3.1. Synthesis Method
It is well-known that the flow of electrical current through a conductor induces Joule heating. This heating will depend on the equilibrium and transport properties of the material, such as the density (), specific heat (), the electrical () and the thermal conductivity (). The electron flow at high current densities could lead to cations’ transport in the conductor via the momentum transfer, which is known as the electromigration process. Herein, we aim at harvesting this phenomenon to produce WO3 micro- and nanostructures on the surface of W wires under ambient conditions when an electrical current is going through the wire.
Figure 1a shows the layout of the designed setup for the resistive heating experiments along with a photograph of the actual system. A 6 cm W wire is placed between two electrical contacts to where the current source is applied. The conductor elements stand over a PVC insulator base. Measurements are conducted under air conditions and at room temperature. The temperature of the whole system when the electrical current is passing through the wire was calculated taking into account both electrical (
) and thermal transport (
) across a specific conductor, and is governed by the following equations:
where
is the total electrical density current,
is the electric field and
Q is the heat energy. The temperature reached at the stationary state can be obtained by solving the coupled equations system in Equation (
1) with a finite element model. In this case, the material properties (density, specific heat, the electrical and thermal conductivity, and dielectric constant) of tungsten as a function of the temperature were taken from the literature [
20,
21,
22].
In electronic conduction, the heat generated by the current flow is the Joule heating, and electromigration is the atomic diffusion driven by a high current density in the direction of the electron flow. Joule heating and electromigration are two coupled mechanisms. On the one side, Joule heating will increase the temperature of the sample that, in the case of filaments or wires, will create a temperature gradient from the central part towards the edges because of heat dissipation in the material. This process affects the atomic diffusion on the material. On the other hand, at high temperatures, a faster atomic diffusion will drive a faster electromigration’s rate in turn [
17,
23]. The electromigration process requires a high current density. Actually, current densities in the range of 10
A/cm
were reported as threshold values to achieve enough electron flow-induced strain to cause the electromigration process in the Cu stripes [
16].
Our objective is to achieve an accumulation of cations at the wire surface that could act as nucleation sites for the growth of WO
3. The simulation results of the electron and heat transport along the tungsten wire have allowed us to target the current values required to achieve the high temperatures enough for the tungsten oxide crystallization. After several experiments, we conclude that there is a critical current density below which no growth will occur; for our treatments, the minimum current that needed to be applied is 16.5 A. This means a density current of 7.5 ×
A/cm
. If we use a higher current, the time needed for the growth will be reduced. The higher temperatures are expected at the central part of the wire, since more heat dissipation occurs closer to the metal contacts ends.
Figure 1b displays the local temperature of the whole system calculated when a current of 16.5 A is applied after 10 s. The scale bar shows a local temperature of 1480 K at the centre of the wire. We performed simulations at several currents to assess the local heating response of the W wire.
Figure 1c shows the temperature profiles along the wire when density currents vary between 5–8 ×
A/cm
. It is observed that the central region of the wire, between 2 and 4 cm depending on the current, is well above 1000 K. Moreover, the native oxide layer at the surface of the W wire may prevent further atomic diffusion, which leads to an accumulation of atoms on the wire surface that favor the nucleation of metal oxide microstructures.
Figure 2a shows the secondary electron (SE) image of the wire after 50 s sustaining a current of 16.5 A. The time for the synthesis could be extended, but at a certain point the wire is going to break because of the material loss by the electromigration process dragging. Under higher currents than 16.5 A, the time needed for the growth is reduced to few seconds. Finally, above 18 A, the wire breaks down almost immediately. The temperature measured with the Infrared thermometer is 1510 K, which fairly agrees with the calculated one (1480 K). The Joule heating of the wire has caused the formation of well-crystallized structures over about 4 cm in the central part of the wire. According to the simulations (
Figure 1c), the temperature in this region is between 1000–1510 K for 16.5 A, which propitiates the nucleation and formation of the oxide crystalline structures. In spite of the thermal gradient along the wire, the morphology of the structures formed on the wire are rather uniform, as is shown in
Figure 2a, although a lower density of WO
3 structures is found as we go further from the central part of the wire. The detailed morphology of these micro- and nanostructures are shown in
Figure 2b,c. The most common shape corresponds to hierarchical microstructures that consist of several tapered rods, often emerging from the same point and built by stacked plates of decreasing size, as is shown in
Figure 2b,c. Eventually, small needles are also formed on the lateral plates as secondary growth (
Figure 2d). These oxide structures form a shell around the metallic wire of a thickness of about a hundred micron, being easy to separate from the wire. Moreover, the remaining W wire under the oxide layer can be reused in subsequent new growths.
In addition, experiments were also conducted under an external electric field applied perpendicularly to the W wire by placing two copper electrodes, as shown in
Figure 1a. This approach not only adds a drift component to the atomic flux towards the wire surface, but also a cations flight from the surface to the electrode plates. This process adds another source for the nucleation and growth of WO
3 nanostructures. As a result, besides the microstructures formed on the wire surface, a big amount of micro- and nanostructures are also deposited on the external copper plates, giving rise to a thin layer of structures of a few hundred microns on the electrode.
Figure 2e shows the secondary electron image of these WO
3 nanostructures. It is worth noticing that new morphologies, as microspheres, appear on the electrodes in comparison with structures formed on the surface wire. These first results demonstrate that the Joule heating of tungsten wires under suitable electrical density currents bring about WO
3 micro- and nanostructures of several shapes, which would imply, as well, several crystalline phases.
3.2. Structural Characterization
In order to assess the crystalline structure of the achieved WO
3 microstructures, X-ray diffraction experiments were carried out. To this purpose, the structures were transferred to a silicon substrate.
Figure 3a shows the XRD diffractogram recorded under the gracing incident for three representative samples: WO
3 formed on the wire surface without an external electric field, with an external electric field and WO
3 deposited on copper electrodes, revealing the presence of multiple crystalline phases. Some peaks are common to the three samples due to the small variations among the phases [
4]; thus, due to the coexistence of phases and the overlapping of the patterns, the analysis becomes difficult. However, focusing on the 22–25º interval, the characteristic peaks of each phase can be resolved.
Figure 3b shows the zoom up of XRD data in this angle interval for the three samples in which peaks assigned to the
-WO
3 (ICSD 10872385),
-WO
3 (ICSD 80056) and
-WO
3 (ICSD 1620) phases are marked with stars, squares and diamonds, respectively. The XRD results show that WO
3 formed on the wire consists of a mixture of
-WO
3 and
-WO
3 phases, while WO
3 deposited on copper electrodes contains the
-WO
3 phase, as well.
To corroborate these results, synchrotron XRD measurements were also carried out at the ESRF (BM25-SpLine beamline) in WO
3 structures deposited on copper electrodes and on the wire surface. To that end, samples were inserted in a quartz capillary that rotates during the measurements, to avoid any anisotropic effects. Synchrotron radiation XRD results of two representative samples: structures formed on the W wire surface under an external electric field and on the external copper electrodes have been performed. The results are shown in
Figure 3c, in which the yellow plot stands for WO
3 from the the wire surface and the pink plot for structures deposited on electrodes. From these results, there is also a clear evidence of the presence of the
-phase in the WO
3 microstructures deposited on copper electrodes.
The polymorphous nature of the WO
3 microstructures obtained by resistive heating was also assessed from micro-Raman measurements performed in a confocal microscope.
Figure 4 shows the Raman spectra of structures from the electrodes (green and light blue lines) and from the wire surface (purple and dark blue lines). The main characteristic Raman bands of the
-WO
3, at 133, 272, 716 and 803 cm
, are present in all the analyzed samples. The low energy bands, 133 cm
and 272 cm
, are associated to O-W-O lattice vibrations and bending modes, respectively, while the high energy bands (716–803 cm
) may be related to the O-W-O stretching vibrational modes. This phase has been reported to be stable at room temperature in microcrystalline WO
3 [
7]. Besides this phase, a careful inspection of the high energy region highlighted with a red square in
Figure 4 reveals that characteristic modes of the reported low temperature phases can also be resolved [
4,
5,
7]. In particular, a band at 605 cm
, marked with a diamond and characteristic of the
-WO
3, is observed in some structures formed on the surface wire, and bands at 636 and 673 cm
, marked with stars and assigned to the
-WO
3, are detected in structures deposited on the copper electrodes. These results agree with the above-shown diffraction data.
In order to complete the structural characterization, a TEM study of individual microstructures is conducted.
Figure 5a shows a low magnification TEM image of the nanostructures deposited on the electrode, in which two types of morphologies, rounded and octahedron-shaped particles marked with circles, are clearly identified.
Figure 5b,c show the HRTEM image and the SAED pattern from an spherical particle. Diffraction spots have been identified under the [001] zone axis of the
-phase of WO
3.
Figure 5d shows the sketch of the atomic structure in this orientation and agrees with experimental results. An analogue HRTEM and SAED analysis has been made for the octahedral particles, resulting in the
-phase as the WO
3 structural order (
Figure 5e–g).
The above characterization results allow us to draw the following conclusions. The resistive heating of W wires under suitable density currents (5–8 × 10 A/cm) and assisted by an external electric field, gives rise to a high amount of WO3 microstructures adopting several shapes. This growth process is far from the equilibrium and yields polymorphic material with up to three structural phases, that otherwise would require low temperature conditions. The -WO3 phase, which is stable at room temperature, is the dominant one in all samples, as could be expected. However, the -WO3 phase is also found both in structures formed on the wire surface and on the electrodes, depending on the local area investigated. In addition, the -WO3 phase, which has only been reported at low temperatures, has also been identified in microstructures deposited on the copper electrodes.
3.3. X-ray Photoelectron Spectroscopy (XPS) Study
The perovskite-like crystalline structure of WO
3 allows for withstanding a considerable amount of oxygen vacancies. This fact could even lead to structural changes with the stabilization of nonstoichiometric compounds, WO
, and/or to changes in the oxidation state of cations, for the sake of the charge balance [
24,
25]. X-ray photoemission spectroscopy is a very suitable tool to assess the electronic and chemical properties at the surface, and, in particular, the oxidation state of chemical elements. To this end, we have carried out XPS measurements on the ESCAmicroscopy beamline at Elettra on WO
3 microstructures deposited on the electrode. The XPS analyses were carried out qualitatively in order to understand the band structure from the oxidation state of the sample for a later study of its optical properties.
Figure 6a shows the W 4f core level spectrum, which have contributions of both W
and W
oxidation states to the W 4f
and the W 4f
lines. The W
to W
ratio was deduced from the relative integrated areas of the corresponding peaks in the XPS spectra shown in
Figure 6a, following the procedures described elsewhere [
26] and yielding a 0.32 ratio. This could be evidence of the presence of defects, as oxygen vacancies close to the surface, or of some nonstoichiometric phases. The latter would not match with the above structural results that supports the
,
, and
phases of WO
3. However, the evident presence of W
would support a high concentration of oxygen vacancies at the surface level. On the other hand, the fitting of O 1s’ core level (
Figure 6b) yields three peaks: 530.7, 531.5 and 532.8 eV peaks. The 530.7 eV peak is the most intense and is attributed to the O
ions of the WO
3. The components at higher binding energies can be attributed to OH
(531.5 eV) and O
ions (532.8 eV) [
27]. The hydroxide component is weak, but needed for an accurate fitting calculation, and could be originated from some water adsorption from the ambience. The component at the higher energy binding is associated with the presence of oxygen vacancies [
27].
Therefore, the XPS results support that WO
3 nanostructures contain a significant amount of oxygen vacancies, while keeping the structural arrangements of stoichiometric compounds. These defects often introduce electronic levels that behave as donors but also act as recombination sites that could hamper the performance of applications based on electron transfer, as photoelectrochemical water splitting or photocatalytic applications [
28]. The donor electronic levels related to oxygen vacancies in oxides are usually involved in the luminescence mechanisms of the material. In highly defective samples, the associated band could even be the dominant emission. Herein, we have carried out cathodoluminescence measurements in the SEM, in order to excite all possible recombination paths by electron bombardment. The CL spectrum of WO
3 nanostructures at room temperature reveals a broad emission in the visible range (1.6–2.9 eV) that could be deconvoluted into several components (
Figure 7). Taking into account that the bandgap energy of bulk
-WO
3 has been reported to be about 2.62 eV, the complex CL spectrum achieved may have contributions from a number of electronic levels related to defects in the polymorphic WO
3. Further work will be needed to more precisely understand the luminescence behaviour of WO
3.