum-modified ${\mathrm{TiO}}_2$ replicas of P. helenus Linnaeus butterfly wings can function as ${\mathrm{H}}_2$-generating photocatalysts.³⁸ The wing-pattered photocatalyst strongly enhances ultrabandgap light absorption and shows a $\sim 7\times$ improvement in ${\mathrm{H}}_2$ generation relative to nanoparticulate ${\mathrm{TiO}}_2$, although it was unclear how much of this enhancement could be attributed solely to surface area. Demonstrating the hierarchical possibilities of biostructure-derived photocatalytic architectures, Yan and coworkers⁴⁷ deposited plasmonic gold nanorods onto bismuth vanadate (BVO) structures templated from wings of the butterfly Paplio nirius (Fig. 8) and achieved substantial photocatalytic activity at red and near-infrared light illumination. Zhou et al.³⁹ replicated the hierarchical structure of Anemone vitifolia Buch leaves, which contain various components that serve to focus, scatter or guide light and enhance light–matter interactions. Using a two-step infiltration process, they produced an N-doped ${\mathrm{TiO}}_2$ inorganic replica of the leaf structure that showed a roughly $3\times$ to $8\times$ enhancement in ${\mathrm{H}}_2$ generation compared to nanoparticulate ${\mathrm{TiO}}_2$; the activity was enhanced by another order of magnitude when modified with Pt nanoparticles.

Fig. 8

SEM images of (a) top view of original P. nirius wing scale, (b) top view of templated BVO replica of the P. nirius wing scale, (c) cross section of the original scale, (d) cross section of the BVO wing, (e) TEM image of the gold nanorods (AuNR), (f) HRTEM image of a single AuNR, (g) AuNRs-loaded BVO wing with insets showing the simplified model (upper) and an enlarged image (lower), (h) HRTEM image of the AuNRs-loaded BVO wing, and (i) schematic diagram of the controlled assembly of AuNRs onto the BVO unit. Reproduced from Ref. 47 (open access). Copyright 2016 Nature Publishing Group.

6.

Conclusions and Comments

Exploitation of the slow photon effect of photonic bandgap-containing materials requires rigorous matching of the red edge of the stop band with the desired excitation wavelength while avoiding absorption losses due to nonpropagation of light at the stop band. When PCs are correctly applied, photocatalytic activity enhancement factors of about 2 to 3 can be readily achieved, and in some cases enhancement of $5\times$ or more may be possible. It is also possible to achieve light-utilization enhancement via the more disorganized process of internal light scattering, which is less demanding in structural precision, but also more difficult to design in a rational way.

A couple of important caveats should be mentioned. First, application of the slow photon effect to photocatalysis is illumination angle dependent, and as such can only be effectively applied in film geometries in which the illumination angle can be well-controlled. Practical photocatalysts are often used as fluid-dispersed (suspended) powders to maximize both mass transport and light utilization. The reactor geometry limitation inherent to PCs (i.e., a thin-film form-factor) means that careful consideration must be given to mass transport and light harvesting when designing the interfaces. Additionally, solar fuels photocatalysis involves many challenging photophysical and photochemical steps. Light harvesting is critical, but even with substantial improvements in light harvesting, many photocatalyst materials are still 2 to 3 orders of magnitude (or more)—rather than a factor of 2 to 3—from practical viability. One important class of solar fuels photocatalytic reactions omitted from this section involve reduction of carbon dioxide (${\mathrm{CO}}_2$) to hydrocarbons as its final step, rather than protons to ${\mathrm{H}}_2$. Photochemical and electrochemical reductions of ${\mathrm{CO}}_2$ comprise an enormous research challenge in its own right. Improvements on all photocatalytic fronts, however, continue apace; a number of composite solar fuels photocatalyst materials are approaching activity competitive with alternative routes to solar fuels and photonics could conceivably contribute enough additional efficiency to make them viable.