Silicon Carbide Photonic Crystal Photoelectrode

Abstract

The immense challenge of large-scale implementation of photoelectrochemical (PEC) water splitting and carbon fixation lies in the need for a cheap, durable, and efficacious photocatalyst. Cubic silicon carbide (3C-SiC) holds compelling potential due to its auspicious band positions and high-volume, high-quality, single crystal industrial manufacturing, but is hindered by its inferior light absorptivity and anodic instability. A slanted parabolic pore photonic crystal (spbPore PC) architecture with graphitic carbon nitride (g-CN), nickel(II) oxide (NiO), or 6H silicon carbide protective coatings is proposed to overcome the drawbacks of 3C-SiC photoelectrodes. A 30 µm- and 62 µm-thick 3C-SiC spbPore PC of lattice constant 0.8 µm demonstrates maximum achievable photocurrent density (MAPD) of 9.95 and 11.53 mA cm^-2 in the [280.5, 600] nm region, respectively, representing 75.7% and 87.7% of the total available solar photocurrent density in this spectral range. A 50 nm-thick g-CN or NiO coating forms type-II heterojunctions with the 3C-SiC spbPore PC, facilitating the charge transport and enhancing the corrosion resistivity, all together demonstrating the MAPD of 9.81 and 10.06 mA cm^-2, respectively, for 30 µm-thick PC. The scheme advances the low-cost, sustainable, real-world deployment of PEC cells for green solar fuel production.

Keywords: photoelectrochemical cell; photoelectrode; photonic crystal; silicon carbide; water splitting.

Figure 1

Illustration of silicon carbide photonic crystal photoelectrode. a) The photoelectrode is composed of 3C‐SiC slanted parabolic‐pore layer on bulk 3C‐SiC, with g‐CN, NiO, or 6H‐SiC surface coating and SiO₂ rear passivation layer, deposited on a highly reflective conductive substrate and submerged in aqueous solution. Periodic metal protrusions separated by a distance close to the hole diffusion length of 3C‐SiC connect the bulk 3C‐SiC and the metal substrate for charge transport, while minimizing surface recombination of carriers. The slanted parabolic pore is obtained from a rotation of a normal pore of depth h around y‐axis pivoting at the surface edge by an angle θ. The PC consists of a square lattice of such (slanted) pores with periodicity a. The thicknesses of the surface coating, 3C‐SiC and SiO₂ layers measured along vertical direction z are d _coat, d _PC, and d _psv, respectively. The total thickness of the optically active materials is D = d _coat + d _PC. b) Band positions of n‐type 3C‐SiC, p‐types g‐C₃N₄ and NiO in empty space or, equivalently, point of zero charge aqueous environments (thick horizontal lines), and those together with hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) energies in pH = 0 electrolyte (colored bands). In other words, the separately color‐filled bands for each material depict the energies in the pH = 0 aqueous environment, not in physical vacuum. The pH_PZC is 4.9 for 3C‐SiC,^{[ 52 ]} 5.2 for g‐CN,^{[ 53 ]} and 10.3 for NiO.^{[ 54 ]} The physical vacuum (empty space), a potential level of 0 V, is defined as −4.44 V with respect to SHE. This latter energy coincides with the minimum energy of an electron in a pH = 0 solution to cause HER. The illustrated band positions assume flat band conditions, following E = E _PZC + 0.059(pH − pH_PZC) eV, where E _PZC is the band positions of semiconductors in electrolytes of pH = pH_PZC (equated with the band positions in an empty space). The second term of E is due to the formation of a Helmholtz layer on the semiconductor‐electrolyte interface,^{[ 48} , ⁴⁹ , ⁵⁵ , ⁵⁶ , ^{57 ]} causing a static electric field to occur in the surface region of the semiconductor. In general, energy shifting rates (due to this field) with respect to SHE other than 59 meV pH⁻¹ are also possible.^{[ 49} , ^{56 ]} In non‐flat band conditions, more extensive space charge regions near the semiconductor surfaces emerge and the concomitant band bendings (not shown in the figure) may or may not vary with pH,^{[ 58 ]} depending on surface reactions. E _F (dashed lines) denotes typical experimental Fermi levels of the semiconductors, and the rainbow‐shaded region of NiO corresponds to its in‐gap states of d‐d transitions, both illustrated in an empty space (or equivalent pH_PZC) environment.

Figure 2

Light harvesting of square lattice normal parabolic‐pore PC in aqueous solution in the wavelength region of λ ∈ [280.5, 600] nm using the dielectric Model‐L. The MAPD as a function of the lattice constant a, for 5 µm‐ (blue solid circle and dark‐yellow diamond) and 15 µm‐thick (square) PCs with 1 µm pore depth, with (lines) and without (scatter) 50 nm‐thick SiO₂ rear passivation layer.

Figure 3

MAPD as a function of 3C‐SiC thickness d _PC for lattice constant of a = 0.8 µm (blue solid square), 1.15 µm (open circle and triangle), 1.35 µm (cross), and pore depth of h = 1 µm (solid lines), 2 µm (dashed lines), with a 50 nm‐thick SiO₂ rear passivation layer.

Figure 4

MAPD as a function of 3C‐SiC pore depth h from 0 to 15 µm for lattice constant a = 0.8 µm (navy blue) and 1.15 µm (green), and 3C‐SiC thickness 30 µm, with 50 nm‐thick SiO₂ rear passivation layer.

Figure 5

MAPD as a function of 3C‐SiC pore slant angle θ and mesa width w for lattice constant a = 0.8 µm, vertical pore depth h′ = 1 µm or h′ = 1 µm× cosθ and 3C‐SiC thickness d _PC = 30 µm, with 50 nm‐thick SiO₂ rear passivation layer and no front coating layer.

Figure 6

Absorptivity (solid lines) and reflectivity (dashed lines) spectra of 3C‐SiC slanted parabolic pore PC with dielectric Model‐L for 3C‐SiC thickness d _PC = 30 µm, slanted angle θ = 1.75° (navy blue lines), and d _PC = 62 µm, θ = 0° (dark yellow lines). Common parameters are lattice constant a = 0.8 µm, parallel slanted pore depth h = 1 µm, SiO₂ rear passivation layer thickness d _psv = 50 nm and no front coating layer.

Figure 7

MAPD as a function of g‐CN and NiO coating thicknesses on optimized 3C‐SiC slant parabolic pore PC with Model‐L.

Figure B1

Experimental data on the refractive indices n (dotted lines with color fillings) and extinction coefficients κ (lines without color fillings) of a,b) 3C‐SiC and c,d) g‐C₃N₄. The vertical dashed lines demark the frequency range of interest, i.e., λ ∈ [280.5, 600] nm. a, b) Data sets of 3C‐SiC are denoted by author‐year notation followed by the sample name, i.e., “Hofmeister09, impure”: the “impure 3C SiC” in ref. [79]; “Alterovitz91, 10 µm(0.61 µm)”: the 10 µm(0.61 µm)‐thick sample in ref. [137]; “Shaffer71”: 3C‐SiC in ref. [134]; “Logothetidis96”: 3C‐SiC in ref. [139]; “Solangi92, A(B)”: Sample A(B) in ref. [138]; “Nishino75, S1(S2)”: Sample S1(S2) in ref. [135]; “Patrick69, A(B)”: Sample A(B) in ref. [133]; “Choyke88”: 3C‐SiC in ref. [136]; “Sridhara99”: 3C‐SiC in ref. [140]; “Philipp58”: 3C‐SiC in ref. [132]. Model‐L is based on n (navy blue‐sand yellow dashed line with medium purple color filling) and κ (medium purple solid line) of “Hofmeister09, impure”. Model‐M is based on n of “Hofmeister09, impure” and κ of “Solangi92, B” (blue solid line), extrapolated and complemented (deep yellow thick solid line) in part by κ of “Hofmeister09, impure”. Model‐H is based on n of “Hofmeister09, impure” and κ of “Nishino75, S2” (hunter green solid line), complemented (light orange thick solid line) in part by κ of “Nishino75, S1” (light green dashed line). c, d) Data sets of g‐C₃N₄ are denoted by author‐year notation followed by the structure or sample name, i.e., “Datta20, hg‐AA”: the DFT result of AA‐stacked heptazine g‐C₃N₄ in ref. [89]; “Datta20, tg‐AA(AB)”: the DFT result of AA(AB)‐stacked triazine g‐C₃N₄ in ref. [89]; “Reshak14, tg‐AA(AB)”: the DFT result of AA(AB)‐stacked triazine g‐C₃N₄ in ref. [141]; “Fujita16, 483(534, 573, 603) °C”: the sample prepared under the substrate temperature of 483(534, 573, 603) °C in ref. [142]. The Lorentz dielectric model of g‐C₃N₄ is based on n of “Datta20, hg‐AA” (red‐brown‐sand yellow dashed line with medium purple color filling) and κ of “Fujita16, 573 °C” (black solid line), extrapolated to the entire 280 to 600 nm spectral range (light red‐brown thick solid line). The DFT results are the average of three components of n(κ) for electric field polarizing along and perpendicular to the optical axis, which have similar values to that obtained from averaging three components of the dielectric constant. The extinction coefficients of “Datta20, h(t)g‐AA(AB)” presented here are from raw data of ref. [89] via private communication.

Figure B2

Dielectric models of lightly doped (Model‐L), moderately doped (Model‐M), and highly doped (Model‐H) 3C‐SiC, and of g‐CN (or specifically, g‐C₃N₄) and NiO, in the spectral region of λ ∈ [280, 600] nm. Notations of the data sets are the same as in Figure B1. a) The cyan filling, medium purple filling, and walnut solid line correspond to the refractive indices of experimental measurement on impure 3C‐SiC in ref. [79], DFT calculation on AA‐stacked heptazine g‐C₃N₄ in ref. [89], and experimental measurement on thin‐film NiO in ref. [113]. b) The navy blue, caribbean green, red, black and walnut solid lines correspond to the extinction coefficients of experimental measurements on impure 3C‐SiC in ref. [79], 3C‐SiC Sample B in ref. [138], 3C‐SiC Sample S2 in ref. [135], g‐C₃N₄ prepared under 573 °C substrate temperature in ref. [142], and thin‐film NiO in ref. [113], respectively. The highlighted light orange, light yellow, and light red‐brown thick solid lines correspond to extrapolated and/or complemented experimental data sets in order to build dielectric models in the entire 280 to 600 nm range. The dashed and dash‐dot lines in (a) and (b) are analytical curves fitted to the experimental and DFT data sets using the Lorentz model (B1) under the parameters specified in Tables  B1 , B2 , B3 for 3C‐SiC, in Table  B4 for g‐C₃N₄, and in Table  B5 for NiO.

Figure C1

A light harvesting comparison of lightly doped (Model‐L), moderately doped (Model‐M), and highly doped (Model‐H) 3C‐SiC (slanted) parabolic pore PCs of lattice constant 0.8 µm, SiO₂ rear passivation layer thickness 50 nm and no front coating layer. a) Absorptivity and reflectivity spectra for spbPore PC with slanted pore depth h = 1 µm, slant angle θ = 1.75°, and 3C‐SiC thickness d _PC = 30 µm. b) MAPD of spbPore PC with pore depth h = 1 µm and slant angle θ = 0°.

Figure F1

The model of the square lattice slanted parabolic‐pore PC. a) A normal parabolic‐pore with apex P _A (vertical coordinate z _apx), base radius R and height h. b) The slanted parabolic‐pore obtained by rotating a normal parabolic‐pore by angle of θ in the x‐z plane around y axis with respect to a point P ₀. h and h′ are the pore depths of the normal and slanted parabolic pore, respectively.

Figure G1

Convergence analysis of FDTD simulations. a) The MAPD as a function of the lattice constant a for various FDTD mesh resolutions for the simulation of the light harvesting of normal parabolic‐pore PC with d _PC = 5 µm‐thick 3C‐SiC, $d_{\mathrm{coat}}=100$ nm‐thick g‐CN coating, h = 1 µm pore depth and no rear passivation layer. b) The implied MAPD and E _x decay factor as functions of physical light propagation time. The simulated structure is a normal parabolic‐pore PC of 0.8 µm lattice constant, on a highly reflective substrate in aqueous environment, with 30 µm‐thick 3C‐SiC (dielectric Model‐L), 1 µm pore depth, 50 nm SiO₂ rear passivation and no surface coating.