Metastable Intermediates in Amorphous Titanium Oxide: A Hidden

Jul 24, 2018 - amorphous TiO2; atomic-layer deposition; Metastable intermediates; photoelectrochemical stability; surface protection. The Supporting ...
1 downloads 0 Views 6MB Size
Download PDF
Letter Cite This: Nano Lett. XXXX, XXX, XXX−XXX

pubs.acs.org/NanoLett

Metastable Intermediates in Amorphous Titanium Oxide: A Hidden Role Leading to Ultra-Stable Photoanode Protection Yanhao Yu,†,‡ Congli Sun,† Xin Yin, Jun Li, Shiyao Cao, Chengyu Zhang, Paul M. Voyles,* and Xudong Wang* Department of Materials Science and Engineering, University of Wisconsin−Madison, Madison, Wisconsin 53706, United States

Downloaded via DURHAM UNIV on July 27, 2018 at 08:17:21 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

S Supporting Information *

ABSTRACT: Metastable intermediates represent a nonequilibrium state of matter that may impose profound impacts to materials properties beyond our understandings of monolithic and equilibrium systems. Here, we report a discovery of hidden metastable intermediates in amorphous TiO2 thin films and their critical role in electrochemical damage. These intermediates have a non-bulk crystal-like structure and exhibit significantly higher electrical conductivity than both the amorphous and the crystalline phases. When these TiO2 films are applied to protect Si photoelectrochemical (PEC) photoanodes, the intermediates can induce localized high electrical currents that largely accelerate the etching of the TiO2 film and the Si electrode underneath. The intermediates can be effectively suppressed by raising their nucleation barrier via reducing the film thickness from 24 to 2.5 nm. The homogeneous amorphous TiO2-film-coated Si photoanodes achieved more than 500 h of PEC water oxidation at a steady photocurrent density of over 30 mA·cm−2. KEYWORDS: Metastable intermediates, amorphous TiO2, atomic-layer deposition, surface protection, photoelectrochemical stability

M

solar cells,20,21 and photoelectrochemical (PEC) electrodes.22,23 They have been used to dramatically improve device performance by stabilizing the high-energy material surfaces, facilitating interfacial charge and mass transfer and reactivity, and promoting orders of magnitude longer device lifetime. These amorphous ALD thin films are generally considered as homogeneous systems.24 Control of their performance was typically implemented via classic routes of thickness and crystallinity variation. Nevertheless, unexpected performance changes were often discovered from those systems beyond conventional understandings of homogeneous coating. One representative example is the use of ALD amorphous TiO2 thin film to protect silicon PEC electrodes for solar fuel generation. A wide variety of performance was obtained from ALD TiO2 films when the precursor-, growth temperature, and thickness were changed,25−28 while the underlying mechanism is still a mystery. Based on our recent observations of multiple intermediates present in TiO2 nanorod crystallization,29 we envision that the non-equilibrium crystallization paths and the metastable intermediates might be essential but overlooked elements in the synthesis and property control of amorphous ALD oxide thin films. In this paper, we uncovered the existence of highly conductive intermediates in amorphous ALD TiO2 thin films. These intermediates were mesoscopic structural inhomogeneity imbedded in the amorphous phase and showed

aterials science studies structure−property relationships generally in a thermodynamic equilibrium system. Facing the rapid evolution of new materials and functionalities, monolithic and equilibrium systems soon become insufficient in addressing fundamental and applied scientific questions in modern functional-material development. Beyond equilibrium matters, kinetically trapped states emerge as the new foci of materials are discovered, which open vast opportunities in enabling new and enhanced properties that are not possible in equilibrium systems.1 Powerful electron microscopy and theoretical studies recently revealed the existence of metastable intermediates in a number of material systems, such as oxides, metals, and polymers.2−4 These intermediates (e.g., atom clusters, amorphous particles, and non-bulk crystal-like structures) are kinetically trapped structures between the amorphous and the crystalline phases.5,6 They exhibit distinct physical and chemical properties compared with their thermodynamically stable counterparts due to their unique local bonding and short- or medium-range order.7,8 A small amount of intermediates in an amorphous or crystalline matrix may bring large changes to the film properties. Their presence can lead to substantial performance gain in many application areas including ion conductors, pharmaceuticals, and highentropy alloys.9−11 Among many functional materials, amorphous metal oxides, particularly the thin-film morphology synthesized by atomic layer deposition (ALD), are critical components in many modern energy systems, such as batteries,12,13 catalysts,14−19 © XXXX American Chemical Society

Received: June 24, 2018 Published: July 24, 2018 A

DOI: 10.1021/acs.nanolett.8b02559 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 1. Structure and electron orbital features of metastable intermediates in amorphous TiO2 thin film. (a) Top-view SEM image of the 24 nm ALD grown amorphous TiO2 thin film on Si substrate. (b) 3D AFM topography of the TiO2 film shown in panel a. The white dashed box highlights a broad bump area. (c, d) Cross-sectional ABF STEM images featuring two distinctive domains inside the TiO2 layer. One was completely imbedded in the film (shown in panel c), and the other one induced a surface bump (shown in panel d). (e) STEM EELS fine structure of Ti L2, L3 edge for a series of structural features in the amorphous TiO2 thin film and single crystalline anatase TiO2. Partial splitting of the Ti peaks were discovered from both intermediates, as marked by the red dashed lines. (f) STEM EELS fine structure of O K of different TiO2 structures. Intermediates exhibit earlier edge onset and increased onset intensity compared to the amorphous portion, indicating a reduced bandgap energy and increased density of states at the band edge. (g) CBED of different structural features inside amorphous TiO2 film and the anatase TiO2 crystal, confirming the non-bulk crystal-like structure of the intermediates.

To further reveal the morphology and crystallinity information inside the particles and bumps, the sample was cut into thin slices using focused ion beams (FIBs). The structural and electron orbital details of the TiO2 film were then studied by scanning transmission electron microscopy (STEM), electron energy loss spectroscopy (EELS) fine structure, and convergent-beam electron diffraction (CBED). The particles observed by SEM and AFM (Figure 1a,b) were found crystalline under the cross-sectional annular bright-field (ABF) STEM characterization (Figure S1). Inside the crystalline particles, two lattice distances of 0.24 and 0.38 nm can be identified from the fast Fourier transform (FFT) image (Figure S2), matching well to the (004) and (011) planes of anatase TiO2, respectively. In addition to the crystallized particles, the cross-sectional ABF STEM images featured two distinctive domains, in which one was completely embedded in the film and the other one induced a surface bump (Figure 1c,d). Different from the crystalline particles shown in Figure S1, no lattice fringes could be detected from these domains. The different contrast in the ABF images suggests that the domains may have different O concentrations or possess different structural characteristics compared with the amorphous surroundings. Because EELS quantification did not show any detectable difference on Ti-to-O ratios between the domain and surrounding area (Figure S3), the ABF contrast could be attributed to different structural configurations. Thus, the domains are denoted as intermediates I and II.

drastically different electronic property compared with the amorphous matrix and crystalline particles. When the amorphous TiO2 thin film was used as the protective coating of Si PEC electrodes, the intermediates were identified as the trigger of the damage spots in the film. Raising the nucleation energy barrier by reducing the film thickness was found able to effectively suppress the formation of these intermediates, which significantly promoted the electrode longevity in a corrosive PEC environment. Amorphous TiO2 thin films with a thickness of ∼24 nm were deposited on silicon (Si) substrates using ALD (see the Methods section in the Supporting Information for details). The top-view scanning electron microscopy (SEM) image of the TiO2 thin film shows some randomly distributed particlelike features (Figure 1a). According to three-dimensional (3D) atomic force microscopy (AFM) tomography (Figure 1b), these particles were embedded in the amorphous matrix, and they were 40−86 nm in size and ∼10 nm tall. These particles are common features for a thick amorphous ALD thin films due to the high bulk free energy of amorphous oxides compared with their crystalline counterparts.30,31 While most of the other area was extremely flat (the amorphous region, ∼ 0.25 nm in roughness), a broader bump (∼110 nm × 80 nm) with a height of only ∼4 nm was discovered in the topography image (white dashed box in Figure 1b). These observations suggest the existence of substantial structural heterogeneity in this amorphous TiO2 thin film. B

DOI: 10.1021/acs.nanolett.8b02559 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 2. Electrical property of metastable intermediates in amorphous TiO2 thin film. (a) AFM topography image of the TiO2 surface. The bright dots refer to the crystalline particles. The white dashed box marks a bump domain (∼110 nm × ∼80 nm) with a height of ∼4 nm. (b) AFM current mapping of the film recorded under a bias of −3 V. The white and yellow dashed boxes mark the highly conductive regions. (c, d,) Topography line profiles (green and blue) and current line profiles (magenta and red) across the (c) crystalline particle and (d) conductive domains. (e) The local I−V characteristics recorded at the different features marked in panel b.

The EELS fine structures of Ti L2,3-edges and O K-edges for amorphous, intermediate I, intermediate II, and crystalline (shown in Figure S1) regions were collected and compared in Figure 1e,f. As a reference, spectra from a single crystalline anatase TiO2 (i.e., the common crystalline phase of ALD TiO2) sample were also included. The Ti L2 and L3 edges of the amorphous region (black dot in Figure 1c) located at 465.3 and 460.1 eV, respectively. Upon the transition from amorphous to crystalline, the Ti L2 and L3 edges gradually split from one peak to two peaks with a maximum splitting energy of 1.6 eV for the anatase TiO2. This splitting is a common character for TiO2 phase transition, and was believed to be a result of crystallization-induced free energy reduction.32,33 The splitting energy of intermediate I was only 0.8 eV, while intermediate II had a relatively larger splitting energy of 1.3 eV, approaching the anatase TiO2 case. The distinctive splitting energies for intermediate I and II placed them into different crystallization states with different local energy minimums. Figure 1f shows the O K-edges integrated from the spectrum image. O K-edge spectrum can be interpreted as the unoccupied O p-density of states from the first-order

approximation. The O K-edge onsets of amorphous, intermediate I, intermediate II, particle and anatase TiO2 located at 529.4, 527.6, 528.1, 528.1, and 528.6 eV, respectively. Compared to the amorphous area, the O Kedges of the intermediates exhibited a slightly early edge onset with increasing onset intensities. Because the O K-edge reflects the portion of the conduction band projected onto O atoms, the edge shift indicates a reduced band-gap energy and increased the density of states at the band edge, both of which could result in different electronic properties. For the crystalline particle, EELS detected an additional peak centered at 529.7 eV, which was an indication of the Ti3+ point defect located below the conduction-band minimum.34 The local structural features of the amorphous, intermediates I and II, and anatase TiO2 were further investigated by CBED (Figure 1g). The amorphous area gave rise to a cloudlike diffraction surrounding the central beam, confirming the random atom arrangement. The intermediate I showed a strong amorphous CBED feature but with several crystal-like diffraction spots around the central beam, confirming the existence of a certain-level structural ordering at this stage. However, the weak and asymmetric diffraction spots indicate C

DOI: 10.1021/acs.nanolett.8b02559 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 3. Intermediate-induced failure of the amorphous TiO2-protected Si electrode. (a) Cyclic voltammetry (CV) and (b) chronoamperometry of the Si/TiO2/Ni electrode measured in 1.0 M NaOH aqueous solution under 1 sun of illumination. (c) Top-view SEM images of the Si/TiO2/Ni surface after 80 h of PEC reaction. Inset amplifies the circular-shaped damages and identifies a small hole at the center of the dark circle, as marked by the red dashed circles. (d, e) Cross-sectional SEM images of the Si/TiO2/Ni electrode after 80 h of PEC reactions at the locations with (d) a newly formed and (e) an expanded Si etching area. (f−h) Cross-sectional ABF STEM images of the Si/TiO2/Ni electrode after 40 h of PEC reaction at (f) an untacked area, (g) a crystalline particle, and (h) an initially damaged spot. (i−k) Corresponding high-resolution STEM images of the marked regions in panels f−h (yellow dashed squares). The white dashed line in panel j highlights a crystalline particle with clear lattice fringes. The yellow dashed line and red arrow in panel k pointed out the contrast boundary of a slightly darker feature surrounding the etching spot. (l) Schematic of a possible failing mechanism of the amorphous TiO2 film due to structural heterogeneity and high electrical conductivity of the intermediates.

conductivity of the TiO2 thin film was probed by conductive AFM (c-AFM). Similar to the 3D AFM topography (Figure 1b), the top-view AFM topography also identified the isolated crystalline particles (bright spots in Figure 2a) and the broad bump (white dashed box in Figure 2a). The size and geometry of this bump matched well with that of the intermediate observed in Figure 1d, suggesting that the bump area was probably one of the intermediate states. From the c-AFM mapping (Figure 2b), the crystalline particles only exhibited a slightly higher electrical conductivity compared with the amorphous regions, whereas an exceptionally higher conductivity was detected in the bump area (white dashed square in Figure 2b). Line profiles of topography signal matched well with the current profiles for both the sharp crystalline particle and the broad bump domain, confirming the strong correlation between morphology and conductivity features (Figure 2c,d). It is important to note that high conductivity was also recorded at two places with no topography features (yellow dashed square and the upper left portion of the white dashed square in Figure 2b), suggesting the broad bump-related structure could be completely imbedded in the amorphous film and the

that the local structure was probably disordered atomic planes without long-range periodicity. The intermediate II had fewer amorphous features and contained more crystal-like diffraction spots, suggesting a more-developed structure ordering. The plane distances for two main diffraction spots near the central beam were 0.192 and 0.129 nm, which were close to that of the (200) and (107) planes of anatase TiO2 (0. 189 and 0.128 nm, respectively, JCPDS card no. 01-071-1166). However, no zone-axis CBED pattern of single-crystal anatase TiO2 can be matched, meaning that intermediate II was probably composed of multiple crystalline clusters without preferred orientations. For single-crystal anatase TiO2, typical zone-axis [110] CBED patterns with no amorphous features were recorded. The CBED characterizations clearly identified the non-bulk crystallike structure of the intermediates, evidencing the presence of kinetically trapped intermediate states during the deposition of amorphous TiO2 thin films. The EELS characterizations indicate different electronic properties in the intermediate regions compared with the amorphous and crystalline regions and, thus, an inhomogeneous film property. To verify this speculation, electrical D

DOI: 10.1021/acs.nanolett.8b02559 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

and spread over the entire electrode surfaces after 80 h of water oxidation (Figure 3c). A small hole can be observed at the center of the dark circular feature (see the inset of Figure 3c). The underneath morphology of the circular damage was revealed by cross-sectional SEM observations. A bowl-like corrosive region in Si and a suspended TiO2 thin film with a wedged cavity presented at different stages of etching (Figures 3d,e and S7). The thickness of the suspending layer was measured to be 26 nm, close to the as-prepared TiO2 layer thickness (24 nm in Figure S4). The orifice of the cavity was 42 nm, which was remained at the similar size when the Si cavities underwent different corrosion stages, e.g., 93 nm (Figure 3d), 172 (Figure 3e), and 272 nm (Figure S7a). Interconnected Si cavities were observed in heavily damaged locations (Figure S7b). Due to the physical separation of photoactive Si and TiO2/Ni catalysts, the PEC performance could be drastically impaired in these destructed areas.25 After visualization of the damage geometry, the key question to understand the possible intermediate correlation is how and where these damages were initiated. We tackled this question by atomic scale investigation of the cross-sections of the Si/ TiO2/Ni electrode at different damage evolution stages using STEM. The initial stage of damage was imaged from a device after 40-h PEC operation. First of all, no physical pinholes were detected along the entire observed interface area (up to 25 μm), statistically suggesting that the observed orifices were created during the electrochemical reaction rather than formed during the ALD deposition of TiO2.41 As shown in Figure 3f,i, most areas of the electrode exhibited a well-preserved layered structure. A small interfacial layer with a thickness of ∼0.5 nm was observed between Si and TiO2, which was typically believed to be SiOx.42 Crystalline TiO2 particles could be observed being partially imbedded in the amorphous film (Figures 3g,j and S1), which corresponded to the surface particulate features (Figure 1a,b). Underneath the particle, the thickness of interfacial layer was subtly increased to ∼2 nm. These observations revealed that the electrode heterostructure under the crystalline region were well maintained. Therefore, the crystalline particles could tolerate the corrosive electrolyte as good as the homogeneous amorphous region. Surprisingly, a large interfacial separation (up to 5 nm) between Si and TiO2 was observed under a slightly darker region that share the same ABF STEM feature of intermediate I (as marked by the yellow dashed line and red arrow in Figure 3h,k), suggesting a strong correlation with the metastable intermediates. STEM EELS mapping revealed that Ti was deficient at both the top and the bottom regions of the film (Figure S8). Thus, the TiO2 film here exhibited a two-way etching tendency from the top and bottom interfaces toward the center. O was enriched at the bottom region, while the rest of the film rendered a homogeneous O distribution. The large interfacial morphology change and the nonstoichiometric Tito-O ratio indicate this area was likely the spot where the film damage began. Directed by above experimental discoveries, we suggest that the damage to the Si/TiO2/Ni photoanode was prone to occur at the highly conductive intermediates in TiO2 thin films, and we propose a possible mechanism as schematically illustrated in Figure 3l. The photoactive Si absorbed photons and produced electrons and holes upon light illumination. The build-in potential at Si/TiO2/electrolyte interface drove hole migration through the TiO2 protection layer, oxidizing the

conductivity inhomogeneity should not simply be a result of morphological fluctuation. The specific current−voltage (I−V) characteristics for different features were compared by sweeping the bias from 3 to −3 V (Figure 2e). A rectifying curve was obtained due to the metal-oxide-semiconductor (MOS) junction between Si, TiO2 and the Pt AFM tip. At −3 V, current obtained from the amorphous, crystalline, and bump regions (i.e., intermediates) were 0.08, 0.20, and 0.83 nA, respectively, highlighting the drastic electrical conductivity enhancement by the intermediates. Averagely, the discrepancy between intermediates (2.8 nA) and the other areas (0.1 nA for amorphous matrix and 0.2 nA for crystalline particles) were more conspicuous (Table S1). The high conductivity detected at intermediates could be ascribed to the increased density of states of the intermediates (Figure 1f) and the trap-assisted tunneling.35,36 The discovery of metastable intermediates and their distinct property offers a new clue to the unknown performance variation of amorphous ALD TiO2 thin films (for instance, the protection of Si electrodes for PEC water splitting). To study the role of intermediates in photoelectrode protection, ALD TiO2 thin film-protected Si photoanodes were investigated extensively. The electrode was composed of a n-Si wafer as the light absorber, a 24 nm ALD TiO2 as the protective layer and a 12 nm Ni metal as the catalyst for the oxygen evolution reaction (denoted as Si/TiO2/Ni; see the Methods section in the Supporting Information and Figure S4 for details). Ni catalyst is inevitable because the oxygen evolution kinetics of Si and TiO2 are sluggish (Figure S5). The influence of Ni layer is minimized by keeping the identical deposition condition and the film property for all samples. The PEC water splitting performance was evaluated in 1 M NaOH aqueous solution under 1 Sun illumination, and the corresponding cyclic voltammetry (CV) curves at varied reaction time are shown in Figure 3a. The first curve was recorded after 50 cycles (about 1 h) of Ni catalyst activation. The redox peaks located between 1.0 and 1.1 V versus reversible hydrogen electrode (RHE) were the Ni(OH)2 and NiOOH oxidation and reduction peaks.37 The water oxidation initiated at 1.22 V versus RHE and approached the saturation photocurrent density of 30.0 mA·cm−2 at 1.65 V versus RHE. The water oxidation current after 50 h still showed a similar trace as the initial curve. When the reaction proceeded to 80 h, the photocurrent density decreased to 24.9 mA·cm−2 at 1.65 V versus RHE, and the onset potential slightly shifted toward the negative potential direction. Accordingly, the chronoamperometry test at an external bias of 1.8 V versus RHE depicted a stable photocurrent density at 30.0 mA·cm−2 for the first 70 h and then gradually diminished, implying the beginning of electrode damage (Figure 3b). The onset potential and photocurrent density of this device were in the same level as other reported Si-based single junction PEC anodes,38,39 confirming that this device operated at its full capacity. Electron microscopy characterizations were then conducted to uncover the mechanism of the photocurrent decrease. Sparsely distributed circular-shaped damages with diameters of up to 660 nm were found after 40 h of PEC testing (Figure S6a). The circular geometry implies isotropic corrosion, which was similar to the etching situation of Si covered with Ni catalysts.40 The energy dispersive X-ray spectroscopy (EDS) detected slightly reduced Ti signal in the damaged region, implying the TiO2 thin film was preserved (Figure S6b). The number and size of the circular damages significantly increased E

DOI: 10.1021/acs.nanolett.8b02559 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 4. Improve the stability by suppressing the formation of intermediates via thickness reduction. (a) Cross-sectional ABF STEM images of a Si/TiO2/Ni electrode with a 2.5 nm TiO2 film, while the Ni film was remained at 12 nm. (b) High-resolution ABF image at the interface showing uniform contrast and no intermediate features were observed. (c) AFM topography image of the 2.5 nm TiO2 thin film showing an extremely flat surface. (d) C-AFM current mapping of the same TiO2 area, revealing a homogeneous current distribution. (e) CV curves of the 2.5 nm TiO2protected Si electrode measured at a series of continuous operation time points in 1 M NaOH aqueous solution under one sun illumination. (f) Corresponding chronoamperometry curve measured in 1 M NaOH aqueous solution under one sun illumination with an external bias of 1.8 V vs RHE. The photocurrent density was maintained steadily at ∼30 mA·cm−2 for over 500 h. Insets are the top-view SEM images of the electrode surface at three representative reaction time as marked in the curve showing a very slow evolution of surface damage.

property. Therefore, a straightforward strategy to improve the protection longevity is to maximize electrical homogeneity of the amorphous thin film by suppressing the formation of intermediates. One effective kinetic principle is to confine the material in an extremely small volume, which could largely slow down the nucleation rate of a new phase as a result of reduced bulk free energy and limited space for atom rearrangement.44,45 This principle was thus implemented in the amorphous TiO2 thin films to restrict the formation of intermediates by reducing the film thickness to 2.5 nm, while all of the other ALD growth conditions remained identical (Figure 4a). Cross-sectional STEM image showed that the ultrathin TiO2 film was completely amorphous, and no phase contrast was detected (Figure 4b). AFM topography and SEM characterizations demonstrated a sub-nanometer-level smoothness with a surface roughness of ∼0.15 nm (Figure 4c and Figure S11). Electrically, c-AFM revealed a highly uniform current distribution (Figure 4d) with slightly higher value (−0.10 nA) compared to the amorphous region (−0.08 nA) in the 24 nm TiO2 film (Figure S12). Figure 4e presents the CV curves of the Si/2.5 nm TiO2/Ni PEC electrode measured in 1 M NaOH aqueous solution under 1 sun of illumination. The photocurrent of water oxidation started to take off at 1.30 V versus RHE and was saturated at 1.62 V versus RHE. It was noticed that, with the increase of reaction time, the onset potential of water oxidation was shifted to 1.21 V versus RHE. This favorable potential shift was believed to be a consequence of the spatially nonuniform barrier height and the large coverage of a high-barrier-height region, which was likely to form in the device with thin

hydroxyl group to oxygen with the help of Ni catalyst. Due to the distinct charge-transfer capabilities, photocurrent density in the intermediate region could be substantially higher than that of other areas. The excess amount of hole supply was able to attract more negative ions (e.g., OH−) from the electrolyte and thus largely increased the local concentration of OH− at the areas of intermediates (stage 1 in Figure 3l). The concentrated OH− were expected to have two consequences (stage 2 in Figure 3l). First, TiO2 would become solvable in the presence of high-density OH−, leading to the top interfacial corrosion. This fact was verified by placing 10 M NaOH aqueous droplets on the surfaces of TiO2 films with and without Ni catalysts. After 2 h, both samples were significantly corroded (Figure S9). Second, driven by the strong local concentration gradient and electric field, OH− could diffuse through the TiO2 film, resulting in the oxidation of Si and the etching of TiO2 at the Si−TiO2 interface. Upon the etching of TiO2 at the intermediate area, a pinhole was gradually developed through the TiO2 film, where the electrolyte could infuse and etch Si microscopically (stage 3 in Figure 3l). This process could be visualized by a series of cross-sectional electron microscope images from the initial etching of TiO2 at the interfaces to the final formation of interconnected cavities in Si (Figure S10). In situ characterizations, such as scanning tunneling microscopy (STM), could further validate this hypothesis with more direct and quantitative insights.43 The TiO2 corrosion is a dissolving process with no valence change because no additional redox peaks present in the CV curves (Figure 3a). Above studies revealed the detrimental role of intermediates in PEC electrode protection due to their distinct electronic F

DOI: 10.1021/acs.nanolett.8b02559 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters protection or catalytic films.38,46 In terms of the photocurrent density, the plateau after 1 h of PEC reaction reached 30.6 mA· cm−2, which was slightly higher than that of the 24 nm TiO2 sample. Unlike the 24 nm TiO2 protection, this saturated photocurrent density was stably maintained for a significantly longer time. No conspicuous photocurrent decrease was observed until 560 h of continuous operation (Figure 4f). This performance largely exceeded other reported results of TiO2 protected planar Si electrodes in alkaline conditions.25 The reproducibility of 2.5 nm TiO2 protected Si electrodes is shown in Table S2, where a lifetime of over 450 h was obtained from 3 samples. Very few circular-shape damages were created on 2.5 nm TiO2 electrode after 110 and 250 h PEC reactions (inset of Figure 4f), evidencing that the improvement of electrode longevity was a result of the largely retarded TiO2 layer damage. The damaged area in the 2.5 nm TiO2 film exhibited almost-identical geometry to those in the 24 nm film (Figure S13). This phenomenon supports the likelihood of 2.5 nm TiO2 underwent a similar damaging process as that of the 24 nm case. Therefore, the extreme stability of the 2.5 nm film could be attributed to the effective suppression of the detrimental metastable intermediates. For 2.5 nm TiO2, the preferred cross-sectional electron microscopy characterizations for individual etching stages are impractical due to the extremely small film thickness and the instantaneous etching process. To further confirm this conclusion, the PEC performance of a 10 nm TiO2 film-coated Si PEC electrode was tested and its stability was measured as ∼160 h, sitting between the 24 and 2.5 nm films (Figure S14a). In addition to film thickness, the deposition temperature was also changed to 120 °C, while the TiO2 film remained at 24 nm thick. The PEC stability of this sample reached ∼110 h (Figure S14b). As revealed by SEM, both samples had much fewer crystalline particles compared to the 24 nm film (Figure S14c,d). Nevertheless, c-AFM of the sample surface confirmed the presence of highly conductive intermediates with a density of ∼6 per 25 μm2 (Figure S14e), which was less than that of 160 °C TiO2 (∼17 per 25 μm2). These film thickness and temperature influences further confirmed that kinetics routes of suppressing the formation of intermediates were directly correlated to the protection performance. In addition to the control of film thickness and growth temperature, manipulating other deposition parameters (e.g., Ti precursors, pulse and purge times, and doping) could potentially further improve the protection performance by impeding the formation of intermediates. In summary, we have discovered the presence of metastable intermediates in amorphous TiO2 thin films. These intermediates had a non-bulk crystal-like structure and could induce significant inhomogeneity in the film’s electronic properties. When the film was employed for PEC electrode protection, the intermediates raised local concentration of hydroxyls and promoted the formation of pinholes through the amorphous matrix. Formation of intermediates could be effectively suppressed by confining the nucleation space to raise the nucleation energy barrier. Reducing the film thickness to 2.5 nm yielded an amorphous thin film with excellent homogeneity. The 2.5 nm protection film accomplished over 500 h electrode longevity at a photocurrent density of ∼30 mA·cm−2, largely exceeding the stability obtained from the 24 nm TiO2 protection. This discovery is contradictory to the common sense that thicker film should render better protection. Such an “abnormal” phenomenon underlines the significance of the

kinetically trapped intermediates in amorphous protection thin films. Beyond PEC photoanode protection, further understanding and control of metastable intermediates in the nanometer scale might be a critical but overlooked strategy for property and performance optimization in many electronic and energy devices that are sensitive to the charge and ion dynamics, such as transistors, catalysts, photovoltaics, and batteries.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.8b02559. Experimental details, cross-sectional STEM images and EELS elemental mappings, photocurrent−voltage curves of Si/TiO2 electrodes, SEM characterizations, digital images and a SEM image of amorphous TiO2, local I−V characteristic of the 2.5 nm TiO2, characteristics of TiO2 thin films, a table shows the average current values of different TiO2 phases, and a table showing PEC performances for three 2.5 nm TiO2-protected Si electrodes. (PDF)

■

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. ORCID

Yanhao Yu: 0000-0003-2494-3006 Xin Yin: 0000-0002-4759-0226 Jun Li: 0000-0002-7498-6736 Paul M. Voyles: 0000-0001-9438-4284 Xudong Wang: 0000-0002-9762-6792 Present Address ‡

John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, 02138, United States Author Contributions †

Y.Y. and C.S. contributed equally to this work.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work is primarily supported by U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES), under award no. DE-SC0008711. STEM characterization (C.S. and P.M.V.) was supported by the U.S. DOE (DE-FG02-08ER46547). Facilities for microscopy and spectroscopy were supported by the University of Wisconsin− Madison Materials Research Science and Engineering Center (DMR-1720415)

■

REFERENCES

(1) Sosso, G. C.; Chen, J.; Cox, S. J.; Fitzner, M.; Pedevilla, P.; Zen, A.; Michaelides, A. Chem. Rev. 2016, 116, 7078−7116. (2) Nielsen, M. H.; Aloni, S.; De Yoreo, J. J. Science 2014, 345, 1158−1162. (3) Loh, N. D.; Sen, S.; Bosman, M.; Tan, S. F.; Zhong, J.; Nijhuis, C. A.; Kral, P.; Matsudaira, P.; Mirsaidov, U. Nat. Chem. 2016, 9, 77− 82. G

DOI: 10.1021/acs.nanolett.8b02559 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters (4) Schnurr, B.; Gittes, F.; MacKintosh, F. C. Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip. Top. 2002, 65, 061904. (5) Kitchaev, D. A.; Ceder, G. Nat. Commun. 2016, 7, 13799. (6) Russo, J.; Tanaka, H. MRS Bull. 2016, 41, 369−374. (7) Sun, W.; Jayaraman, S.; Chen, W.; Persson, K. A.; Ceder, G. Proc. Natl. Acad. Sci. U. S. A. 2015, 112, 3199−3204. (8) Sun, W.; Dacek, S. T.; Ong, S. P.; Hautier, G.; Jain, A.; Richards, W. D.; Gamst, A. C.; Persson, K. A.; Ceder, G. Sci. Adv. 2016, 2, e1600225. (9) Rodriguez-Hornedo, N.; Murphy, D. J. Pharm. Sci. 1999, 88, 651−660. (10) Li, Z.; Pradeep, K. G.; Deng, Y.; Raabe, D.; Tasan, C. C. Nature 2016, 534, 227−230. (11) Baumgartner, J.; Dey, A.; Bomans, P. H.; Le Coadou, C.; Fratzl, P.; Sommerdijk, N. A.; Faivre, D. Nat. Mater. 2013, 12, 310−214. (12) Han, X.; Gong, Y.; Fu, K. K.; He, X.; Hitz, G. T.; Dai, J.; Pearse, A.; Liu, B.; Wang, H.; Rubloff, G.; Mo, Y.; Thangadurai, V.; Wachsman, E. D.; Hu, L. Nat. Mater. 2016, 16, 572−579. (13) Kozen, A. C.; Lin, C. F.; Pearse, A. J.; Schroeder, M. A.; Han, X.; Hu, L.; Lee, S. B.; Rubloff, G. W.; Noked, M. ACS Nano 2015, 9, 5884−5892. (14) Chen, X.; Liu, L.; Yu, P. Y.; Mao, S. S. Science 2011, 331, 746− 750. (15) Lu, J.; Fu, B.; Kung, M. C.; Xiao, G.; Elam, J. W.; Kung, H. H.; Stair, P. C. Science 2012, 335, 1205−1208. (16) Xia, T.; Zhang, C.; Oyler, N. A.; Chen, X. Adv. Mater. 2013, 25, 6905−6910. (17) Xia, T.; Zhang, W.; Wang, Z.; Zhang, Y.; Song, X.; Murowchick, J.; Battaglia, V.; Liu, G.; Chen, X. Nano Energy 2014, 6, 109−118. (18) O’Neill, B. J.; Jackson, D. H. K.; Lee, J.; Canlas, C.; Stair, P. C.; Marshall, C. L.; Elam, J. W.; Kuech, T. F.; Dumesic, J. A.; Huber, G. W. ACS Catal. 2015, 5, 1804−1825. (19) Yan, X.; Tian, L.; He, M.; Chen, X. Nano Lett. 2015, 15, 6015− 6021. (20) Chang, C.-Y.; Lee, K.-T.; Huang, W.-K.; Siao, H.-Y.; Chang, Y.C. Chem. Mater. 2015, 27, 5122−5130. (21) Sutherland, B. R.; Hoogland, S.; Adachi, M. M.; Kanjanaboos, P.; Wong, C. T.; McDowell, J. J.; Xu, J.; Voznyy, O.; Ning, Z.; Houtepen, A. J.; Sargent, E. H. Adv. Mater. 2015, 27, 53−58. (22) Hu, S.; Shaner, M. R.; Beardslee, J. A.; Lichterman, M.; Brunschwig, B. S.; Lewis, N. S. Science 2014, 344 (6187), 1005−1009. (23) Yu, Y.; Zhang, Z.; Yin, X.; Kvit, A.; Liao, Q.; Kang, Z.; Yan, X.; Zhang, Y.; Wang, X. Nat. Energy 2017, 2, 17045. (24) George, S. M. Chem. Rev. 2010, 110, 111−131. (25) Bae, D.; Seger, B.; Vesborg, P. C.; Hansen, O.; Chorkendorff, I. Chem. Soc. Rev. 2017, 46, 1933−1954. (26) Chen, Y. W.; Prange, J. D.; Duhnen, S.; Park, Y.; Gunji, M.; Chidsey, C. E.; McIntyre, P. C. Nat. Mater. 2011, 10, 539−544. (27) Zhou, X.; Liu, R.; Sun, K.; Papadantonakis, K. M.; Brunschwig, B. S.; Lewis, N. S. Energy Environ. Sci. 2016, 9, 892−897. (28) McDowell, M. T.; Lichterman, M. F.; Carim, A. I.; Liu, R.; Hu, S.; Brunschwig, B. S.; Lewis, N. S. ACS Appl. Mater. Interfaces 2015, 7 (28), 15189−15199. (29) Shi, J.; Li, Z.; Kvit, A.; Krylyuk, S.; Davydov, A. V.; Wang, X. Nano Lett. 2013, 13, 5727−5234. (30) Reichel, F.; Jeurgens, L. P. H.; Mittemeijer, E. J. Acta Mater. 2008, 56, 659−674. (31) Gu, J.; Yan, Y.; Young, J. L.; Steirer, K. X.; Neale, N. R.; Turner, J. A. Nat. Mater. 2016, 15, 456−460. (32) Zhang, Y.; Guerra-Nunez, C.; Utke, I.; Michler, J.; Agrawal, P.; Rossell, M. D.; Erni, R. Chem. Mater. 2017, 29, 2232−2238. (33) Bertoni, G.; Beyers, E.; Verbeeck, J.; Mertens, M.; Cool, P.; Vansant, E. F.; Van Tendeloo, G. Ultramicroscopy 2006, 106 (7), 630−635. (34) Di Valentin, C.; Pacchioni, G.; Selloni, A. J. Phys. Chem. C 2009, 113, 20543−20552. (35) DiMaria, D. J.; Cartier, E.; Arnold, D. J. Appl. Phys. 1993, 73, 3367−3384.

(36) Runnion, E. F.; Gladstone, S. M.; Scott, R. S.; Dumin, D. J.; Lie, L.; Mitros, J. C. IEEE Trans. Electron Devices 1997, 44, 993−1001. (37) Lin, F.; Boettcher, S. W. Nat. Mater. 2014, 13, 81−86. (38) Laskowski, F. A. L.; Nellist, M. R.; Venkatkarthick, R.; Boettcher, S. W. Energy Environ. Sci. 2017, 10, 570−579. (39) Ji, L.; McDaniel, M. D.; Wang, S.; Posadas, A. B.; Li, X.; Huang, H.; Lee, J. C.; Demkov, A. A.; Bard, A. J.; Ekerdt, J. G.; Yu, E. T. Nat. Nanotechnol. 2015, 10, 84−90. (40) Han, T.; Shi, Y.; Song, X.; Mio, A.; Valenti, L.; Hui, F.; Privitera, S.; Lombardo, S.; Lanza, M. J. Mater. Chem. A 2016, 4, 8053−8060. (41) Didden, A.; Hillebrand, P.; Dam, B.; van de Krol, R. Int. J. Photoenergy 2015, 2015, 1−8. (42) Yang, J.; Cooper, J. K.; Toma, F. M.; Walczak, K. A.; Favaro, M.; Beeman, J. W.; Hess, L. H.; Wang, C.; Zhu, C.; Gul, S.; Yano, J.; Kisielowski, C.; Schwartzberg, A.; Sharp, I. D. Nat. Mater. 2017, 16, 335−341. (43) Pfisterer, J. H. K.; Liang, Y.; Schneider, O.; Bandarenka, A. S. Nature 2017, 549, 74−77. (44) Amstad, E.; Gopinadhan, M.; Holtze, C.; Osuji, C. O.; Brenner, M. P.; Spaepen, F.; Weitz, D. A. Science 2015, 349, 956−960. (45) Whittaker, M. L.; Dove, P. M.; Joester, D. MRS Bull. 2016, 41, 388−392. (46) Hill, J. C.; Landers, A. T.; Switzer, J. A. Nat. Mater. 2015, 14, 1150−1155.

H

DOI: 10.1021/acs.nanolett.8b02559 Nano Lett. XXXX, XXX, XXX−XXX