Polyoxoniobates as molecular building blocks in thin films

Abstract

Niobium oxide thin films have been prepared by spin-coating aqueous solutions of tetramethylammonium salts of the isostructural polyoxometalate clusters [Nb₁₀O₂₈]⁶⁻, [TiNb₉O₂₈]⁷⁻ and [Ti₂Nb₈O₂₈]⁸⁻ onto silicon wafers, and annealing them. The [Nb₁₀O₂₈]⁶⁻ cluster yields films of Nb₂O₅ in the orthorhombic and monoclinic crystal phases when annealed at 800 °C and 1000 °C, respectively, whereas the [TiNb₉O₂₈]⁷⁻ and [Ti₂Nb₈O₂₈]⁸⁻ clusters yield the monoclinic crystal phases of Ti₂Nb₁₂O₂₉ and TiNb₂O₇ (titanium–niobium oxides) in different ratios. We also demonstrate a protocol for depositing successive layers of metal oxide films. Finally, we explore factors affecting the roughness of the films.

---

Introduction

To sustain the fast-paced energy- and technologically driven society we live in, we need to develop new materials, and improve on those currently in use. Many electronic devices, from mobile phones to MRI machines, rely on niobium pentoxide (Nb₂O₅) and tantalum pentoxide (Ta₂O₅).¹ The inertness of niobium and tantalum oxides is associated with good biocompatibility and corrosion resistance.^2,3 Important applications where niobium, tantalum and Nb/Ta-titanium oxide films are used include orthopaedic/orthodontic devices,^4,5 waveguides,⁶ electromagnetic interference shielding,⁷ and coatings for corrosion protection and biomedical devices.^8–11 Due to the extensive use of these metal oxides and a lack of alternatives, niobium and tantalum are classified as technologically critical elements.

Key challenges lie in sourcing – leading to coltan, the mineral mined to produce niobium and tantalum, being termed a ‘blood mineral’¹² – and implementing sustainable recycling protocols due to the inertness of niobium and tantalum oxides. Hence, better atom economy in usage, coupled with improved recovery methods, are needed.

There is also a need to develop lead-free piezoelectric materials to circumvent the use of lead zirconium titanate (PZT) piezoelectric devices. Sodium potassium niobate (KNN) and lithium niobate (LiNbO₃) are considered suitable alternatives to PZT.¹³ However, the methods of synthesising KNN thin films may be energy intensive, or lack reproducibility and require the use of toxic precursors,¹⁴ and the ability to vary the alkali–metal content and type is very limited, which is of relevance to this study.

Translating recent advances in the understanding of polyoxoniobate and -tantalate chemistry¹⁵ into novel methods for making thin films therefore promises to allow better control over the composition of the films, better atom economy, and access to films with improved or novel properties.

Polyoxoniobates and -tantalates are examples of polyoxometalates (POMs),^16–18 discrete anionic oxide clusters of group 5 and 6 metals in their highest oxidation states, and they are attractive as potential building blocks in the preparation of extended materials, such as thin films. In particular, the discrete nature of POMs means that their structures can be known with confidence, and their structures and composition can in many cases be tuned by targeted synthetic modifications.^15,19–25 At a minimum, this means that the elemental composition of the film can be controlled to a high degree, with the added possibility of the POM structure being wholly or partially retained in the solid film, so that the relative locations of individual atoms can be controlled.

Many POMs are also highly soluble – the solubility of tetramethylammonium ([N(CH₃)₄]⁺; TMA) salts of the polyoxoniobates typically exceeds 1 g ml⁻¹ in water, and are moderately soluble in methanol and ethanol, making them easy to handle and manipulate in the lab. Recently developed protocols for microwave synthesis offer short reaction times (15 to 60 minutes),²⁶ capable of yielding tens of grams of pure product in one-pot reactions in the lab. In addition, organic counter ions, such as TMA, can be removed by thermal decomposition,^27,28 something which is not possible with alkali ions. This allows for the production of films that can be freely doped with different ions and at different ratios. Finally, while there is a range of existing methods for making metal oxide films,^29–32 not all scale well, or can be applied to a wide range of substrates. Sol–gel,²⁹ sputter deposition³⁰ or laser ablation methods^31,32 may suffer from issues of reproducibility and/or cost.³² The use of polyoxometalates circumvent these limitations, while offering the possibility of heteroatom doped Nb₂O₅ or Ta₂O₅ to be fabricated through soft methods. Film preparation through spin-coating of benign, non-volatile, water-compatible precursors that can easily be made in bulk quantities, is thus a very appealing idea.

Anhydrous niobium pentoxide (Nb₂O₅) – the most common form of niobium oxide – exhibits temperature-dependent polymorphism and can be made from amorphous hydrous Nb₂O₅ (or niobic acid). Pseudohexagonal TT-Nb₂O₅ forms between 500–600 °C, orthorhombic ortho-Nb₂O₅ (or T-Nb₂O₅) is obtained at 600 °C–800 °C, tetragonal M-Nb₂O₅ forms at 850 °C–950 °C, and monoclinic mono-Nb₂O₅ (or H-Nb₂O₅) is seen at >950 °C.^33–35ortho- and mono-Nb₂O₅ prevail in the literature, with TT- and M-Nb₂O₅ being considered as metastable forms.^33–35

Anhydrous Ta₂O₅ also exhibits polymorphism. It exists in an orthorhombic crystal system as L-Ta₂O₅ or ortho-Ta₂O₅ (low temperature form) below 1360 °C and reversibly converts to the H-Ta₂O₅ phase (high temperature form) above 1360 °C.^36–38 Other Ta₂O₅ polymorphs are accessible but require high pressures to promote their formation.³⁹ As with niobium, there exists a hydrated amorphous form, tantalic acid. The reactivity of these amorphous oxides is much higher than that of their anhydrous counterparts,²⁶ and have been demonstrated to be good precursors in polyoxoniobate and polyoxotantalate synthesis.^{15,22–24,40}

Titanium–niobium oxides are also found as orthorhombic and monoclinic crystal systems. Crystalline Ti₂Nb₁₀O₂₉ is dimorphic and exists in both the orthorhombic and monoclinic crystal systems,⁴¹ while the crystalline TiNb₂O₇ exists only in the monoclinic crystal system.⁴² Both can be made from hydrous Nb₂O₅ (niobic acid) and TiO₂ directly at ≥1000 °C,^41–44 niobium citrate and Ti(ⁱPrO)₄,⁴⁵ or NbCl₅ and Ti(ⁱPrO)₄ as sources of hydrous Nb₂O₅ and TiO₂, respectively,^46,47 These methods are limited due to the extensive heating time and temperature required to form the titanium–niobium oxides from solid Nb₂O₅ and TiO₂. Notably, the titanium–niobium oxides are considered as prospective anode materials for lithium-ion batteries, to replace graphite or Li₄Ti₅O₁₂ anodes currently used.^48,49 Studies of the electrochemical properties of Ti₂Nb₁₀O₂₉ have determined it has a high rate capability and diffusion coefficient, which facilitates Li⁺ and electron transport.^47,50,51 The TiNb₂O₇ is of particular interest because of its energy density and lithiation properties.⁴⁸

Decaniobate (Nb₁₀) has been used as an aqueous precursor for depositing Nb₂O₅ and Nb₂O₅-ITO thin films on glass to elicit optical, electrochromic and electrochemical properties.^52–54 These studies were limited to temperatures below 400 °C in the annealing process of Nb₂O₅ due the thermal properties of ITO. Chemical condensation of Nb₂O₅ using formic acid was also investigated at room temperature.⁵⁴

A comparative study by Fullmer et al.²⁷ of the Nb₂O₅ and Ta₂O₅ thin films deposited from aqueous solutions of hexaniobate (Nb₆) and hexatantalate (Ta₆), respectively, found that smooth Ta₂O₅ films always formed from Ta₆ – regardless of annealing temperature – unlike that of the Nb₂O₅ film from Nb₆. There was, in particular, a noticeable increase in the roughness of the Nb₂O₅ films formed at 600 °C and 800 °C, when compared to Ta₂O₅ films. This was attributed to a preference of either a lower activation energy of grain growth in Nb₂O₅, or the formation of proton-bridged Ta₆ dimers vs. linear proton-bridged Nb₆ chain-oligomers, with the latter serving as nucleation seeds.²⁷

Mansergh et al.²⁸ further probed the orthorhombic Ta₂O₅ thin films deposited from Ta₆. Film thickness was determined to be directly proportional to the concentration of Ta₆, while film thickness decreased at elevated temperatures. The latter decrease is associated with the removal of water, and decomposition of the TMA as the film condenses. The Ta₂O₅ films also become courser following heating above 700 °C due to grain growth, which coincides with a decrease in the refractive index of the film above 600 °C.²⁸ More recently, Saez Cabezas et al.⁵⁵ have developed an electrochemical method for acid-catalyzed condensation of Nb₂O₅ and Ta₂O₅ thin films, from Nb₆ and Ta₆, respectively, via the potentiostatic oxidation of water. They determined both Nb₂O₅ and Ta₂O₅ film thickness to increase with the applied potential and that film thickness increases rapidly during the first two minutes of the electrochemical deposition process, after which it plateaus. Finally, Fast et al.¹⁴ spin-coated multiple layers of potassium sodium niobate thin films from sodium/potassium hexaniobate, yielding homogeneous films. Hexaniobate is only stable at alkaline pH above 9–10, which limits the chemical compatibility with potential dopants. In addition, the use of alkali salts leads to the forced inclusion of alkali cations in the finished film, whether desirable or not.

In this study, we used solution deposition of polyoxoniobates to develop surfaces where the crystallinity, thickness and roughness can be tuned by changing the starting material, concentration, and annealing temperature. A multi-step protocol has been developed to ensure layered metal oxide thin films can be deposited through an iterative spin-coating and annealing process.

Experimental

Chemicals were used as received. [NC(H₃)₄]₆[Nb₁₀O₂₈]·2H₂O (Nb₁₀), [NC(H₃)₄]₇[TiNb₉O₂₈]·3H₂O (TiNb₉), [NC(H₃)₄]₈[Ti₂Nb₈O₂₈]·6H₂O (Ti₂Nb₈), [NC(H₃)₄]₈[Nb₆O₁₉]·2H₂O (Nb₆) and [NC(H₃)₄]₈[Ta₆O₁₉] (Ta₆) were synthesised using microwave irradiation according to published procedures.^26,56 Aqueous solutions (0.2 M) of Nb₁₀, TiNb₉, Ti₂Nb₈, Nb₆ and Ta₆ were spin-coated onto silicon wafers (50 × 50 mm, p-type, 10–20 Ω cm, Siegert Wafer GmbH, Germany) pre-treated with piranha solution and atmospheric-pressure plasma. See ESI† for details. In the typical experiment, 100 μl of polyoxometalate solution (0.2 M) was placed at the centre of the substrate, and spun at 2000 rpm for four minutes, before annealing in a furnace using a ramped temperature program (see ESI† for details). Raman spectra of bulk solutions of polyoxoniobates and of annealed powders were collected using a 532 nm laser on an Anton-Paar Cora 5600 spectrometer. Raman spectra of spin-deposited films were collected using a Renishaw Qontor Raman microscope, using a 405 nm laser. Powder X-ray diffraction measurements were done on a PANalytical X'Pert³ diffractometer with a CuK_{[small alpha, Greek, macron]} source. Ellipsometry measurements were done using an Alpha-SE Ellipsometer (J.A. Woollam Co. Inc.). Atomic force microscopy imaging was done with a Bruker BioScope Catalyst instrument operating in peak force mode in air. See ESI† for additional details.

Results and discussion

We have explored the spin-coating of tetramethylammonium (TMA) salts of a series of polyoxoniobate and polyoxotantalate clusters – Nb₆, Nb₁₀, TiNb₉, Ti₂Nb₈ and Ta₆ (Fig. 1). These targets were chosen as the structures are known with great confidence, and their solution phase chemistries have been studied extensively.^57–60 While Ta₆ and Nb₆ are stable at pH 9, Nb₁₀ is only stable in a very narrow pH region, ca. 5–7.5,⁵⁷ above which it dissociates to Nb₆. The replacement of one niobium atom by a titanium atom in Nb₁₀, however, extends this stability range to pH 6.5–12.0 in TiNb₉,⁵⁹ and for Ti₂Nb₈ no dissociation or speciation is evident even at pH 12.5.⁶⁰ While the microwave syntheses of Nb₁₀, Nb₆, Ta₆ and TiNb₉ have been reported,^26,56 for this work we also developed the microwave synthesis of Ti₂Nb₈. Although these molecules can be made through hydrothermal methods by heating overnight,^15,23,40 microwave synthesis affords these clusters after irradiation for 30 minutes, and can easily be scaled up to yield 10–15 grammes per one-pot batch. Rapid production of large quantities of compound is an important factor in making these types of clusters in industrial thin film synthesis both attractive and viable.

Fig. 1 The [Nb₁₀O₂₈]⁶⁻ (a), [TiNb₉O₂₈]⁷⁻ (b), [Ti₂Nb₈O₂₈]⁸⁻ (c), [Nb₆O₁₉]⁸⁻ (d) and [Ta₆O₁₉]⁸⁻ (e) polyoxometalate ions. Red, light blue, grey and dark blue spheres represent oxygen, niobium, titanium and tantalum atoms, respectively.

Silicon wafers were chosen as the substrate in spin-coating deposition experiments as they have high smoothness, good chemical inertness, and can tolerate the elevated temperatures needed for annealing. Silicon substrates were cleaned with piranha solution – a 3 : 1 solution of concentrated sulfuric acid and 30% hydrogen peroxide – before being treated with atmospheric pressure plasma in a plasma chamber. This plasma treatment was necessary for the spin-coated polyoxometalates to adhere to the surface of the substrate.

First, the effects of the annealing temperature and polyoxometalate concentration were investigated. Based on spectroscopic ellipsometry, there is a consistent decrease in film thickness with increased annealing temperature up to 800 °C. The most notable change in thickness, however, occurs when the films are heated between 200 °C–400 °C and corresponds to removal of water and decomposition of the TMA counter ion. As expected, the film thickness increases at higher polyoxometalate concentrations. This is illustrated for the Nb₁₀ and Ta₆ films in Fig. 2. At high polyoxometalate concentrations (0.5 M), the films are prone to spontaneous crystallisation after spin-coating and prior to annealing, leading to the formation of a transparent crystalline layer on the surface of the film. After annealing, this crystallisation is manifested in the presence of grain boundaries – especially at higher temperatures – that induce film coarseness and total internal reflection of incident light. The latter phenomenon limits the reproducibility of thickness measurements. Hence, 0.2 M polyoxometalate solutions were used throughout the study as this concentration yielded sufficiently thick films that could still be probed via spectroscopic ellipsometry.

Fig. 2 Thickness of Nb₂O₅ (a) and Ta₂O₅ (b) films obtained from annealing Nb₁₀ and Ta₆, respectively, and determined by ellipsometry.

Knowing that a phase change in Nb₂O₅ occurs between 800 °C and 1000 °C, we spin-coated aqueous solutions (0.2 M) of all polyoxometalates (Nb₁₀, TiNb₉, Ti₂Nb₈, Nb₆ and Ta₆) in water onto silicon wafers, and annealed the films at 800 or 1000 °C in air. Thereafter, the films were characterised by Raman spectroscopy, SEM and AFM.

Raman spectra of Nb₁₀ films showed that mainly ortho-Nb₂O₅ is obtained after annealing at 800 °C, as indicated by the broad peak at ca. 690 cm⁻¹ (Fig. 3a). Annealing at 1000 °C, similarly yielded mono-Nb₂O₅ as indicated by the Nb=O stretch at ca. 993 cm⁻¹. The crystal form of the film can thus be tuned by the choice of annealing temperature.

Fig. 3 (a) Raman spectra of [Nb₁₀O₂₈]⁶⁻ films annealed at 800 °C and 1000 °C, (b) Raman spectra and (c) PXRD patterns of powdered [Nb₁₀O₂₈]⁶⁻ annealed at 200–1000 °C and (d) ortho-Nb₂O₅ and mono-Nb₂O₅ polymorphs of Nb₂O₅. Red spheres represent oxygen atoms and light blue polyhedra niobium atoms. See ESI (Fig. S2-5†) for Raman signals of powdered [Nb₁₀O₂₈]⁶⁻ annealed at 200–400 °C.

Powders of pure Nb₁₀ were also annealed, to see whether this yielded the same products as annealed films. The Raman spectra of annealed powders showed the same behaviour as the thin films, with ortho- and mono-Nb₂O₅ dominating at 800 and 1000 °C, respectively (Fig. 3b). There is no evidence of residual TMA counter ion in either powder or film, indicating that it has been fully combusted during annealing. Powder XRD further confirmed the assignment of the crystal phases at 800 and 1000 °C, as well as showing that powders annealed at 200 and 400 °C remained amorphous (Fig. 3c). This agrees with previous results that determined amorphous Nb₂O₅ crystallisation commences at temperatures >400 °C.⁶¹Fig. 3d depicts the difference in packing of NbO₆ octahedra within Nb₂O₅ crystal lattice of ortho- and mono-Nb₂O₅. Films of Nb₆ behaved broadly the same as films of Nb₁₀, yielding ortho-Nb₂O₅ and mono-Nb₂O₅ at 800 °C and 1000 °C, respectively (Fig. S2-2†).

Raman spectra of films of TiNb₉ annealed at 800 and 1000 °C looked very similar to those obtained from Nb₁₀. However, a narrow band at ca. 997 cm⁻¹ was observed at both temperatures (Fig. 4a). The thin film spectra were the same as those obtained from annealed TiNb₉ powder (Fig. 4b), which allowed us to determine the composition of the film using PXRD to be ca. 80% 1 : 1 orthorhombic and monoclinic Ti₂Nb₁₂O₂₉ (Fig. 4c), as well as 20% 1 : 1 orthorhombic and monoclinic Nb₂O₅, at both 800 °C and 1000 °C. The source of the Nb₂O₅ is two-fold: the titanium : niobium (Ti : Nb) ratio in Ti₂Nb₁₂O₂₉ is higher than in the starting material, and Nb₆ is present as a minor impurity in the one-pot synthesis of TiNb₉ (Fig. S7-4†). The peak at ca. 997 cm⁻¹ in the Raman spectrum was assigned to Nb=O symmetric stretching and the broad band in the 600 cm⁻¹ region is shifted in Ti₂Nb₁₂O₂₉ relative to Nb₂O₅, owing to the substitution of some of the NbO₆ octahedra by TiO₆ (Fig. 4d).

Fig. 4 (a) Raman spectra of [TiNb₉O₂₈]⁷⁻ films annealed at 800 °C and 1000 °C, (b) Raman spectra and (c) PXRD patterns of powdered [TiNb₉O₂₈]⁷⁻ annealed at selected temperatures and (d) the monoclinic Ti₂Nb₁₂O₂₉ titanium–niobium oxide. Red spheres represent oxygen atoms, light blue and grey octahedra represent niobium and titanium atoms, respectively. See ESI (Fig. S2-5†) for Raman signals of powdered [TiNb₉O₂₈]⁷⁻ annealed at 200–400 °C.

Films of Ti₂Nb₈ annealed at 800 °C and 1000 °C also exhibited Raman spectra that were consistent with the spectra of annealed of Ti₂Nb₈ powders at the same temperatures (Fig. 5a and b). The Raman peak at ca. 1000 cm⁻¹ is symptomatic of Nb=O symmetric stretching in TiNb₂O₇ and is consistent with the monoclinic crystal system. Raman peaks at ca. 646 cm⁻¹ and 543 cm⁻¹ are similarly attributable to stretches of NbO₆ octahedra and Nb–O–Nb bonds, respectively. PXRD confirmed the presence of monoclinic TiNb₂O₇ at 800 °C and 1000 °C (Fig. 5c). However, Rietveld refinement showed that the monoclinic TiNb₂O₇ obtained after annealing at 800 °C contained ca. 16% 1 : 1 orthorhombic and monoclinic Ti₂Nb₁₂O₂₉, while ca. 34% 1 : 1 orthorhombic and monoclinic Ti₂Nb₁₂O₂₉ were present after annealing at 1000 °C. The Ti₂Nb₁₂O₂₉ likely stems from thermal decomposition of Ti₂Nb₈ during the annealing process and the Ti : Nb ratio of TiNb₂O₇ is higher than the starting material. Decomposition of Ti₂Nb₈ was confirmed by the observation that treatment of an aqueous solution of Ti₂Nb₈ at 160 °C for one hour yielded signals in the ¹⁷O NMR spectrum that corresponded to a very low concentration of TiNb₉ (Fig. S7-5†). Notably, the shifting of the NbO₆ octahedra stretching to higher frequencies is similarly attributable to the presence of TiO₆ octahedra (substituting for NbO₆ octahedra) within the TiNb₂O₇ crystal lattice (Fig. 5d); much like Ti₂Nb₁₂O₂₉.

Fig. 5 (a) Raman spectra of [Ti₂Nb₈O₂₈]⁸⁻ films annealed at 800 °C and 1000 °C, (b) Raman spectra and (c) PXRD patterns of powdered [Ti₂Nb₈O₂₈]⁸⁻ annealed at selected temperatures and (d) the monoclinic TiNb₂O₇ titanium–niobium oxide. Red spheres represent oxygen atoms, light blue and grey octahedra represent niobium and titanium atoms, respectively. See ESI (Fig. S2-5†) for Raman signals of powdered [Ti₂Nb₈O₂₈]⁸⁻ annealed at 200–400 °C.

In contrast, annealing spin-coated films of Ta₆ did not yield films with Raman spectra that agreed with their annealed powder counterparts, mainly because Ta₆ films annealed at 800 °C and 1000 °C were Raman silent (Fig. 6a; only the background signal from the silicon wafer is observed). The Raman spectrum of powder obtained from annealed Ta₆ at 800 °C and 1000 °C showed that it consisted of ortho-Ta₂O₅ (Fig. 6b), an assignment which was confirmed by PXRD (Fig. 6c). The broad peak at 895 cm⁻¹ originates from the stretching of Ta–O bonds with different bond strengths due to corner/edge-sharing in tantalum oxide polyhedra. ortho-Ta₂O₅ exhibits a Raman peak at 845 cm⁻¹ that is attributable to Ta–O stretching of TaO₆ and TaO₇ (pentagonal bipyramidal) polyhedra. The Raman peaks at 620 cm⁻¹ and 700 cm⁻¹ are from stretches of TaO₆ octahedra, with the stretching at 700 cm⁻¹ arising from Ta–O bonds with different bond strengths due to corner/edge-sharing polyhedra. The structure of crystalline ortho-Ta₂O₅ (Fig. 6d) depicts the different TaO₆ and TaO₇ polyhedra present in the crystal lattice.

Fig. 6 Raman spectra of [Ta₆O₁₉]⁸⁻ films annealed at 800 °C and 1000 °C, (b) Raman spectra and (c) PXRD patterns of powdered [Ta₆O₁₉]⁸⁻ annealed at selected temperatures and (d) the ortho-Ta₂O₅ polymorph of Ta₂O₅. Red spheres represent oxygen atoms and dark blue polyhedra represent tantalum atoms. See ESI (Fig. S2-5†) for Raman signals of powdered [Ta₆O₁₉]⁸⁻ annealed at 200–400 °C.

The lack of a Raman signal from the Ta₆ films was attributed to attenuation specific to the tantalum oxide films. This was confirmed by depositing films of solutions of 1 : 1 Nb₆: Ta₆, which yielded strongly attenuated spectra relative to pure Nb₆, whereas powders of the same solution showed no such attenuation (Fig. S2-6†).

Decomposition of the TMA counter ion in the annealed films was probed through energy-dispersive X-ray spectroscopy (EDS) (Fig. 7). The elemental composition of C after annealing at 400 °C decreases below 10%, and did not exceed 1% after annealing at 800–1000 °C. This correlates to rapid decomposition of the TMA counter ion at 200–400 °C and the continued decomposition of any carbon residues (char) at higher temperatures. There is a notable increase in the metal (Nb, Ti or Ta) composition of the annealed films at 200–400 °C, in agreement with decomposition of TMA, which later plateaus at 800–1000 °C.

Fig. 7 Atomic composition of annealed Nb₁₀ (a), TiNb₉ (b), Ti₂Nb₈ (c), and Ta₆ (d) films. See ESI (Table S5-1†) for details of full atomic composition. The carbon content decreases with annealing temperature in all cases.

The Ti content of annealed Ti₂Nb₈ films is higher than the Ti composition of annealed TiNb₉, which corroborates the higher Ti content of Ti₂Nb₈ films. A spatial distribution of Nb and Ti in TiNb₉ and Ti₂Nb₈ films annealed at 1000 °C, obtained from EDS mapping, is used to illustrate this in Fig. 8. Notably, the distribution of the elements confirms that the films are homogeneous.

Fig. 8 EDS map of the spatial distribution of Nb (a), Ti (b) and Nb/Ti (c) in TiNb₉, and Nb (d), Ti (e) and Nb/Ti (f) in Ti₂Nb₈ films annealed at 1000 °C.

Multiple depositions of films

We also investigated repeated deposition of layers of polyoxometalates. Nb₁₀ was deposited on silicon wafers that had been prepared by plasma treatment, and then annealed. At 800 °C, the film after a single deposition consisted of ortho-Nb₂O₅, had a thickness of 147 ± 5 nm according to ellipsometry. This film was again treated by ambient pressure plasma, and a new layer of Nb₁₀ was spin-coated onto the first one. Pre-treatment of the film by plasma was necessary to ensure good adhesion between the layers. This process of repeated spin-coating and plasma treatment was repeated to deposit four successive layers of film, and we did this with all five polyoxometalates included in this study. After four layers of Nb₁₀ had been spin-coated onto the silicon wafer, a final thickness of 235 ± 5 nm, as determined by SEM, was obtained when annealing at 800 °C (Fig. 9a). Notably, SEM showed that the film possessed voids with no obvious layering, and the films were determined to be smooth (roughness of ±24 nm) via AFM.

Fig. 9 Cross-section images for annealing: Nb₁₀ at 800 °C (a) and 1000 °C (b); TiNb₉ at 800 °C (c) and 1000 (d); Ti₂Nb₈ at 800 °C (e) and 1000 (f) films after multiple depositions. Films annealed at 800 °C showed no layering, while the films annealed at 1000 °C exhibited two layers above the silicon substrate.

In contrast, when successive deposition was carried out with annealing at 1000 °C, a film of mono-Nb₂O₅ with a total thickness of 642 ± 46 nm was obtained (Fig. 9b). Here, two layers were observed, with a lower layer of thickness 199 ± 6 nm, and an upper layer with thickness 442 ± 44 nm. Voids were also apparent, but only in the upper layer and it had a roughness of ±55 nm. The presence of a distinct first layer in the mono-Nb₂O₅ film is indicative of successive layered deposition and subsequent stacking of the mono-Nb₂O₅ film. Though stacking of layers is not discernible for the ortho-Nb₂O₅ film, the larger thickness of this film (235 ± 5 nm), compared to the first layer of the ortho-Nb₂O₅ film (147 ± 5 nm), may be due to coalescence of the successively deposited layers of the ortho-Nb₂O₅ film. The increase in roughness as the films were annealed at 1000 °C is attributable to enhanced grain growth at higher temperatures, which occurs simultaneously with the phase change.

Layering of TiNb₉ and Ti₂Nb₈ films was also studied (Fig. 9c–f) and the thickness and roughness are summarised in Table S4-1.† After four successive depositions of TiNb₉ and Ti₂Nb₈ followed by annealing at 800 °C, Ti₂Nb₁₂O₂₉ and TiNb₂O₇ films were obtained, respectively, with no discernible layering, as with Nb₁₀. Furthermore, as with Nb₁₀, two layers were obtained at 1000 °C for both titanium–niobium oxide films and roughness increased for Ti₂Nb₁₂O₂₉ (±35.3 nm) but not TiNb₂O₇ (±13.8 nm). For both Ti₂Nb₁₂O₂₉ and TiNb₂O₇ films annealed at 1000 °C, the lower layers of the films were homogeneous and free of voids, while the voids were apparent only in the upper layer, like with Nb₁₀.

Successive deposition of four layers of Nb₆ and Ta₆ films annealed at 800 °C also yielded ortho-Nb₂O₅ and ortho-Ta₂O₅ films, respectively, which did not exhibit layering. The ortho-Nb₂O₅ film had voids (Fig. 10a) while the ortho-Ta₂O₅ film was homogeneous throughout (Fig. 10b). Furthermore, the Nb₆ film annealed at 1000 °C produced a layered mono-Nb₂O₅ film – similarly to Nb₁₀ – with both layers of film being homogeneous. The Ta₆ film annealed at 1000 °C also formed a layered ortho-Ta₂O₅ film that was smooth and free of voids in both layers. The tendency for Ta₆ to form smooth, homogeneous Ta₂O₅ films is likely a consequence of its tendency to gel, instead of crystallising like Nb₆, as observed with the removal of water during rotary evaporation.²⁷

Fig. 10 Cross-section images for annealing: Nb₆ at 800 °C (a) and 1000 °C (b); Ta₆ at 800 °C (c) and 1000 °C (d). The tantalate films are noticeably smoother and more homogeneous than the niobate films.

The ortho-Nb₂O₅ film possessed more voids than the mono-Nb₂O₅ film, when comparing the ortho-Nb₂O₅ film as a whole to the lower layer of the mono-Nb₂O₅ film. In the solid state, Si–O–Nb bond formation is encouraged at the silica–Nb₂O₅ interface at higher temperatures.^62,63 Therefore, more Si–O–Nb bonds will form when annealing the Nb₂O₅ film at 1000 °C, compared to 800 °C. This contributes to film densification at higher temperatures. The homogeneity of the lower layer of mono-Nb₂O₅ film in contrast to the presence of voids in the ortho-Nb₂O₅ film and in the upper layer of the mono-Nb₂O₅ film, is thus likely due to poor densification of the latter.

The formation of a distinct first layer in Ti₂Nb₁₂O₂₉ and TiNb₂O₇ is similarly likely dependent on Si–O–Nb bond formation. However, Si–O–Ti bonds are considered to contribute to the binding of the film to the substrate as complete incorporation of Ti into the silica network occurs at high temperatures (1000 °C).⁶⁴ The prevalence of this bonding is inherently lower than that of the Si–O–Nb bonds, based on the partial occupancy of Ti atoms within the crystal lattice of Ti₂Nb₁₂O₂₉ and TiNb₂O₇.^41,42 The presence of the Ti atoms within the crystal lattice contributes to formation of Nb–O–Ti bonds, which also facilitate the adhesion of subsequent layers of the titanium–niobium oxides and supports film densification. Hence, the upper layers of the Ti₂Nb₁₂O₂₉ and TiNb₂O₇ films annealed at 1000 °C possessed fewer voids than the analogous layer of mono-Nb₂O₅ film.

Conclusions

The controlled deposition of metal oxide thin films is possible via an iterative spin-coating and annealing process of POMs. This approach allows deposition of alkali-free, crystalline metal oxide thin films with tuneable thickness and roughness. Notably, the optical and energy storage properties of metal oxide films is dependent on their crystal structures. The TMA salts of Nb₆, Ta₆, Nb₁₀, TiNb₉ and Ti₂Nb₈ clusters allow the deposition of orthorhombic and monoclinic Nb₂O₅, Ta₂O₅, Ti₂Nb₁₂O₂₉ and TiNb₂O₇ from aqueous solutions, depending on precursor and conditions. The substitution of a combustible organic counter cation, here tetramethylammonium, for the more common alkali–metal counter ions not only allows for the deposition of alkali-free metal oxide films, but also allows for increased flexibility in the tuning of the composition and properties of the films. Furthermore, the ability to deposit films with tuneable crystallinity is crucial to developing high quality, pristine metal oxide thin films for advancing nanotechnology and semiconductor development. We also showed that Raman spectroscopy is a powerful and convenient method of determining the phase purity of niobium pentoxide (orthorhombic or monoclinic).

The inclusion of heterometal atoms in polyoxometalates leads to different solid-state structures, even when the difference in substitution is limited to a single atom. The site-specific inclusion of Ti within the decametalate structure is exemplary of this and it can be inferred that the site-specific inclusion of other heterometals may elicit a similar effect. There is thus great potential for developing this technique to include other types of heterometals, such as those that have been shown to be catalytically active in different reactions. By including these in a chemically inert substrate such as niobium pentoxide, chemical processes may be adapted to take place in environments that would otherwise be too demanding.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

C. A. O thanks the Kempe Foundation (JCK-2029.1) and the Swedish Research Council (2018-07039) for generous support. M. A. R thanks the Department of Chemistry at Umeå University for a postgraduate fellowship. Prof. William H. Casey at UC Davis and Dr Sven Albrecht from TANIOBIS GmbH, are equally thanked for generous gifts of hydrous Nb₂O₅ and hydrous Ta₂O₅, respectively. We also thank Prof. Markus Broström for providing a furnace for annealing experiments, and Drs Hussein Kanbar and Cheng Choo Lee (Umeå Core Facility of Electron Microscopy) for assisting with PXRD and FIB-SEM/EDS measurements, respectively. Support towards the Vibrational Spectroscopy Core Facility by the Chemical Biological Centre and the Department of Chemistry at Umeå University is greatly acknowledged.

References

1. Y. Pozdeev, Qual. Reliab. Eng. Int., 1998, 14, 79–82.
2. D. Pradhan, A. W. Wren, S. T. Misture and N. P. Mellott, Mater. Sci. Eng., C, 2016, 58, 918–926.
3. N. Wang, H. Li, J. Wang, S. Chen, Y. Ma and Z. Zhang, ACS Appl. Mater. Interfaces, 2012, 4, 4516–4523.
4. J. Ge, F. Wang, Z. Xu, X. Shen, C. Gao, D. Wang, G. Hu, J. Gu, T. Tang and J. Wei, J. Mater. Chem. B, 2020, 8, 2618–2626.
5. Y. Guo, K. Xie, W. Jiang, L. Wang, G. Li, S. Zhao, W. Wu and Y. Hao, ACS Biomater. Sci. Eng., 2019, 5, 1123–1133.
6. Q. Zhao, R. O. Behunin, P. T. Rakich, N. Chauhan, A. Isichenko, J. Wang, C. Hoyt, C. Fertig, M. h. Lin and D. J. Blumenthal, APL Photonics, 2020, 5(116103), 1–8.
7. N. Maruthi, M. Faisal, N. Raghavendra, B. P. Prasanna, S. R. Manohara and M. Revanasiddappa, Synth. Met., 2021, 275(116744), 1–9.
8. M. Dinu, L. Braic, S. C. Padmanabhan, M. A. Morris, I. Titorencu, V. Pruna, A. Parau, N. Romanchikova, L. F. Petrik and A. Vladescu, J. Mech. Behav. Biomed. Mater., 2020, 103, 103582.
9. Y. Ge, Y. Wang, J. Chen, Y. Zou, L. Guo, J. Ouyang, D. Jia and Y. Zhou, J. Alloys Compd., 2018, 767, 7–15.
10. W. Hu, J. Xu, X. Lu, D. Hu, H. Tao, P. Munroe and Z.-H. Xie, Appl. Surf. Sci., 2016, 368, 177–190.
11. J. Xu, X. k. Bao, T. Fu, Y. Lyu, P. Munroe and Z.-H. Xie, Ceram. Int., 2018, 44, 4660–4675.
12. J. W. Mantz, Global Studies Review, 2008, 4, 12–14.
13. T. Ibn-Mohammed, S. C. L. Koh, I. M. Reaney, A. Acquaye, D. Wang, S. Taylor and A. Genovese, Energy Environ. Sci., 2016, 9, 3495–3520.
14. D. Fast, M. Clark, L. Fullmer, K. Grove, M. Nyman, B. Gibbons and M. Dolgos, Thin Solid Films, 2020, 710(138270), 1–6.
15. M. Nyman, L. J. Criscenti, F. Bonhomme, M. A. Rodriguez and R. T. Cygan, J. Solid State Chem., 2003, 176, 111–119.
16. M. Nyman, Dalton Trans., 2011, 40, 8049–8058.
17. L. Zhang and Z. Chen, Int. J. Energy Res., 2020, 44, 3316–3346.
18. H.-Y. Zhao, Y.-Z. Li, J.-W. Zhao, L. Wang and G.-Y. Yang, Coord. Chem. Rev., 2021, 443(213966), 1–23.
19. M. Amiri, N. P. Martin, C. L. Feng, J. K. Lovio and M. Nyman, Angew. Chem., Int. Ed., 2021, 60, 12461–12466.
20. N. P. Martin and M. Nyman, Angew. Chem., Int. Ed., 2021, 60, 954–960.
21. N. P. Martin, E. Petrus, M. Segado, A. Arteaga, L. N. Zakharov, C. Bo and M. Nyman, Chem. – Eur. J., 2019, 25, 10580–10584.
22. C. A. Ohlin, E. M. Villa, J. C. Fettinger and W. H. Casey, Angew. Chem., Int. Ed., 2008, 47, 5634–5636.
23. C. A. Ohlin, E. M. Villa, J. C. Fettinger and W. H. Casey, Dalton Trans., 2009, 2677–2678,  10.1039/B900465C.
24. J.-H. Son and W. H. Casey, Chem. – Eur. J., 2016, 22, 14155–14157.
25. J.-H. Son, J. Wang and W. H. Casey, Dalton Trans., 2014, 43, 17928–17933.
26. M. A. Rambaran, M. Pascual-Borràs and C. A. Ohlin, Eur. J. Inorg. Chem., 2019, 3913–3918,  DOI:10.1002/ejic.201900750.
27. L. B. Fullmer, R. H. Mansergh, L. N. Zakharov, D. A. Keszler and M. Nyman, Cryst. Growth Des., 2015, 15, 3885–3892.
28. R. H. Mansergh, L. B. Fullmer, D.-H. Park, M. Nyman and D. A. Keszler, Chem. Mater., 2016, 28, 1553–1558.
29. P. Amaravathy, S. Sowndarya, S. Sathyanarayanan and N. Rajendran, Surf. Coat. Technol., 2014, 244, 131–141.
30. V. K. Balla, S. Banerjee, S. Bose and A. Bandyopadhyay, Acta Biomater., 2010, 6, 2329–2334.
31. V. K. Balla, S. Bodhak, S. Bose and A. Bandyopadhyay, Acta Biomater., 2010, 6, 3349–3359.
32. M. Grobelny, M. Kalisz, M. Mazur, D. Wojcieszak, D. Kaczmarek, J. Domaradzki, M. Świniarski and P. Mazur, Thin Solid Films, 2016, 616, 64–72.
33. E. I. Ko and J. G. Weissman, Catal. Today, 1990, 8, 27–36.
34. R. A. Rani, A. S. Zoolfakar, A. P. O'Mullane, M. W. Austin and K. Kalantar-Zadeh, J. Mater. Chem. A, 2014, 2, 15683–15703.
35. H. Schäfer, R. Gruehn and F. Schulte, Angew. Chem., Int. Ed Engl., 1966, 5, 40–52.
36. S. Lagergren and A. Magnéli, Acta Chem. Scand., Ser. A, 1952, 6, 444–446.
37. N. C. Stephenson and R. S. Roth, Acta Crystallogr., Sect. B: Struct. Sci., 1971, 27, 1037–1044.
38. N. C. Stephenson and R. S. Roth, J. Solid State Chem., 1971, 3, 145–153.
39. I. P. Zibrov, V. P. Filonenko, M. Sundberg and P.-E. Werner, Acta Crystallogr., Sect. B: Struct. Sci., 2000, 56, 659–665.
40. C. A. Ohlin, E. M. Villa and W. H. Casey, Inorg. Chim. Acta, 2009, 362, 1391–1392.
41. A. D. Wadsley, Acta Crystallogr., 1961, 14, 664–670.
42. A. D. Wadsley, Acta Crystallogr., 1961, 14, 660–664.
43. R. B. V. Dreele, A. K. Cheetham and J. S. Anderson, Proc. R. Soc. London, Ser. A, 1974, 338, 311–326.
44. X. Wu, J. Miao, W. Han, Y.-S. Hu, D. Chen, J.-S. Lee, J. Kim and L. Chen, Electrochem. Commun., 2012, 25, 39–42.
45. N. G. Eror and U. Balachandran, J. Solid State Chem., 1982, 45, 276–279.
46. S. Deng, Z. Luo, Y. Liu, X. Lou, C. Lin, C. Yang, H. Zhao, P. Zheng, Z. Sun, J. Li, N. Wang and H. Wu, J. Power Sources, 2017, 362, 250–257.
47. S. Lou, X. Cheng, J. Gao, Q. Li, L. Wang, Y. Cao, Y. Ma, P. Zuo, Y. Gao, C. Du, H. Huo and G. Yin, Energy Storage Mater., 2018, 11, 57–66.
48. K. J. Griffith, Y. Harada, S. Egusa, R. M. Ribas, R. S. Monteiro, R. B. Von Dreele, A. K. Cheetham, R. J. Cava, C. P. Grey and J. B. Goodenough, Chem. Mater., 2020, 33, 4–18.
49. Y. Yang and J. Zhao, Adv. Sci., 2021, 8(2004855), 1–24.
50. D. Pham-Cong, J. Kim, V. T. Tran, S. J. Kim, S.-Y. Jeong, J.-H. Choi and C. R. Cho, Electrochim. Acta, 2017, 236, 451–459.
51. C. Yang, S. Yu, Y. Ma, C. Lin, Z. Xu, H. Zhao, S. Wu, P. Zheng, Z.-Z. Zhu, J. Li and N. Wang, J. Power Sources, 2017, 360, 470–479.
52. A. Llordés, G. Garcia, J. Gazquez and D. J. Milliron, Nature, 2013, 500, 323–326.
53. A. Llordés, A. T. Hammack, R. Buonsanti, R. Tangirala, S. Aloni, B. A. Helms and D. J. Milliron, J. Mater. Chem., 2011, 21, 11631–11638.
54. A. Llordés, Y. Wang, A. Fernandez-Martinez, P. Xiao, T. Lee, A. Poulain, O. Zandi, C. A. Saez Cabezas, G. Henkelman and D. J. Milliron, Nat. Mater., 2016, 15, 1267–1273.
55. C. A. Saez Cabezas, K. Miller, S. Heo, A. Dolocan, G. LeBlanc and D. J. Milliron, Chem. Mater., 2020, 32, 4600–4608.
56. H. Chaudhary, I. A. Iashchishyn, N. V. Romanova, M. A. Rambaran, G. Musteikyte, V. Smirnovas, M. Holmboe, C. A. Ohlin, Z. M. Svedruzic and L. A. Morozova-Roche, ACS Appl. Mater. Interfaces, 2021, 13, 26721–26734.
57. E. M. Villa, C. A. Ohlin, E. Balogh, T. M. Anderson, M. D. Nyman and W. H. Casey, Angew. Chem., Int. Ed., 2008, 47, 4844–4846.
58. E. M. Villa, C. A. Ohlin, E. Balogh, T. M. Anderson, M. D. Nyman and W. H. Casey, Am. J. Sci., 2008, 308, 942–953.
59. E. M. Villa, C. A. Ohlin and W. H. Casey, J. Am. Chem. Soc., 2010, 132, 5264–5272.
60. E. M. Villa, C. A. Ohlin, J. R. Rustad and W. H. Casey, J. Am. Chem. Soc., 2009, 131, 16488–16492.
61. J. M. Jehng and I. E. Wachs, Chem. Mater., 1991, 3, 100–107.
62. M. S. P. Francisco and Y. Gushikem, J. Mater. Chem., 2002, 12, 2552–2558.
63. S. Morselli, P. Moggi, D. Cauzzi and G. Predieri, in Stud. Surf. Sci. Cat, ed. B. Delmon, P. A. Jacobs, R. Maggi, J. A. Martens, P. Grange and G. Poncelet, Elsevier, 1998, vol. 118, pp. 763–772.
64. C. C. Perry, X. Li and D. N. Waters, Spectrochim. Acta, Part A, 1991, 47, 1487–1494.

---

Footnote

† Electronic supplementary information (ESI) available: Details related to synthesis and film deposition, Raman, PXRD, ellipsometry, ¹⁷O-NMR, EDS, SEM and AFM measurements. See DOI: 10.1039/d1dt03116c

---