Essential role of treatment in hydrogen atmosphere on creation of ferromagnetic order in carbon incorporated TiO₂ nanocomposite

Titanium oxide (TiO₂) is widely used in many applications. The physical and chemical properties cause TiO₂ to contribute in different fields such as photovoltaics, photocatalysis, gas sensing, pigments, protective coatings, dielectric materials, etc, [1–3]. It can have excellent photocatalytic properties depending on synthesis details including kinds of controlled chemical impurities [3, 4]. However, it has a large energy gap (∼3.2 eV) in the UV spectral region. Therefore, in order to let TiO₂ functioning as photocatalytic in visible light, its optical band gap has to be decreased. The control of band gaps is usually performed by suitable doping. Moreover, the doping process could create in TiO₂ oxide exotic properties that widen its applications. For instance, the doping by transition/rare-earth ions could create room-temperature ferromagnetic (RT-FM) properties. Occasionally the doping with nonmagnetic ions can also create RT-FM. Carbon ions were used at certain conditions as dopant ions to create RT-FM in many transparent conducting oxides (TCOs) like ZnO, In₂O₃, SnO₂, etc [5–9]. Many Authors attributed the creation of FM in C-doped ZnO material to the substitution of structural O by C (C_O) ions i.e. the magnetism emerged due to the Zn–C system in ZnO environment and the magnetic moment is contributed mainly by the carbon 2p orbitals (0.80 μB) [5]. In general, the reasons for the creation of RT-FM in C-incorporated TCOs are not clear. The present work deals with the effects of C incorporation on the structural, optical and magnetic properties of host TiO₂ nanocomposite, focusing on the conditions necessary to create RT-FM.

Pure and C-incorporated TiO₂ (TiO₂–C) nanocomposite powders were synthesized by using pure titanium tetrachloride (TiCl₄) liquid and refined sucrose (C₁₂H₂₂O₁₁) (Sigma-Aldrich products) as starting complexes. The preparation was carried out by a simple thermo-chemical process including co-decomposition of the complexes. To synthesis pure TiO₂, a precursor of 3 ml of TiCl₄ liquid mixed with 30 ml of pure ethanol was prepared in a beaker. The solution was magnetically stirred for ∼30 min at RT before the addition of the ∼120 ml of deionized distilled water. Then, the temperature of the main solution was raised slowly to ∼60 °C with stirring for ∼20 h until a dry powder was formed in the beaker. The powder was collected and a flash sintering in the air at 500 °C/2 h followed by cool with the oven to the RT. The formed powder was found to be pure TiO₂. To synthesis C incorporated TiO₂, a controlled amount of sucrose as a source of carbon was added to the main solution. Three nominal (C/Ti) mass ratio powders were synthesized: 1%, 2% and 8.5%, which were referred as TiO₂–Cn (n = 1, 2, 3), i.e. TiO₂–C1, TiO₂–C2, and TiO₂–C3, respectively. Moreover, samples of each powder were annealed in a hydrogen atmosphere (hydrogenated) at 400 °C for 20 min and referred to TiO₂–Cn–H, i.e. TiO₂–C1–H, TiO₂–C2–H, TiO₂–C3–H, respectively. Pure TiO₂ and hydrogenated TiO₂–H were prepared as references. All the prepared samples were pelletized for characterization.

The structural analysis was performed by the x-ray diffraction (XRD) with a Rigaku Ultima-IV diffractometer. The optical properties over the range 200–1000 nm were studied by the diffuse reflectance spectroscopy (DRS) method with a Shimadzu UV-3600 spectrophotometer equipped with an integrating sphere. The magnetic properties were measured by the vibrating sample magnetometer (VSM) method with a PMC MicroMag-3900 over the range ±10 kOe at a step of 25 Oe/1 s at RT.

TiO₂ crystallizes in three main structures; the stable phase (rutile-R) and two metastable phases (anatase and brookite). The brookite phase is usually difficult to synthesize [10]. Figure1(a) shows the XRD patterns of pure and C-incorporated TiO₂ powder samples. Figure1(b) shows the XRD of the hydrogenated samples. The structural analyses revealed a formation of anatase (A) phase with amounts of rutile (R) and brookite (B). The Bragg reflections were indexed according to the early known crystallographic data [11, 12]:

$$\begin{eqnarray*}&&\begin{array}{l}\begin{array}{l}\begin{array}{l}{\rm{Anatase}}\left({\rm{A}}\right):\,\mathrm{tetragonal},\,{\rm{SG}}:\,{\rm{I}}{4}_{1}/{\rm{amd}}\,\left({\rm{No}}.\,141\right),a={\rm{b}}={\rm{0.3785}}\,{\rm{nm}},\\ {\rm{c}}=0.9514\,{\rm{nm}}{\rm{.}}\end{array}\\ \begin{array}{l}{\rm{Rutile}}\,\left({\rm{R}}\right):\,{\rm{tetragonal}},\,{\rm{SG}}:\,{\rm{P}}{4}_{2}/{\rm{mnm}}\,\left({\rm{No}}.136\right),{\rm{a}}={\rm{b}}=0.4594\,\mathrm{nm},\\ {\rm{c}}=0.2959\,{\rm{nm}}{\rm{.}}\end{array}\end{array}\\ {\rm{Brookite}}\,\left({\rm{B}}\right):\,{\rm{orthorhombic}},\,{\rm{SG}}:\,{\rm{Pbca}}\,\left({\rm{No}}.61\right),{\rm{a}}=0.9184\,{\rm{nm}},\\ {\rm{b}}=0.5447\,{\rm{nm}},{\rm{c}}=0.5145\,{\rm{nm}}{\rm{.}}\end{array}\end{eqnarray*}$$

Zoom In Zoom Out Reset image size

However, it is not easy to find the exact % content of each phase in the synthesized powder sample due to the strong XRD signal overlapping. For instance, the main anatase (101) diffraction peak at 2θ = 25.28° overlaps with the brookite (120) and (111) peaks at 25.34° and 25.69°, respectively [13]. The Rietveld refinement technique was used by the built-in PDXL software to analyze the crystalline structure and find the lattice parameters. The average crystallite size (CS) was calculated through Hadler-Wagner (H-W) equation [14]: ${({\beta }_{hkl}/\tan {\theta }_{hkl})}^{2}=\left(k\lambda /D\right)({\beta }_{hkl}/\,\tan \,{\theta }_{hkl}\,\sin \,{\theta }_{hkl})+16{\varepsilon }^{2},$ where θ is the Bragg's angle, k = 4/3, λ_Κα = 0.154 nm, D is the CS, β_hkl (rad) is the reflection peak width and ε is the microstrain. A graph of ${({\beta }_{hkl}/\tan {\theta }_{hkl})}^{2}$ versus $({\beta }_{hkl}/\,\tan \,{\theta }_{hkl}\,\sin \,{\theta }_{hkl})$ should be a straight line and the results are presented in the table 1 including the found Rietveld refinement parameters; the weighted profile factor (R_wp) and the goodness-of-fit (S). The values of S were 1–2 indicating good fitting results.

For pure TiO₂, figure 1 shows very weak reflections of R and B phases, therefore, the XRD pattern revealed mainly the crystallization of A phase. However, the intensity of reflections of the R phase increased by increasing of C content in the sample, referring to increase the R phase in the samples. Moreover, figure 1 shows the almost constant relative intensity of (121) reflection of phase B, which refers to roughly constant relative % content of phase B in the samples. Meanwhile, the phase R content significantly increased with increasing of C inclusion. A similar result has been previously observed, that the incorporation of C into TiO₂ lattice enhanced A to R transformation through the creation of O-vacancies [10]. Therefore, the data analyses presented in the table 1 were obtained by considering that the samples have consisted of a mixture of A and R phases. Moreover, no diffraction peaks of pure carbon or any related compounds were detected. Thus, the crystalline traces of secondary phases were not found. The quantitative ratios of R/A as calculated by PDXL program were ∼4%, 25%, 31%, and 49% for TiO₂, TiO₂–C1, TiO₂–C2, TiO₂–C3, respectively.

Figure 1(b) shows the XRD patterns of the hydrogenated TiO₂–Cn–H samples. For all the samples, any structural modification by the hydrogenation was not detected. However, the % amount of R phase significantly increased by the hydrogenation so that the quantitative ratios of R/A were ∼11%, 163%, 117%, and 72% for TiO₂–H, TiO₂–C1–H, TiO₂–C2–H, TiO₂–C3–H, respectively. The hydrogenation can be considered by some of its actions, like low-oxygen partial pressure treatment, which could generate a high concentration of oxygen vacancies (V_O) and enhance A to R transformation [10].

Data of table 1 show that the CS of A and R phases increased by C inclusion attaining the largest values with TiO₂–C2 sample. The unit-cell volume (V_cell) slightly varied within the experimental accuracy due to the small size of the incorporated C ion-species. The hydrogenation did not introduce measurable systematic variations in V_cell and CS.

The type of C inclusion into the TiO₂ lattice can be studied through geometrical principles. The ionic radius of C⁴⁺(0.016 nm) is smaller than that of Ti⁴⁺(0.0605 nm [15]) by ∼3.8 times, therefore, the substitution of C for Ti ions (Ti_C) is not favourable, since it would create a strong crystalline distortion, according to Hume-Rothery rules [16]. Thus, the inserted small C^v+ (where v ≤ 4) ion-species could mainly occupy crystalline interstitial positions. Such occupation must create O-vacancies due to the re-distribution of static point charges.

DRS technique was used to characterize the prepared pure and C-incorporated TiO₂ nanocomposite powders. The measured DR spectra (R_ef (λ)) were studied through Kubelka-Munk (K-M) equation;$(K/S)=F(\lambda )\,={\left(1-{R}_{ef}\right)}^{2}/2{R}_{ef}$ where K is the K-M absorption coefficient; K = 2α, α is the linear absorption coefficient and F (λ) is the spectral K-M absorption function. The scattering coefficient (S) can be considered approximately constant, therefore, F (λ) ∼ α. The graphical dependence of the spectral K-M absorption function, F(λ) are demonstrated in figures 2(a), (b). The optical band gap (E_g) can be estimated through the absorption threshold wavelength (λ_g). Each optical band gap was estimated by E_g(eV) = 1242.3/ λ_g(nm) and the results are listed in table 1. For pure TiO₂, the estimated optical band gap (3.1 eV) is in agreement with the known published values. Thus, the optical band gap was red-shifted by carbon incorporation, which refers to the insertion of carbon ions into the crystalline lattice of TiO₂. Such optical band gap narrowing (BGN) of TiO₂ was also observed in many investigations [17, 18]. This BGN was attributed to the creation of structural point defects like oxygen vacancies (V_Os) of energy levels located close the bottom and overlapped with the conduction band of host TiO₂ causing a slight BGN.

Zoom In Zoom Out Reset image size

The data presented onto the table 1 show that the hydrogenation of pure and C-incorporated TiO₂ samples slightly widen the corresponding optical gap (BGW). Such BGW was also observed in [19]. Generally, the BGW in TCOs is attributed to the increase in the concentration of conduction electrons, according to Moss-Burstein effect [20]. Moreover, the blue-shift is also resulted by the improvement of the crystallinity since the hydrogenation reduces the density of structural defects by removing many types of structural defects, imperfections, and dangling atoms/ions, etc. Moreover, the hydrogenation would homogenize the incorporated C distribution throughout TiO₂ lattice via tossing them. Therefore, the observed results refer to the resultant consequences of the hydrogenation process.

Pristine TiO₂ has diamagnetic (DM) properties [21, 22], however, nanocrystalline TiO₂ synthesized by different procedures might retain some weak magnetic properties due to the created structural defects [23]. The prepared pure TiO₂ in the present work showed weak paramagnetic (PM) properties of susceptibility ∼10⁻⁷ cgs g⁻¹. Such PM properties were attributed to the formatted structural and local electrostatic point defects [23]. By hydrogenation of the pure TiO₂, weak FM properties superimposed on DM behaviors were observed, as shown in figure 3(a). The generation of FM properties by hydrogenation without participating of chemical impurities can be attributed to the creation of structural O-vacancies, which in turns change the oxidation state of structural Ti ions (i.e. formation of electrostatic defects). This conclusion agrees with that proposed in [23] about the formation of local [Ti³⁺–O_6−x] complex defects i.e. the formation of one unpaired 3d-electron creates local magnetic moments.

Zoom In Zoom Out Reset image size

The occupation of C ion-species the interstitial positions of TiO₂–C1 lattice slightly increased the O-vacancies concentration, which slightly enhanced the PM susceptibility of the host TiO₂ (table 2). For further addition of C ions, as in case of TiO₂–C2 and TiO₂–C3 samples, the PM susceptibility was totally defeated resulting of net DM susceptibility (table 2), which was increased in magnitude by increasing of C inclusion (figure 3(b)). The effective magnetic moment per C ion dopant can be estimated from the net (corrected) experimental PM susceptibility (χ_g) of C1 ions, i.e. (6.7–3.4) × 10⁻⁷ cgs g⁻¹ = 3.3 × 10⁻⁷ cgs g⁻¹ by using Curie equation, $\rho {\chi }_{g}={n}_{ion}{\mu }_{C}^{2}/3{k}_{B}\,T,$ where ρ is the density of TiO₂ (3.78 g cm⁻³), n_ion=1.14 × 10²¹ cm⁻³, μ_C is the magnetic moment per C ion, k_B is the Boltzmann constant, and T is the working temperature. Hence, μ_C = 1.24 μ_B/C, which is close to the previously found magnetic moment, (1.5–3.0) μ_B/C [5]. If the localized magnetic moment of 2p of carbon (μ_2p ∼ 0.85 μ_B/C) was considered, then the measured average magnetic moment (1.24 μ_B/C) must involve μ_2p in addition to some contribution of O-vacancies [24].

The hydrogenation of TiO₂–Cn–H (where n = 1, 2 and 3) samples slightly increased the concentration of O-vacancies. Moreover, H₂ molecules could toss some of the C ion-species and drop them into some of the O-vacancies. Similar conclusions have been mentioned in the literature [25]; the annealing under O-poor conditions enhanced the substitution of oxygen by carbon ions (C_O) in addition to the creation of O-vacancies.

Type and strength of spin-spin (S_p. S_p) super-exchange interactions between C dopant ion-species via vacancies in TiO₂–Cn–H samples should depend on the average C–C separation (R_C–C). In the present work, RT-FM was found only with TiO₂–C1–H sample (table 2 and figure 3(c)). The relationship between R_C–C and the possibility of the creation of FM interaction can be concluded according to the following discussions. In general, the strength of the spin-spin (S_p.S_p) exchange interaction between C dopant ions depends on R_C–C and on the properties of the electronic interaction medium, which conduct that exchange interaction. For a uniform distribution of solute C ions of concentration n_C, the R_C–C can be estimated by n_CV = 1, where V = (4/3)πR³_C–C. As the concentration n_C was equal to 1.14 × 10²¹ cm⁻³, 2.28 × 10²¹ cm⁻³, 9.69 × 10²¹ cm⁻³ thus the corresponding R_C–C was equal to 5.9 Å, 4.7 Å and 2.9 Å, for TiO₂–C1–H, TiO₂–C2–H, and TiO₂–C3–H, respectively. The present results showed that only for R_C–C = 5.9 Å (TiO₂–C1–H), the interaction was FM while for lower values (4.7 Å and 2.9 Å) that interaction was converted to DM at RT. Similar observation on magnetic coupling in connection with R_C–C parameter was theoretically concluded in [26], where the Authors proved that when the R_C–C < 1.34 Å, the system is in nonmagnetic state and the dopants tend to form a cluster through a direct C–C bonding interaction, however when the R_C–C is in-between 3.8 Å and 5.5 Å, then there exists a strong ferromagnetic (FM) or antiferromagnetic (AFM) coupling between C ions. The dependence of magnetic character on R_C–C value was also studied earlier in [8], where the Authors concluded theoretically that when R_C–C ∼ 3–4 Å the coupling between the dopant C ions was leading to either FM or AFM. The dependence of the type of magnetic interaction on the R_C–C was also studied for other host crystalline medium like in ZnO [5], where the Authors theoretically deduced the realization of FM interaction when R_C–C ∼ 7.76 Å.

In table 2 the magnetic parameters of the samples including the FM of pure TiO₂–H and TiO₂–C1–H sample were presented. The saturation magnetization (M_s) increased by ∼20 times due to C1 doping, while the magnetic energy (U_m = M_r.H_c) remained without a considerable change. The M_s data onto table 2 can be used to estimate the effective concentration $\left({{\rm{n}}}_{{\rm{C}}}^{* }\right)$ of C ion-species participated in FM interaction by using the following equation, ${\mu }_{{\rm{C}}}={{\rm{M}}}_{{\rm{s}}}({\rm{emu}}/{\rm{g}})\times \rho /{{\rm{n}}}_{{\rm{C}}}^{* },$ where ${\mu }_{C}$ is the magnetic moment of C ion-species, ${\mu }_{C}=1.24\,{\mu }_{B}$ found from PM behavior. Thus, only ∼0.3% of the total doped C into

TiO₂–C1–H sample participated in the FM interaction. This revealed small participation fraction of dopant C ions that occupied O-vacancies and interacted ferromagnetically with each other. It is useful to compare the magnetization (M_s) of the present samples with that of TiO₂ powder doped with other types of ions. The present work result (∼10 memu g⁻¹) is comparable to the previously reported value for host TiO₂ doped with magnetic ions; for TiO₂−3%Co (7.42 memu g⁻¹), TiO₂−3%Mn (9.46 memu g⁻¹) [22], TiO₂−5%Ni (21.79 memu g⁻¹) [27], and TiO₂:5%Fe (20 memu g⁻¹) [28]. Thus, it is obviously clear the significance of hydrogenation in supporting the embedded magnetic properties.

Titanium tetrachloride (TiCl₄) and sucrose (C₁₂H₂₂O₁₁) complex were used to synthesize pure and C-incorporated TiO₂ nanocomposite powders by thermal co-decomposition method. Structural tests of pure TiO₂ nanocomposite confirm the formation of major A phase, however, the incorporation of C ions and hydrogenation supported the A to R transformation. The optical study revealed that the C inclusion redshifted the optical band gap due to the creation of point defects like O-vacancies while the hydrogenation blueshifted the optical band gap according to the Moss-Burstein effect. The inserted C ion-species occupied interstitial positions of TiO₂ lattice. During the hydrogenation, H₂ molecules tossed the inserted C ions and dropped them into O-vacancies. Then, the super-exchange interaction between C2p–C2p electrons via oxygen vacancies could create FM depending on the average C–C separation (R_C–C). This model explained the creation of FM just for ∼1% C- incorporated host TiO₂ lattice. The local effective magnetic moment was found to be 1.24 μ_B/C, which was attributed to the magnetic moment of C2p in addition to some contribution of O-vacancies.