Atomic scale mapping of impurities in partially reduced hollow TiO2 nanowires

The incorporation of impurities during the chemical synthesis of nanomaterials is usually uncontrolled and rarely reported because of the formidable challenge that constitutes measuring trace amounts of often light elements with sub nanometre spatial resolution. Yet these foreign elements influence functional properties, by e.g. doping. Here we demonstrate how the synthesis and partial reduction reaction on hollow TiO2 nanowires leads to the introduction of parts-per-millions of boron, sodium, and nitrogen from the reduction reaction with sodium borohydride at the surface of the TiO2 nanowire. This doping explains the presence of oxygen vacancies at the surface that enhance the activity. Our results obtained on model metal-oxide nanomaterials shed light on the general process leading to the uncontrolled incorporation of trace impurities that can have a dramatic effect on their potential use in energy-harvesting applications.

can be used when NaBH4 is involved in the reduction process, which makes this method applicable to TiO2 grown on temperature sensitive transparent glass electrodes such as fluorine-doped tin oxide (FTO). TiO2/FTO assemblies are typically used for photoelectrochemical-and solar-cell applications, and they would lose their good conductivity if exposed to higher temperature 29,30 . However, the presence and location of trace elements from the inert gas (e.g. nitrogen) or impurities from the reductant (sodium and boron) and their effects on the structure and properties of TiO2 are still unclear 31 . Basically, all these elements can introduce disorder in the TiO2 crystal lattice and can lead to a lower band-gap. N can act as an anionic dopant of TiO2 since its atomic p-orbital levels are appropriate for narrowing the otherwise wide band-gap of TiO2 38 . B in TiO2 can have a synergy effect as OTiBN bonds narrow the bandgap and compensate the excess charge from N doping 50,51 . Interstitial B has been suggested to stabilize Ti 3+ species N6 . The role of Na for the TiO2 conductivity seems not clear yet, although Na can act as a recombination center or limiting the crystallization of TiO2 47 . Besides being incorporated in the lattice, foreign elements from the synthesis can also be located at the very surface, either as clusters or individual atoms, which is in particular important for porous nanomaterials. Therefore, detecting and locating trace impurities with a sensitivity in range of parts-per-million as well as visualizing the distribution of oxygen vacancies and Ti 3+ defects in TiO2 nanoporous materials is crucial to understand the influence of the synthesis route on the functional properties.
Complex three-dimensional (3D) nanomaterials are usually analyzed by transmission electron microscopy (TEM) enabling to assess their size, morphology, and crystal structure N2 .
Recent development in electron tomography, i.e. through acquisition of a tilt-series in TEM or scanning TEM (STEM), allows advanced 3D structural characterization 8 . Furthermore, the combination with energy dispersive X-ray spectroscopy (EDS) or electron energy loss spectroscopy (EELS) offers the opportunity for chemical analysis 9 . However, the distribution of surface disorder and oxygen vacancies has been barely analyzed in well-defined metal oxide nanostructures locally N1,N7 .
Atom probe tomography (APT) provides 3D elemental information with near-atomic resolution and a higher sensitivity than EDS and EELS, in particular for light elements 10,11 .
Moreover, combined APT and high-resolution (S)TEM have recently enabled quantitative measurements of solute segregation at grain boundaries as a function of grain boundary character and misorientation 12,13 . Although APT has been mostly used to analyze bulk materials 14 and some semiconductor nanowires with dopants [15][16][17][18] , it is still a burgeoning technique to investigate freestanding nanomaterials, in particular metal catalyst, metal oxide battery particles, or quantum dots [19][20][21][22][23] . Analyzing porous metal oxide nanomaterials using APT is highly challenging due to the low conductivity of metal oxides and the presence of pores. The latter renders the electrostatic field distribution very inhomogeneous and exacerbate reliable sample preparation technique. A step towards solving these issues could be the use of electrodeposition to fill the pores in microporous materials with a metal 20,32,33 In the present work we studied partially reduced hollow TiO2 1D nanowires (R-HTNWs) using (S)TEM and APT as a model system to demonstrate the feasibility of our approach. The 3D morphology of the R-HTNWs was characterized using electron tomography while the local distribution of oxygen vacancies, more specifically the correlated Ti 3+ species, was analyzed using EELS in STEM mode. To enable APT analysis, electrodeposition was used to embed R-HTNWs in a Ni matrix, which also fills the hollow core of the NW. The 3D hollow structure of partially reduced TiO2 was successfully reconstructed and the distribution of trace elements such as B, Na, and N, resulting from the reduction with NaBH4 under N2 atmosphere, were quantified. Our results show the ability to characterize the 3D structure and spatial distribution of impurities in porous metal oxide nanomaterials at an atomic resolution.

Results
Characterization of R-HTNWs using EM Hollow TiO2 nanowires (HTNWs) were synthesized on a FTO glass substrate using a hydrothermal method and reduced to R-HTNWs   Figure S2a presents STEM-EELS spectra of TiL2,3 and OK edges. The TiL2,3 edge has four peaks, resulting from the spin-orbit splitting into TiL3 (2p3/2) and TiL2 (2p1/2) edges with further crystal field splitting of Ti 3d states into t2g and eg. The four peaks here are characteristic for crystalline TiO2 such as anatase and rutile 39,40 . The left shoulder of the eg peak of TiL3 indicate the rutile nature of R-HTNWs N1 . The OK edge splitting indicates O 2pTi 3d hybridization states of t2g and eg. Elemental maps were obtained using TiL2,3 and OK edges.

Characterization of R-HTNWs using APT
The low electrical conductivity of TiO2 and the presences of pores in the R-HTNWs were obstacles for conventional APT characterization. Filling pores with a highly conductive material has been shown to be a solution for stable and homogeneous electrostatic field distribution 20,32,33 , but has not been demonstrated for poorly conducting, porous nanomaterials. Here, we co-electrodeposited R-HTNWs, detached from the FTO along with Ni to form a film on a Si substrate, which allowed to prepare APT specimen using focused ion beam (FIB) milling N4 . Metal oxide nanoparticles (i. e. Al2O3, TiO2, and La2O3) have been reported as inert particles during co-electrodeposition process of Ni films [41][42][43] . Therefore, this single-step method, without the electrophoresis step, is simpler than the previously reported two-step electrodeposition method 20 . The electrodeposition depends strongly on the surface heterogeneity 44,45 . In our case, we observed protrusions on the co-electrodeposited Ni surface, which was indicative of the presence of the embedded nanowires (see Figure S3). The presence of the nanowires inside of the protrusion is confirmed by cross-sectioning one of these regions with the FIB (see Figure S4a). The hollow and 1D nanowire morphology of the R-HTNWs are preserved after co-electrodeposition with Ni. The pores in the R-HTNWs are successfully filled with Ni without any voids, which could induce inhomogeneous field evaporation during APT measurements. Every protrusion we investigated contained R-HTNWs helping us to guide the fabrication of site-specific APT specimen. One of these regions was then lifted out, deposited onto a support, sliced to only contain the R-HTNWs, and sharpened into a needle-like specimen (see Figure S4b and 4c).
Details of the APT analysis of the R-HTNWs embedded in Ni can be seen in Figure   S5 and Figure S6. The major peaks in the mass-to-charge-ratio spectrum can be assigned to TiO molecular ions and the electrodeposited Ni matrix in single and double charged states (see Table S1). A strong O + peak originating from the R-HTNWs is detected at 16 Dalton (Da). Several peaks can be clearly assigned to atomic species introduced during the reduction of R-HTNWs. The peaks at 14 and 23 Da are N + and Na + ions, respectively, and the peaks at   Table S2. They differ by a factor of ~2. This is because for the calculation all four {110} surface planes are considered as the exposed surface area, two of which are actually blocked by other nanofingers ( Figure S8).
To address the origin of these minor elements further, an APT analysis was conducted on the as-grown and non-reduced HTNWs. Figure S7 shows the atom map and element composition profile of such a HTNW. Boron is not detected (see Table S2), confirming that the source of boron impurity is the NaBH4 used for reduction. A low amount of Na is detected in the HTNWs, indicating that the presences of Na in R-HTNWs can be from the NaBH4 used for the reduction as well as from the diffusion of Na from the glassbased FTO substrate during the hydrothermal synthesis of the HTNWs 46,47 . The low amount of N detected can originate from the nitrogen atmosphere during reduction or from adsorption during specimen transfer or their processing. A proximity histogram from a single nanofinger region in R-HTNW shows similar result to the hollow nanowire region (Figure 5b). The result reveals that impurities of Na, B, and N have a strong tendency to be deposited on the exposed {110} surface of the R-HTNWs during synthesis or the reduction. We have successfully resolved the 3D morphology of R-HTNWs using EM and APT.
The distribution of oxygen vacancies on single nanofinger was characterized using EELS and trace impurities of B, Na, and N were mapped in atomic scale using APT. We could unequivocally demonstrate that the impurities originate from NaBH4 that was used as a reductant for TiO2. This uncontrolled ingress of impurities was revealed by APT, as enabled by an advanced sample preparation technique whereby these hollow nanomaterials were embedded within a metallic matrix. Both oxygen vacancies and these impurities were concentrated at the surface of the R-HTNW.
These surface defects need to be considered when discussing structure -property relationships of R-HTNWs as their impact can be very large N9 . First, the surface impurities, which are unintentionally added to the system, contribute to distorting the rutile crystal struture. This distortion is responsible for higher carrier mobility leading to metallic-like conductivity [N10] . The surface impurities also stabilize Ti 3+ on the surface as interstitial B allows to have sufficient Ti 3+ species in TiO2 (B-O-Ti 3+ ) N6N8 . This is supported by out STEM-EELS measurements. B and N synergistically reduce the band-gap of TiO2, increasing the performance for photo(electro)catalyst and energy storage applications 50,51 . Additionally, surface B increases electrocatalytic properties in fuel cell applications due to the high oxygen adsorption ability N8 . All these examples show that the detection of trace elements enabled by the combination of advanced electron microscopy and atom probe tomography is necessary to further understand the functional properties of these and other nanomaterials. The lack of precise, quantitative characterization of impurities likely explain contradicting results reported in the literature, since different amounts of trace elements might be introduced during the synthesis, and were, up to now, most often not considered. The approach discussed herein could readily be deployed to a vast array of different materials systems and hence lead to a better control over the integration of impurities with an aim to enhance the materials' functional properties.

Synthesis of R-HTNWs
Rutile TiO2 nanowires were grown on a FTO glass substrate using a modified hydrothermal method 34,35 . They are composed of finer nanofinger bundles surrounded by compact and bigger outer nanofingers ( Figure S1a and S1b). HTNWs were obtained by selective core etching of TiO2 nanowires with a 1:1 (v/v) mixture solution of deionized water and hydrochloric acid (HCl, 36 wt%, Sigma Aldrich) in an autoclave reactor at 180ºC for 2h. The porous and finer inner nanofinger bundles could selectively be etched out in HCl solution because of their higher TiOH surface area where Cl  can easily attack ( Figure S1c), resulting in hollow morphologies 36,37 . Finally, R-HTNWs were prepared by the reduction of HTNWs with sodium borohydride (NaBH4, Sigma Aldrich) under N2 flow at 400 ºC for 2h with a heating rate of 10 ºC/min. The obtained R-HTNWs were washed using deionized water and dried at room temperature. Aldrich) and 2 g of boric acid (H3BO3, Sigma Aldrich) were dissolved in 50 mL of distilled water (18.2 MΩ-cm) from the AQUA Solution type 1. Then, as-synthesized R-HTNWs were dispersed in the Ni electrolyte using a sonicator for 10 min and the complex solution was poured into a vertical cell for co-electrodeposition process. Co-deposition was carried out in a specially designed vertical cell including a Cu substrate and a Pt-mesh counter electrode as shown in Figure S9a and S9b. In order to homogenously deposit Ni on the Cu substrate and effectively deposit the nanowires within the Ni layer, a vertical cell is designed in a conical shape having that the surface area of Pt counter electrode (2 cm 2 ) is larger than that of the Cu working electrode (0.2 cm 2 ). The one-step co-electrodeposition was performed at a constant current of 19 mA for 500s. In Figure S9c, a photograph of the co-electroplated Ni and R-HTNWs on a Cu substrate is shown.

Co-deposition of R-HTNWs within an electrodeposited Ni
Electron microscopy characterization SEM was performed to investigate the morphology of the R-HTNWs (Gemini 500, Zeiss, in-lens detector, 2 kV). (S)TEM was performed using a FEI 60-300 Titan Themis operated at 300 kV with a Cs-corrector for the probe forming lens.
The chemical composition of R-HTNWs was analyzed by EDS in the STEM mode. EELS data were acquired in the STEM mode using a dual channel acquisition mode 55  APT characterization A needle-shape specimen was prepared from the co-deposited sample using FIB (Helios NANOLAB 600i, FEI) for APT measurement to identify the chemical composition and position for each element 62 . APT analyses were performed using a local electrode atom probe (Cameca LEAP 5000 XS system) in pulsed UV laser mode at a detection rate of 1 %, a laser pulse energy of 80 pJ, and a pulse frequency of 125 kHz. The specimen temperature was set to ~50 K during analysis. The set parameter was based on the recent report on natural rutile (TiO2) measurement for atom probe tomography 63 . Data reconstruction and analyses were performed using the commercial software Imago visualization and analysis system standard (IVAS) 3.8.2 developed by Cameca Instruments.
All 3D atom maps presented in this paper were reconstructed using the standard voltage reconstruction protocol.

Corresponding Author
Correspondence and requests for materials should be addressed to J.L.