Electron Beam-Induced Writing of Nanoscale Iron Wires on a Functional Metal Oxide

Electron beam-induced surface activation (EBISA) has been used to grow wires of iron on rutile TiO2(110)-(1 × 1) in ultrahigh vacuum. The wires have a width down to ∼20 nm and hence have potential utility as interconnects on this dielectric substrate. Wire formation was achieved using an electron beam from a scanning electron microscope to activate the surface, which was subsequently exposed to Fe(CO)5. On the basis of scanning tunneling microscopy and Auger electron spectroscopy measurements, the activation mechanism involves electron beam-induced surface reduction and restructuring.


■ INTRODUCTION
Surfaces of rutile TiO 2 have for many years been used as model systems to explore the physics and chemistry associated with the varied applications of the material. 1,2 Of particular interest has been the surface science associated with light harvesting processes such as photocatalysis and dye-sensitized photovoltaics. 1,2 Applications in the field of molecular electronics have also been suggested, with TiO 2 being employed as a dielectric substrate. This has motivated studies of metal wire growth as interconnects using physical vapor deposition. 3 Moreover, electron-induced surface modification of TiO 2 (110) 4−6 is of interest as a means of patterning the substrate for wire formation or to template an array of functional molecules. There is a related interest in the generation of memristor elements on TiO 2 . 7 In this work we examine the potential of electron beaminduced surface activation (EBISA) 8 to deposit metallic structures, including wires, on rutile TiO 2 (110) in ultrahigh vacuum (UHV). The latter environment is essential for the analysis methods employed as well as to avoid contamination from residual gases. 9 EBISA is a technique that is part of a more general approach termed focused electron beam-induced processing (FEBIP), which also includes the related technique of electron beam-induced deposition (EBID). 10−13 A step-by-step explanation of the recently developed EBISA process is given in Figure 1. The substrate (a) is first irradiated with energetic electrons, resulting in local activation of the surface by electron-stimulated desorption of O atoms (b). 6,14 In a second step, precursor molecules are decomposed upon contact with the activated area, resulting in a deposition of nonvolatile material (c, d). In addition, the initial deposit grows in size because of autocatalytic decomposition of precursor molecules (e, f). 8 The autocatalytic growth has been demonstrated for iron pentacarbonyl, Fe(CO) 5 , on SiO x in UHV 8,15,16 and recently also for Co 2 (CO) 8 , indicating the general nature of this process for metalcarbonyls. 17,18 Here we demonstrate EBISA-induced interconnect growth on TiO 2 (110). Moreover, additional insight into the mechanism of the EBISA process has been achieved using local Auger electron spectroscopy (AES) to gain chemical information and scanning tunneling microscopy (STM) to provide atomic scale images.

■ EXPERIMENTAL SECTION
All experiments were performed in an Omicron UHV Multiscanlab at room temperature. The base pressure of the instrument was <2 × 10 −10 mbar. The instrument includes an electron column for scanning electron microscopy (SEM) with spatial resolution better than 3 nm. This column is also used in combination with a hemispherical electron energy analyzer for local Auger electron spectroscopy (AES) and scanning Auger microscopy (SAM) with a spatial resolution better than 10 nm.
For AES, a beam energy of 15 keV and a current of 3 nA were used, while all electron exposures for SEM and lithography were done at an electron beam energy of 15 keV and a probe current of 400 pA. The lithographic processes were controlled via a homemade lithography application based on LabVIEW 8.6 (National Instruments) and a high-speed DAC PCIe-card (M2i.6021-exp, Spectrum GmbH, Germany). For line exposures (single pixel width), a single sweep with a step size of 12 nm was applied. The line dose d L is calculated as d L = I PE × t dwell /Δx, where I PE is the beam current, t dwell the dwell time at every position, and Δx the step size along the line. The electron dose was controlled via the respective dwell times. The square patterns were exposed in a single sweep, meander-like pattern at a step size of 6 nm with the area dose d A being defined as d A = I PE × t dwell /(Δx) 2 . The corresponding area dose d A for a given line dose d L can be estimated by dividing d L by the step size.
The backscattered electron (BSE) exit areas and coefficient were simulated with the program Casino V2.42. 19 The diameter corresponds to an area of which 99% of BSEs are emitted. For the simulations, a primary electron (PE) number of 10 7 was applied. The beam diameter was set to 3 nm.
An STM scanner can be inserted between the pole piece of the SEM column and the sample for in situ STM measurements. Tip positioning is done using SEM imaging. W tips were used in this work, negatively biased relative to the sample. ST micrographs were acquired using Matrix V3.0 (Omicron) and evaluated using the latest versions of WSxM 20 (Nanotec Electronica, Madrid, Spain).
Rutile TiO 2 (110) samples (PiKem and Crystec) were prepared by repeated cycles of Ar + sputtering and annealing to 1160−1180 K until a sharp (1 × 1) low-energy electron diffraction (LEED) pattern was obtained with no detectable contaminants in AES. The purity of the Fe(CO) 5 precursor gas (ACROS Organics, 99.5%) was confirmed using a mass spectrometer. The precursor gas was dosed through a nozzle to about 12 mm from the TiO 2 (110) surface at an estimated local pressure of ∼9 × 10 −6 mbar, based on simulations with the software GIS Simulator (version 1.5). 21,22 ■ RESULTS AND DISCUSSION The first goal of this work was to extend the use of the EBISA technique, which thus far has only been reported for silica surfaces, to a different oxide material, namely a TiO 2 (110)-(1 × 1) surface. For that purpose, line patterns and square patterns (2 × 2 μm 2 ) were irradiated with a focused electron beam. Fe(CO) 5 was subsequently introduced to allow growth of iron structures at the irradiated positions. The gas was pumped off after 270 min, corresponding to an exposure of about 1.1 × 10 5 langmuir (1 langmuir is 10 −6 Torr s). Figure 2 shows SEM images of the resulting Fe line deposits (panels a−d) and Fe square deposits (panels e−h). The images show a strong increase in the amount of deposited iron with electron dose. AES measurements 24 h after fabrication (see Figure S1 in Supporting Information) indicate the purity of the iron structures is >90 atom %, which is similar to that achieved on SiO 2 substrates. 8,15,16 Those iron deposits on SiO 2 exhibit a low room-temperature resistivity of 88 μΩ cm as well as ferromagnetic behavior. 23 These properties can be expected to be similar on TiO 2 , potentially making the corresponding iron deposits suitable interconnects.
At higher SEM magnification we observe that the structures are comprised of crystalline iron nanocubes. The density of the cubes within the line and square deposits varies with the applied electron dose, whereas the size of the individual cubes at a given gas exposure seems to be similar. With increasing gas exposure the size of the clusters increases because of autocatalytic growth. For increased electron doses, the cubes start to merge into continuous patches, as can be seen for the 1.1 C/cm 2 square deposit (Figure 2g) or the line deposits (Figure 2a,b). Higher electron doses mark the start of pronounced proximity effects, i.e., unintended deposits near the irradiated patterns due to electron scattering. 11,16,24 Pronounced proximity effects are present at the 3.3 C/cm 2 square irradiation doses ( Figure 2h) and the higher line doses (Figure 2c,d). For EBISA, all of the observed proximity effects can be attributed solely to BSE and associated secondary electrons (SE II ). 16 Proximity effects from forward scattered electrons (FSE) are not observed in EBISA because of the absence of structure growth during the electron beam exposure step. 16 The observation of homogeneously scattered cubes in the nonirradiated surface regions at low electron irradiation doses (Figure 2a,b,e) points to the presence of nucleation points on the TiO 2 (110) surface that were not induced by the electron beam (see also Supporting Information, Figure S2). The defects are oxygen vacancies, which typically are present at an initial concentration of 5% ML (1 monolayer is the number of surface unit cells). The vacancies react with water in the residual vacuum to form the second major form of defects, bridging hydroxyls (one water molecule reacts with one vacancy to produce two hydroxyls). 25 Turning to the EBISA mechanism, we can first look at the effect of the 15 keV electron beam on the reduction state of the surface. Figure 3a shows Auger electron spectra for nonirradiated (red line, "pristine") and strongly irradiated (blue line, "reduced") areas of a freshly prepared TiO 2 (110) surface. The most prominent features are the Ti LMM and Ti LMV peaks at It is immediately apparent that the oxygen signal for the strongly irradiated surface (blue) is considerably smaller than that for the pristine surface (red); the Ti signals show the opposite trend, albeit not as pronounced. This behavior evidences the electron beam-induced desorption of oxygen from the surface and is in line with that expected from electronstimulated desorption (ESD) results from TiO 2 at lower electron beam energies. 6,14 In addition, the change of the peak shape of the Ti LMV signal allows for insight into the chemical state of the Ti atoms of the surface. As discussed by Nishigaki 26 and Gopel, 27 the LMV Auger peak consists of two components: 413 eV for Ti 4+ and 419 eV for Ti <4+ . In a simplified picture, for Ti 4+ , in the complete absence of a Ti 3d valence electron, the valence electron in the LMV Auger process stems from an interatomic transition of an O 2p electron with a higher binding energy; on the contrary, for Ti <4+ , at least a partial electron charge remains in the Ti 3d level (at lower binding energy than the O 2p level) and thus participates in an intra-atomic Auger transition at higher kinetic energy. Therefore, the observed intensity shift from the low-to the high-kinetic energy contribution of the Ti LMV peak is a direct indication of a reduction of the Ti 4+ by the electron beam. The evolution of the LMV peak shape upon increasing electron irradiation is illustrated in Figure 3b from top (pristine, red) to bottom (reduced, blue).
In Figure 3c, the normalized (I pristine = 1) peak areas of the Ti LMM (black) and O KLL (orange) signals and the O:Ti peak ratio (green) are plotted versus the primary electron dose (note the logarithmic scale). The apparent O:Ti peak ratio is normalized so that the value for the pristine surface is set to 2, i.e., TiO 2.0 . At doses lower than 0.1 C/cm 2 , the peak ratio does not change, i.e., the applied dose is not sufficient to induce a reduction of the surface detectable in AES. After crossing the 0.1 C/cm 2 threshold, the oxygen intensity and the O:Ti ratio drop rapidly until a fairly stable lower limit at ∼0.9 is approached at doses larger than 10 3 C/cm 2 . The corresponding relative oxygen intensity is 0.6 of the initial value of the pristine surface. The obtained O:Ti ratio of 0.9 indicates an average surface composition of TiO 0.9 . Note that the chemical composition of the surface is likely to be nonuniform after the reduction process, so the detected Auger signal is a convolution of different Ti oxidation states and chemical surrounding. In addition, decreased damping due to the loss of oxygen atoms from the topmost layers has to be taken into consideration when assessing the Ti LMM intensity. Nevertheless, the O:Ti ratio serves as an indicator for the degree of electron beam-induced reduction of the surface.
The Auger spectra indicate that above 0.1 C/cm 2 the stoichiometry of the surface changes gradually from TiO 2 to TiO 0.9 at 10 3 C/cm 2 . As for the accompanying morphology change, Figure 4 shows the effect of the electron beam as viewed by STM. Prior to electron beam exposure, the STM image shows [001] direction bright rows separated by 0.65 nm that correspond to Ti atoms. 1,28 A line exposure corresponding to an electron area dose of approximately 60−100 C/cm 2 (electron energy, 15 keV; beam current, 400 pA; step size, 12 nm; dwell time, 150 ms; line dose, 50 μC/cm) results in an O:Ti ratio of 1.1−1.2. The exposed region can be clearly identified in the image in Figure 4 as an area with a disordered appearance and a strongly increased number of protrusions (bright spots) arranged within a vertical line with a width of approximately 20 ± 5 nm (indicated by dotted yellow lines); this width is ∼7 times larger than the nominal beam diameter of 3 nm (Gauss profile, 20/80 criterion), indicating the wellknown broadening due to SE and BSE proximity effects. The apparent surface roughness (rms) of the exposed area (0.12 nm) in STM is double that of the nonirradiated surface. The left-hand side of the image in Figure 4 resembles the asprepared surface, with the characteristic [001] direction Ti rows visible (see also Supporting Information, Figure S3). However, there are additional defects in the form of isolated protrusions and in-row depressions (white arrows) that can be attributed to proximity effects, as the density is higher than on the as- The central line is a result of activation by PEs, while the fringe is caused by BSE proximity effects. For the square fields, deposit density increases from a loose grain assembly at 0.27 C/cm 2 via a close-packed assembly at 0.55 C/cm 2 to a fused grain deposit of near ideal geometry at 1.1 C/cm 2 . At even higher doses, proximity effects also cause a fringe growth (3.3 C/cm 2 ). Note that scattered iron clusters are found all over the surface without electron exposure, a fact which is attributed to active defects created during sample preparation. prepared surface. Apart from the increase in mean roughness, the electron beam-induced disorder does not cause a significant increase in average apparent height. This indicates the absence of unintentionally formed deposits from residual gases.
The center of the SEM image in Figure 5a clearly shows a significant electron beam-induced darkening. In previous studies we showed that this darkening of irradiated areas is an intrinsic feature of electron beamed oxide surfaces and is not indicative of contamination. 8 These results indicate that the modifications to the TiO 2 (110) morphology and subsequent reaction with Fe(CO) 5 are a result of electron beam-induced oxygen loss.
Panels b and c in Figure 5 show the STM images of the irradiated 50 μC/cm line pattern before and after Fe(CO) 5 dosing (total exposure: ∼10 langmuir), respectively. After the gas exposure, the formation of iron deposits in the form of clusters is observed, primarily in the regions irradiated with electrons (indicated by yellow lines). Larger clusters appear as slightly asymmetric protrusions in the image, a consequence of the applied scanning speed. These clusters have a typical diameter of 3.2 ± 0.3 nm, including tip convolution, and an apparent height of 1.2 ± 0.1 nm (see Supporting Information, Figure S2). This corresponds to clusters containing 650−1000 Fe atoms.
In addition to the clusters observed within the yellow lines in Figure 5, some deposits are also found in nonirradiated areas to both sides of the irradiated line. These clusters originate from growth due to the intrinsic surface defects in TiO 2 (110) (see above), as well as from growth due to BSE proximity effectinduced defects. The BSE exit range of TiO 2 , here defined as the distance from the point of impact that includes 99% of all emitted BSEs (r 99 ), was extracted from MC simulations and amounts to about 1050 nm. All of the scanned area in Figure 5c is well within this exit range. The BSEs (E > 50 eV by definition) are able to induce ESD of oxygen atoms within the Knotek−Feibelman framework (E min,ESD = ∼34 eV), 14 thus creating isolated defects.  The surface on the left shows only minor damage; the typical row pattern of the TiO 2 (110) surface is preserved. Defects include (1 × 2) reconstructed sections (small arrows), dark defects in the row structure, and isolated protrusions. On the right, a disordered section running top to bottom was identified as the result of a line irradiation with a primary electron dose of 50 μC/cm. The (1 × 1) structure is barely retained in the highly damaged area; no characteristic defects can be identified in the irradiated area. The loss of long-and shortrange order corresponds to an electron-induced surface amorphization.
The Journal of Physical Chemistry C Local Auger electron spectrum of an iron structure prepared by EBISA on TiO 2 ( Figure S1), higher resolution STM data of a line irradiation before and after Fe cluster growth ( Figure S2), and STM images of the freshly prepared rutile TiO 2 (110) surface ( Figure S3). This material is available free of charge via the Internet at http://pubs.acs.org.