Electrostatic Asymmetry of Single-Crystal Nanostructures and Their Photocatalytic Properties

Efficient light absorption and high energy of charge carriers of zinc cadmium sulfide (ZCS) make this semiconductor attractive for many photocatalytic reactions. Despite marked successes in shape-controlled synthesis of ZCS central to their photocatalytic performance, recombination of charge carriers as they migrate through the nanoscale particles result in losses of excitation energy markedly reducing photocatalytic activity of ZCS and other heterogeneous photocatalysts. Here we show that electrostatic asymmetry of single-crystalline ZCS with planar geometry assists charge separation and substantially increase the yield of photocatalytic reactions. The synthesized ZCS nanorods and nanoplates with identical chemical composition were found to have markedly different photocatalytic activity for evolution of hydrogen in water. Despite much smaller specific surface areas, the ~500 nm wide nanoplates displayed hydrogen evolution rate 12 times higher than the ~35 nm long nanorods also outperforming other ZCS photocatalysts. Experimental and computational data indicate that the homo and heterojunction-free ZCS nanoplates with continuous wurtzite lattice behave essentially as nanoscale dipoles. Electric-field-directed migration of charge carriers stimulates their localization on opposite parts of the nanoplates. Direct imaging of intraparticle electrical field using off-axis electron holography confirmed their electrostatic asymmetry. Polarizationenhanced charge separation provides a new pathway to efficient and stable photocatalysts for sustainable energy technologies.

Albeit using largely Edissonian methods, previous studies on CdS give important clues about preferential nanoscale geometry for photocatalytic water splitting. The rods and nanoplates (NPLs) with {001} facets serving as the primary catalytic plane displayed PA higher than the near-spherical CdS. [20][21][22]  Nanostructures with continuous crystallinity and intrinsic static electric field would represent a nearly ideal case for efficient carrier separation in photocatalysts. [23] The discovery of multiple nanostructures with electrostatic asymmetry [24][25][26][27][28][29][30][31] opens a conceptual possibility to enhance the charge separation without adding the charge transport barriers. Here, we show that such possibility can be realized and utilized for monocrystalline ZCS with planar geometries.
We synthesized ZCS NPLs and nanorods (NRs) with identical chemical composition and evaluated their PA in hydrogen evolution from water. The flat 'tops' and 'bottoms' of the NPLs with continuous wurtzite single-crystal lattice are formed by {001} facets. The {001} facets in wurtzite crystal lattice are asymmetric with one terminated with Zn/Cd atoms and the opposite plane terminated with S atoms. [26] The facet with an excess of metal atoms carries a positive charge, while the facet with an excess of sulfur atoms carries a negative charge. This intrinsic structural asymmetry of nanoscale particles [26,[29][30][31] without any other structural engineering of the nanoparticles results in the polarization field in their interior that promotes electron hole separation and their directed migration toward the opposite interfaces, which we utilized here to drastically increase H2 generation.

From Nanoparticle to Nanorods and Nanoplatelets.
The Zn0.5Cd0.5S NRs were prepared via a solvothermal growth at 160 °C. Scanning electron microscopy (SEM) images (Figure 1a and Figure S1, Supporting Information) and transmission electron microscopy (TEM) bright-field (BF) images ( Figure S2, Supporting Information) show that the synthesized NRs are 3-6 nm in width and 30-40 nm in length. When the temperature is raised to 220 °C, the products are dominated by NPLs (Figure 1b In order to elucidate the mechanism of the NR→ NPL transition, TEM BF images of the reaction mixture for different times at 220 °C were taken. The dispersion consisted mainly of NPLs for the reaction time of 72 h ( Figure S5, Supporting Information). The transition from NRs to NPLs proceeds via non-classical crystallization pathways [31][32][33] with platelets of about 5 nm in diameter serving as intermediates (Figure 1d- Therefore, the most likely mechanism for the observed reconstruction involves restacking the nanoplatelets in a process of oriented attachment [31,32] .

Photocatalytic Evolution of Hydrogen.
Homojunctions found in the NRs (Figure 2a) were reported to facilitate photo-generated charge separation, [16,17] and, thus, the NRs were expected to show much higher PA than NPLs. This, however, was not the case. H2 production from aqueous Na2S/Na2SO3 solution under visible light illumination (λ ≥ 420 nm) was substantially higher for the NPLs (Figure 3a and Figure   S9, Supporting Information). Large difference in PA can be also observed by the naked eye: the NPLs produced a large amount of gas bubbles during the illumination, while H2 evolution in case of the NRs was sluggish (Movie S1, Supporting Information).
The great PA difference between the NPLs and the NRs, is particularly surprising considering that the NPLs have ca 4x smaller specific surface area than the NRs. The surface area per gram of the nanomaterial for the NRs and the NPLs is 206 m 2 •g −1 and 52 m 2 •g −1 , respectively, as measured from Brunauer-Emmett-Teller (BET) adsorption-desorption isotherms (Figure 3b). The average pore size is 10 nm for NRs while they range from 20 to 70 nm for the NPLs. It should be noted that the average pore in our case refer to the gap between particles, the smaller average pore size indicates that the aggregation of NRs is more serious than NPLs, which is consistent as we observed in SEM images ( Figure 1a). Nevertheless, the NRs still have larger BET surface than that of the NPLs.
The apparent quantum yields (AQY) of H2 generation for the NPLs were 8.7% and 5.8% at 435 nm and 450 nm, respectively ( Figure S10, Supporting Information). The highest H2 production rate for the NPLs was 671 μmol•h −1 •mg −1 , which is 12 times higher than that of the NRs (55 μmol•h −1 •mg −1 ). It also exceeded numerous other nanostructures with comparable composition reported previously with H2 production rate for the NPLs surpassing ZCS catalysts modified with noble metals (Table S1, Supporting Information). What is also remarkable is that the NPLs maintain a high H2 production rate of 554 μmol•h −1 •mg −1 even when the concentrations of Na2S and Na2SO3 are as low as 0.05 M (Figure 3c), which is in stark contrast with other II-VI photocatalysts. [34] Furthermore, the dispersion of NPLs retained 67% of its catalytic activity after 102 hours of illumination highlighting high resistance of NPLs to photocorrosion (Figure 3d). It should be noted here that 67% is a conservative estimate considering that a small part of catalyst will inevitably be lost via the splashed bubble during each process of vacuum pumping.

Optical Properties.
We initially hypothesized that dramatically increase of PA for NPLs vs NRs is related to the attributed to an increase of the characteristic size of the NPLs nanocrystals, [35] which can be confirmed from the Mott-Schottky plots and ultraviolet photoelectron spectra. The conduction band (CB) minima of the NPLs deduced from the Mott-Schottky plots was found to be -0.76 V, while that of NRs was -1.02 V (vs. SCE) ( Figure S12a-b, Supporting Information). Although the CB energy of the NPLs is reduced by ~0.26 V compared to the NRs, it remains, nevertheless, thermodynamically sufficient for water reduction into hydrogen. [36] The valence band (VB) maxima deduced from ultraviolet photoelectron spectra (Figure S12c-d, Supporting Information) are 0.89 and 0.93 eV for the NRs and the NPLs, respectively. Both of them are also sufficient for redox reaction with hole scavengers.
Looking further into the behavior of charge carriers, photoluminescence (PL) spectra were collected from both the NPLs and the NRs ethanol suspension. The band-edge emission (BE) is observed in 450 nm regions and is quite weak for both NPLs and NRs, which can be attributed to the fast electron-hole separation in both cases. Unlike the NPLs, the NRs exhibit a strong defect emission (DE) peak at 572 nm, [37] (Figure 4g). The lack of defect emission for the NPLs is consistent with its crystal structure homogeneity.
Furthermore, when the dispersion solvent is replaced by water, aqueous solution of NaOH or Na2S/Na2SO3, the intensity of DE from the NRs strongly decreased (Figure 4h and Figure   S13, Supporting Information), which is consistent with the fact that the PA of CdS NRs highly depends on concentration of hydroxyl ions and scavengers [38,39] . On the contrary, the emission intensity of the NPLs have almost no changes both at BE and DE regions when the solvent were changed, indicating the NPLs can maintain fast electron-hole separation in different concentration solvent, which is consistent with the NPLs can maintain a high PA at low concentrations of Na2S and Na2SO3.

Electrostatic Asymmetry of the NPLs.
Comparative measurements of photocurrents generated by films of the NRs and the NPLs deposited on fluorine doped tin oxide (FTO) glass were carried out to characterize how many photoelectrons can successfully reach the particle surface and to better understand the mechanism of high PA. The photocurrent response of the NRs is slow and weak, reaching the maximum of 0.18 μA•cm −1 with a rise time of 20 seconds (Figure 5a and b To understand better the reasons behind more efficient charge separation at the NPL/FTO than the NR/FTO interfaces, we proposed a mechanism in Figure 5. The NPLs and the NRs are oriented with their long axes parallel to the surface (Figure 5 c and d), which is reasonable as we can see from the SEM images that most of NPLs are oriented with their long axes parallel to the silicon wafer surface (Figure S3, Supporting Information), and as is common for nanocolloids with geometrical asymmetries. [40][41][42] Note, however, that seemingly similar orientation has markedly different consequences for the two types of nanostructures. Owning to the asymmetry of the hexagonal crystal lattice, [43] the (001) Table 1). The small DM of 0.0881 D per unit cell arising from the bond polarization is also parallel to the [001] direction, which is in agreement with semi-empirical calculations. [44,45] The largest facets of the NPLs producing its 'top' and 'bottom' surfaces are {001}, while the NRs have a growth direction of [001]. So, the internal polarization field is perpendicular to the NPL/FTO interface but parallel to the NRs/FTO interface (Figure 5e and f). Therefore, the photo-generated electrons tend to move to (001) terminated surface, i.e., the 'top' and 'bottom' surface for the NPLs while the two tips for the NRs. Considering the roughness of the electrode surface, the polarized part NPLs with high density of charge carriers will always find contacts with electrodes (inserts in Figure 5c and d). Consequently, the photo-generated charges can be efficiently transferred from the NPLs to the FTO electrodes. Besides some geometrical disadvantage with charge transfer to electrode that could be small, the electrostatic asymmetry of NRs is substantially reduced by the presence of several ZB segments. When it is large enough to assist the charge separation, the intrinsic polarization drives the charges toward the ends of the NRs rather than toward FTO interface as for NPLs.
The direct confirmation of electrostatic asymmetry of NPLs was obtained by off-axis electron holography (EH) that enables profiling electrostatic potential through the particle with nanometer-resolution (Figure 6). [46,47] To examine the presence of the static potential gradient in NPLs, we embedded the NPLs in the carbon film on Si substrate and then cross-sectioned them by focused ion beam milling. The carbon film mitigates unwanted charging due to the screening of local electrostatic polarization of the NPLs and induced by electron-beam. The phase shift map of NPLs after thickness correction (details in Supporting Information, Figure   S14) shows considerable asymmetry (Figure 6a). The line profile of phase shift ( (x)  ) along the direction of the short axis of the NPLs (Figure 6b) shows vivid increase of phase shift (Δ (x)  = 0.4 rad) for the NPL region (between spots 2 and 3, Figure 6c). The phase gradient in this region is linearly fitted with a slope of 0.002 rad/nm, which corresponds to electric field of 0.01 V/nm pointing from the Zn/Cd-terminated plane (spot 3) to the S-terminated plane (the spot 2). The electrostatic asymmetry in NRs along their longitudinal axis were also directly observed by EH ( Supporting Information, Figure S15). The set of EH data confirms that internal polarization field is perpendicular to the NPL/FTO interface but parallel to the NRs/FTO interface.

Conclusions
Single-crystal ZCS NPLs with dominant {001} facets have been prepared via a solvothermal method. The intrinsic polarization and ability of NPLs to self-assemble at the interfaces dramatically enhance the charge separation efficiency leading to markedly enhanced photocurrent. These findings make possible a new structural design approach for highly efficient photocatalysts that can be extended to many photocatalytic systems and photocurrent devices requiring efficient charge separation and transport.

Experimental
Synthesis of Zn0.5Cd0.5S: All raw materials and reagents in this work were analytical grade and used without further purification. In a typical synthesis, Zn0.5Cd0.5S, 0.5 mmol zinc acetate  is simplified as the following: where t(x) is the local TEM sample thickness.
Finally, the VCP(x) can be obtained by rearranging the Eq. 2: High-angle annular dark-field (HAADF)-STEM has been performed to evaluate the thickness t(x) of the same sample using JEOL ARM 200CF microscope equipped with a cold field-emission gun and spherical aberration corrector operated at 200 kV ( Figure S12). Since the intensity of HAADF signal for same material is proportional to t, the HAADF signal in each material region show the tendency of thickness variation for each region. The linear fitted slopes of three main parts, silicon wafer, ZCS NPL, and platinum are -0.023, -0.036, and -0.088 nm -1 , respectively (Figure 12c). This is typical t variation in cross-section TEM samples prepared by FIB, of which t(x) decreases as close to the start point of milling process (platinum layer) ( Figure S12d) However, the Δ showed different trend between the silicon and the NPL: Δ (-0.0090 rad/nm) < Δ (-0.0062 rad/nm) (Figure S12e).
In order to remove the thickness variation effect and extract the VCP(x) from the phase shift map (Figure S12b), we flattened the phase shift slope in silicon wafer region ( Figure S5, note the slope for silicon: 0 rad/nm). This flattening process assumes a linear thickness variation t(x) for both the silicon and NPL region, solely leaving the VCP(x) with an offset of the local mean inner potential VMIP(x) in the Eq. 3. Note that the nonlinear thickness variation of platinum layer is not completely corrected (as shown negative slope in Fig. 6c for the region) because we did "flattening" with respect to silicon, of which thickness variation is linear. After this flattening process, if we assume that NPL has no electric field, we can expect the flat phaseshift from NPL as the black dashed line in Fig. 6c. However, the slope in NPL increases, which opposes to the phase shift variation caused by thickness reduction, which represents that this increasement can be only caused by charge potential and it could be even slightly underestimated due to imperfect thickness correction. The offset correction was also made by setting vacuum to zero, corresponding to zero phase shift. There was no noticeable phase shift that present a space charge region near the Si edge, otherwise showing a quadratic change in the phase shift. The edges of NPL region shown in phase shift (Figure 6c and Figure S12e) profiles are not sharp because we made line profile with integration width of 20 pixels. Due to the electric field of NPL, the phase shift of the carbon region on platinum side (between lines 3 and 4) is higher than the phase shift of the carbon on silicon wafer side (between lines 1 an 2) (Figure 6c). In order to quantitatively measure the local sample thickness t(x), electron energy loss spectroscopy (EELS) was performed using JEOL 3100R05 STEM operated at 300 keV for the same sample. By importing electron mean-free path (133.5 nm) and absolute thickness of specimen obtained by EELS, the phase shift gradient (0.002rad/nm) has been converted into electrostatic potential gradient (0.01V/nm).
Photocatalytic activity measurement: Photocatalytic H2 production experiments were carried out in a top-irradiation vessel connected to a closed gas-circulation system. In a typical run, 5 mg of the photocatalysts were dispersed by ultrasonic treatment in 20 mL deionized water and then 4 mL of samples (1 mg) were pipetted into 100 mL of aqueous solution containing Na2S (0.75 M) and Na2SO3 (1.05 M) as sacrificial reagents. After degassing the system for ten minutes to remove most of the dissolved oxygen, a 300W xenon lamp (16 A), equipped with a cut-off filter (λ ≥ 420 nm, Figure S14a), was applied to execute the photocatalytic reaction.
The photocatalytic H2 evolution rate was analyzed using an GC-7806 gas chromatograph (GC, TCD detector, and Ar carrier).
The wavelength dependent apparent quantum efficiency (AQE) and stability was measured under similar photocatalytic reactions except the besides was wrapped by tin foil.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.       It should be noted that the negative slope in the Pt region is due to the rounded sample edge, which matches with estimated mean inner potential contribution. However, the positive slope in the NPL is opposite to the thickness variation, which attributes to the charge potential contribution.