Structural, magnetic and electronic properties of iron doped barium strontium titanate

Abstract

The structural, magnetic, and electronic properties of iron doped barium strontium titanate (Ba_0.7Sr_0.3Fe_xTi_1−xO₃ where x = 0, 0.1, 0.2, 0.3; BSTF) solid solutions synthesized via solid state reaction route were investigated. X-ray diffraction patterns of all the samples clearly show phase formation with the absence of impurity peaks. The Rietveld refinement confirmed the coexistence of the tetragonal and cubic phase for samples with Fe content x = 0, 0.1 and pure cubic phase for x > 0.1. The M–H hysteresis curves for samples with composition x = 0.1 and 0.2 exhibit paramagnetic behaviour even at low temperatures and the composition with x = 0.3 shows the nature of weak ferro- and ferri-magnetic orderings at about 2 K. X-ray absorption near edge structure (XANES) spectroscopy reveals the presence of Fe²⁺ with an increase in the Fe content and also the mixed valence states of Fe ions.

---

I. Introduction

Ferroelectric ceramics are of technological promise because of their wide range of applications, such as dynamic random access memory (DRAMS), non-volatile memory, pyroelectric detectors and electro-optic devices, etc.^1–3 Many researchers have reported lead based ferroelectric materials such as PbZrTiO₃, and PbTiO₃ etc., because of their enhanced ferroelectric and dielectric properties.^4–9 But there are major drawbacks of using lead based materials because of volatile nature of lead oxide and hence makes it difficult to maintain stoichiometry. It also shows fatigue oddity in the fabricated devices.¹⁰ Apart from all these drawbacks, Pb based materials are hazardous to health and not environment friendly.

Barium strontium titanate (Ba_1−xSr_xTiO₃, hereafter denoted as BST) being ecofriendly material, is considered as one of the most promising candidates for ferroelectric devices due to its excellent properties of high dielectric constant, low leakage current and adjustable Curie temperature (T_C).^11–13 The doping of magnetic ion ‘Fe’ at ‘Ti’ site can induce the magnetism in ferroelectric material. Although, various researchers^14–18 have reported on Fe doped BaTiO₃ ceramics (host material) and reported ferromagnetic behavior in the samples. But Fe doped BST solid solutions have been rarely reported.¹⁹ Recently, Guo et al. have reported on the structural and multiferroic properties of bulk Fe-doped Ba_0.5Sr_0.5TiO₃ and their results show the simultaneous coexistence of ferroelectric and ferromagnetic orderings.¹⁹

However, very less is known about the magnetic properties and the electronic structures of these Fe-doped BST solid solutions. X-ray absorption spectroscopy especially X-ray absorption near-edge structure (XANES) when invited with Fe doped BST can provide information about the change in the valency of Fe and Ti ions and chemical bonding information. Present investigation focuses on the structural, magnetic properties and electronic structure of Fe doped BST ceramics prepared by solid state route.

II. Experimental details

Bulk samples with composition Ba_0.7Sr_0.3Fe_xTi_1−xO₃ where x = 0, 0.1, 0.2, 0.3 (abbreviated as BST, BSTF1, BSTF2, BSTF3) were prepared using conventional solid state reaction route. The raw materials, (>99.9% purity Sigma Aldrich) barium carbonate (BaCO₃), strontium carbonate (SrCO₃), iron(III) oxide (Fe₂O₃) and titanium(IV) oxide (TiO₂) were weighed in the stoichiometric proportions and ball milled for 6 h in propanol medium at 200 rpm. The mixed powders were then calcined at 1000 °C for 5 h. Polyvinyl alcohol (PVA) (2% by weight) was mixed as binder to these calcined powders. The PVA mixed calcined powders were pressed in the form of pellets of 10 mm diameter. The prepared pellets were then sintered at 1250 °C for 6 hours.

Bulk samples were characterized for structural, morphological and magnetic properties. X-ray diffractograms of polycrystalline sample of the Fe doped BST at room temperature confirmed the phase purity. The structural refinements were performed using the Rietveld refinement program FULLPROF. A Thompson-Cox-Hastings pseudo-Voigt with axial divergence asymmetry function was used to model the peak profiles. The background was fitted using a sixth order polynomial. Except for the occupancy parameters of the atoms, which were fixed corresponding to the nominal composition, all other parameters, i.e., scale factor, zero displacement, isotropic profile parameters, lattice parameters, isotropic thermal parameters and positional coordinates, were refined. All the refinements have used the data over the full angular range. Although in the figures only a limited range is shown for clarity.

The magnetic measurements of bulk samples were characterized using commercial Quantum Design make 9 T PPMS based vibrating sample magnetometer. X-ray absorption near edge fine structure (XANES) measurement was performed at the National Synchrotron Radiation Research Center (NSRRC) in Taiwan. Ti and Fe L-edges were measured at the high-energy spherical grating monochromator (HSGM) at BL20A1 (resolving power E/ΔE = 8000) in an ultrahigh vacuum chamber (<5 × 10⁻⁹ Torr) by recording the total electron yield signal. Ti K edge was collected at BL01C and the energy resolution at Ti K-edge (4966 eV) was about 0.5 eV. Standard oxides were used for energy calibration and also for comparing the different electronic states.

III. Results and discussions

(i) Structural properties

Fig. 1 show the powder X-ray diffraction patterns of Fe doped BST bulk samples with x = 0.0, 0.1, 0.2 and 0.3 at room temperature. All the Bragg reflections present in the powder X-ray diffraction pattern could be indexed as main cubic perovskite reflections and confirmed the phase purity. It is well known that pure BaTiO₃ has tetragonal structure at room temperature and transform in the cubic phase around 396 K (∼123 °C). This structural phase transition is first order in nature. On doping of Sr at Ba site, tetragonal to cubic phase transition temperature decreases. For the present composition Ba_0.70Sr_0.30TiO₃, tetragonal structure exists at room temperature and it undergoes to cubic phase around 308 K (∼35 °C).²⁰ The splitting of {h00} type reflections present in powder diffraction data is signature for tetragonal structure. A careful and detail inspection of diffraction patterns suggest a change in the Bragg's profile with composition (see top panel of Fig. 2). It is evident from this figure that on increasing Fe content in BST, the peaks appeared at lower angle marked with arrow in diffraction patterns as shown in top panel of Fig. 2 disappear for x = 0.20. It may be signature of structural phase transition.

Fig. 1 Observed and refined XRD pattern of powdered samples for BST, BSTF1, BSTF2 and BSTF3.

Fig. 2 XRD pattern for BST, BSTF1, BSTF2 and BSTF3 showing structural phase transition.

To obtain structural parameters, we have refined the powder diffraction data using tetragonal symmetry with space group (P4mm) for pure BST sample. The result of Rietveld refinements is not able to index all observed peak in experimental data (one of them is marked with arrow and see top panel of Fig. 3). We found that all the reflections could be accounted using phase coexistence model (tetragonal symmetry (P4mm) and cubic (Pm3̄m) structure). The fit between the observed and calculated profiles is satisfactory and shown in Fig. 3 (bottom panel of Fig. 3). We also found that Fe doped BST with x = 0.1 has phase coexistence at room temperature. For the composition x > 0.1, {h00} type reflection do not show splitting which confirmed the change in the structure from tetragonal symmetry. Absence of additional reflections and splitting in main Braggs peaks clearly suggests that Fe doped BST with x = 0.2 and 0.3 transforms to cubic phase. Thus, the powder X-ray diffraction data are analyzed using cubic phase with the space group Pm3̄m. The fit between the observed and calculated profiles is satisfactory (Fig. 3) and indicating the correctness of model. We find that tetragonality (c/a ratio) decreases with increase in doping of iron resulting in tetragonal to cubic phase. An obvious slight change in lattice parameters is also seen which can be accredited to ionic radius change.²¹ The refined lattice parameters for BST solid solutions are given in Table 1. This observation reveals that tetragonal to cubic transition temperature lowered with increasing doping.

Fig. 3 XRD pattern showing fitting of BST and BSTF1 with phase coexistence model and pure cubic for BSTF2 and BSTF3.

Table 1 Rietveld refined parameters for BST, BSTF1, BSTF2 and BSTF3

Sample name	Phase	Space group	Lattice parameters	% molar	RB	Rp	Rw–p	Rexp	χ^2
BST	Tetragonal	P4mm	a = b = 3.96773 (4) Å, c = 3.97745 (4) Å	73.85 (0.81)	4.11	11.7	17.0	11.17	2.31
	Cubic	Pm3̄m	a = b = c = 3.98385 (14) Å	26.15 (0.57)	6.60
BSTF1	Tetragonal	P4mm	a = b = 3.97017 (6) Å, c = 3.97307 (13) Å	48.40 (0.63)	5.68	15.3	21.3	11.30	3.56
	Cubic	Pm3̄m	a = b = c = 3.98121 (9) Å	51.60 (0.65)	8.89
BSTF2	Cubic	Pm3̄m	a = b = c = 3.97871 (3) Å	100 (0.69)	8.29	14.5	19.2	12.30	2.43
BSTF3	Cubic	Pm3̄m	a = b = c = 3.98084 (3) Å	100 (0.57)	6.84	13.9	18.6	12.23	2.32

---

(ii) Magnetic properties

The Zero Field Cooled (ZFC) and field cooled (FC) magnetization curves in an applied magnetic field of 100 Oe and 5 kOe for BSTF1, BSTF2 and BSTF3 samples are shown in Fig. 4(a–c). For BSTF1 and BSTF2 samples, the ZFC and FC curves for both fields coincide with each other and do not show any indication of any magnetic ordering. These two samples continue to be in paramagnetic state down to ∼2 K. This is also confirmed by no hysteresis in the M vs. H isotherms recorded at several temperatures as shown in Fig. 5(a) and (b). On other hand in BSTF3 sample, the ZFC curve taken in 100 Oe magnetic field shows peak around 4.8 K and ZFC–FC bifurcation below this temperature. This seems to suggest development of long range magnetic order below 4.8 K in BSTF3 sample. This is further confirmed by the opening of M–H hysteresis loop with a coercivity of ±300 Oe (which is well outside the experimental errors of ±10 Oe). However, the cusp and bifurcation in BSTF3 disappears in the ZFC–FC curves taken at 5 kOe, which is not surprising as it is commonly observed in many magnetic systems that high fields suppress the bifurcation. In order to rule out the presence of superparamagnetic behavior, the M–H loops of BSTF3 shown in Fig. 5(c) are in the form of M vs. H/T in Fig. 6 as for superparamagnetic materials these loops will overlap into one universal curve. It is clearly seen that none of M–H curves overlaps each other, thereby clearly ruling out superparamagnetic behavior at any temperature. These observations clearly show that there is a weak ferro or ferri magnetic ordering in BSTF3 sample at low temperature.

Fig. 4 (a) ZFC and FC magnetization versus temperature curves of BSTF1. (b) ZFC and FC magnetization versus temperature curves of BSTF2 and χ versus T plots for all the samples. (c) ZFC and FC magnetization versus temperature curves of BSTF3.

Fig. 5 (a) Magnetization versus magnetic field of BSTF1. (b) Magnetization versus magnetic field of BSTF2. (c) Magnetization versus magnetic field of BSTF3.

Fig. 6 M versus H/T plot confirming absence of superparamagnetic behaviour in BSTF3.

The temperature dependent magnetization curves (above T_C in case of BSTF3 sample) of all samples were analyzed based on Curie–Weiss (CW) law given by

where C is Curie constant and θ is the Curie–Weiss temperature. The inverse susceptibility (1/χ = H/M) data is plotted as function of temperature and linear fit is performed over a temperature region where χ⁻¹ vs. T is linear. The slope of such fit gives C and intercept on temperature gives θ. The Curie–Weiss (CW) analysis shows that θ is negative (Table 2) for all samples, which indicates that the nature of interactions are antiferromagnetic (AFM). This analysis also shows that, although BSTF1 and BSTF2 samples remain as paramagnetic down to 2 K, there are considerable magnetic (AFM) interactions present below 185 K (BSTF1) and 225 K (BSTF2) as revealed by deviations from CW fits. In case of BSTF3, the deviations from CW fit start from 250 K downwards. The effective paramagnetic moment, μ_eff is calculated from the Curie constant C using the relation μ_eff = 2.828CA/ρ (where A molecular weight and ρ is the density) are shown in Table 2. The theoretical values for effective paramagnetic moment, μ^th_eff are 4.90 μ_B/Fe²⁺ and 5.92 μ_B/Fe³⁺ The effective paramagnetic moments of BSTF1 and BSFT2 are close to Fe³⁺ and Fe²⁺ ions, respectively. However, μ_eff = 3.01 μ_B of BSFT3 is lower than that Fe²⁺ and Fe³⁺ ions. The presence of Fe ions in higher valence states, say Fe⁴⁺, can give reduced μ_eff. So, this large shift in the experimental and theoretical effective magnetic moment may be due to the presence of mixed valence states of the doped Fe ions.¹⁵

Table 2 Results of χ⁻¹ versus T analysis for Fe doped BST samples

Sample name	Temp. range (fit)	Intercept	Slope	θ (K)	Effective magnetic moment
BSTF1	185 K–310 K	14 664	674.8	−21.7	6.42 μB/Fe
BSTF2	225 K–310 K	22 379.3	354.2	−67.2	4.78 μB/Fe
BSTF3	250 K–310 K	50 522.7	384.4	−131.5	3.01 μB/Fe

---

(iii) Electronic structure studies

In order to determine the electronic structures of Fe doped BST compounds, the XANES measurements were carried out at different absorption edges. The XANES spectra at Ti L_3,2-edges of BST, BSTF1, BSTF2, and BSTF3 along with reference samples TiO₂ (anatase and rutile crystalline phases), are shown in Fig. 7. The spectra show several well resolved features that are due to excitations of 2p core electrons into the Ti 3d empty states above the Fermi energy (E_F). The spectra can be divided into two prominent regions in the energy ranges 455–461 eV and 461–468 eV, corresponding to the L₃ (2p_3/2) and L₂ (2p_1/2) absorption edges, respectively. The multiple structures are mainly due to strong Coulomb interaction between poorly screened Ti 3d electrons and the Ti 2p core hole.²² The L₂ edge features are broadened relative to the L₃ edge because of a shorter lifetime of the 2p_1/2 core hole. Under the crystal field the 3d band splits into t_2g and e_g bands in each structure Ti occupies an octahedral site surrounded by O atoms. The line shapes of all the BST spectra do not show significant changes, implying that the Ti remain the tetravalent [Ti⁴⁺] state with 3d⁰ ground-state configuration as Fe ions are doped.²³ However, the FWHM of e_g band presents a gradually broadening from 1.09 eV at x = 0 to 1.18 eV at x = 0.3. This modification of the e_g band in the spectrum is an indication of the distorted octahedral (O_h) symmetry.^22,24 Consequently, as the Fe is doped into the ABO₃ structure, the lattice distortion is observed in TiO₆. This doping induced distortion is also evidenced by the XRD results and the pre-edge features of Ti K-edge XAS as mentioned below.

Fig. 7 The Ti L_3,2-edges XANES spectra of Fe doped BST along with reference samples of TiO₂ (anatase and rutile crystalline phases).

Fig. 8 shows Ti K-edge XANES spectra of TiO₂ (anatase), Ti₂O₃ and Fe doped BST. The feature-rich spectra at the K-edges are originated from transitions from the Ti 1s core level to 4p-derived final states, which are composed of strongly hybridized O 2p and Ti 4sp and 3d orbitals. In general, the quadrupole-allowed transitions appeared in the pre-edge region at about energies between 4964 and 4977 eV, which corresponds to the 1s to 3d transition through 4sp–3d hybridization in the TM oxide system.²⁵ For a detailed comparison of the pre-edge spectra, the inset shows four main features (A1, A2, B1, and B2) in the Ti K-edge spectrum. The splitting of A1&A2 and B1&B2 are due to the 3d orbital under the O_h crystal field.^26,27 According to Ti K-edge spectral profile, no significant change in Ti valence can be observed for various doping levels of Fe. However, an increase of the intensity of peak A1 is observed, indicating the electronic structure is modified as introducing Fe into the BST. Moreover, a decrease in the B1 and B2 peaks as the Fe content increases is revealed, probably resulted from the partial replacement of Ti⁴⁺ by Fe ions, giving rise to a more atomic disorder in TiO₆ octahedron, owing to different ionic size of Fe and Ti. Thus, the introduction of Fe may cause the disordered atomic structure, affecting the hybridization of Ti–O within O_h structure and smearing out the peak B which is consistent with the observation of Ti L-edge in Fig. 7. These pre-edge spectra are strongly dependent on the symmetry of the local atomic arrangement as well as the unoccupied electron states of through sp–d hybridization.²⁷

Fig. 8 Ti K-edge X-ray absorption pre-edge spectra of TiO₂ (anatase) and Fe doped BST samples.

The Fe L₃-edge XANES spectra presented in Fig. 9 are due to the transition from Fe 2p_3/2 levels to unoccupied 3d states owing to spin–orbit coupling. The spectrum of Fe₂O₃, due to the crystal-field splitting, exhibits a multi-peak structure with a main feature around 709 eV and a shoulder at 707.4 eV. The spectral profile and peak positions of BST are found to be very similar to Fe₂O₃, implying the Fe has its trivalent states. However, the spectra of Fe-doped BST, although similar, is different from that of Fe₂O₃, suggesting the Fe exhibits trivalent states but the environment of the Fe in this sample is not as same as Fe₂O₃. The main feature is shifted to a low photon energy and a shoulder appears at 709.2 eV at Fe concentration x = 0.3. As increasing the Fe concentration, the mixed valence states and Fe²⁺ are observed. Analysis of XANES spectra at Fe L-edges indicates that the dopant Fe is reduced from Fe³⁺ to partial Fe²⁺ with the increase of the Fe concentration. This reduction is correlated to the charge redistribution between Fe and Ti. The Fe gains charge and Ti loses charge, which is evidenced by pre-edge of Ti K-edge in Fig. 8. The leading peak of pre-edge is increased in the intensity as a function of Fe concentration, indicating the Ti may transfer out some electrons. These electrons are likely to migrate to Fe 3d orbital.

Fig. 9 The Fe L₃-edge XANES spectra of Fe doped BST samples.

The XANES spectra of the Ba L₃-edge are shown in Fig. 10. The spectral shape, edge position, and intensity of the main edge are almost identical at all Fe content, indicating that the local structure around Ba²⁺ in the system does not show any influence on Fe doping in BST. The Fig. 11 displays the XANES Sr L₃-edge spectra. The obtained results also indicate that the introduction of Fe dopants do not introduce significant changes in the local structure and electronic properties around Sr²⁺ ions. Thus, above XANES analysis reveal that the interaction between dopant Fe and Ti–O bond is more important than other constituent elements. Owing to different orientations of the 3d-orbitals in the structural symmetry of Fe–O, the energy bands experience a complex crystal field splitting in BST structure. Therefore, according to the Fe L-edge XANES results, upon slight doping, the formation of Ti–O–Fe bond is expected. According to the spectral profile and the peak position of XANES spectra, it is suggested that the Ti–O–Fe bond is formed as the Ti ion is substituted by Fe ion. The change of electronic structure around Fe ions appears to have a significant influence on the oxidization of Fe state and hence the magnetic properties, particularly, when the x value increases to 0.3. Above analysis of the Fe L₃-edge reveals that with the increase in Fe concentration, the Fe ions exhibits mixed valence nature and presence of Fe²⁺ ions.

Fig. 10 The XANES spectra of the Ba L₃-edges of Fe doped BST samples.

Fig. 11 The XANES spectra of the Sr L₃-edge of Fe doped BST samples.

IV. Conclusions

X-ray analysis carried out on samples of Fe doped BST show phase formation. Refinement confirmed that the mixture of tetragonal (78.35%) and cubic (26.15%) and mixture of tetragonal (48.40%) and cubic (51.60%) phase for BST and BSTF1 respectively and pure cubic phase (100%) was observed for BSTF2 and BSTF3. Magnetic measurements for BSTF1 and BSTF2 measured at different temperatures i.e. from 310 K (RT) to 2 K showed paramagnetic behavior. But the BSTF3 shows long range magnetic order below 4.8 K. It is inevitable that presence of Fe²⁺ state is responsible for paramagnetism. However, with increasing Fe content mixed valency seem to be setting in. This strange magnetic behavior is due to the presence of mixed valence states and in particular Fe²⁺ state in the samples, as observed from Fe L₃-edge XANES spectra. The Ti L_3,2-edges at the XANES spectra confirmed that the doping of Fe in ABO₃ structure leads to the lattice distortion. This doping induced distortion is also evidenced by the Fe K-edge XAS. The Fe L₃-edge XANES spectra revealed that with increasing the Fe concentration, the mixed valence states and presence of Fe²⁺ are observed. The XANES spectra of the Ba L₃-edge and Sr L₃-edge spectra confirmed that the local structure around Ba²⁺ and Sr²⁺ respectively does not show any influence from the dopants in the BST system.

Acknowledgements

One of the authors Anumeet Kaur would like to thank University Grants Commission (UGC) and IUAC, New Delhi for funding the project UFR 53306.

References

1. J. F. Scott, Science, 2007, 315, 954–959.
2. J. F. Scott, Annual Review of Materials Research, 1998, vol. 28, pp. 79–100..
3. A. Ioachim, M. I. Toacsan, M. G. Banciu, L. Nedelcu, F. Vasiliu, H. V. Alexandru, C. Berbecaru and G. Stoica, Prog. Solid State Chem., 2007, 35, 513–520.
4. A. Singh and R. Chatterjee, Appl. Phys. Lett., 2008, 93, 3–6.
5. V. R. Palkar and S. K. Malik, Solid State Commun., 2005, 134, 783–786.
6. E. V. Barabanova, O. V. Malyshkina, a. I. Ivanova, E. M. Posadova, K. M. Zaborovskiy and a. V. Daineko, IOP Conf. Ser.: Mater. Sci. Eng., 2013, 49, 012026.
7. K. Pytel, J. Suchanicz, M. Livinsh and A. Sternberg, Condens. Matter Phys., 2013, 16, 31706.
8. H. Meštrić, R. A. Eichel, T. Kloss, K. P. Dinse, S. Laubach, S. Laubach, P. C. Schmidt, K. A. Schönau, M. Knapp and H. Ehrenberg, Phys. Rev. B: Condens. Matter Mater. Phys., 2005, 71, 1–10.
9. Z. Ren, G. Xu, X. Wei, Y. Liu, X. Hou, P. Du, W. Weng, G. Shen and G. Han, Appl. Phys. Lett., 2007, 91, 063106.
10. C. B. Samantaray, A. Dhar, D. Bhattacharya, M. L. Mukherjee and S. K. Ray, In Situ, 2001, 2, 365–370.
11. H. Frayssignes, B. L. Cheng, G. Fantozzi and T. W. Button, J. Eur. Ceram. Soc., 2005, 25, 3203–3206.
12. V. Somani and S. J. Kalita, J. Electroceram., 2007, 18, 57–65.
13. H. Y. Tian, W. G. Luo, X. H. Pu and A. L. Ding, J. Mater. Sci. Lett., 2000, 19, 1211–1213.
14. F. Lin, D. Jiang, X. Ma and W. Shi, J. Magn. Magn. Mater., 2008, 320, 691–694.
15. B. Deka, S. Ravi, A. Perumal and D. Pamu, Phys. B, 2014, 448, 204–206.
16. S. Y. Qiu, W. Li, Y. Liu, G. H. Liu, Y. Q. Wu and N. Chen, Trans. Nonferrous Met. Soc. China, 2010, 20, 1911–1915.
17. A. Mishra, Int. J. Mater. Sci. Appl., 2012, 1, 14.
18. Z. Guo, L. Yang, H. Qiu, X. Zhan, J. Yin and L. Cao, Mod. Phys. Lett. B, 2012, 26, 1250056.
19. Z. Guo, L. Pan, C. Bi, H. Qiu, X. Zhao, L. Yang and M. Y. Rafique, J. Magn. Magn. Mater., 2013, 325, 24–28.
20. D. Y. Wang, J. Wang, H. L. W. Chan and C. L. Choy, J. Appl. Phys., 2007, 101, 1–7.
21. Y. Ye and T. Guo, Ceram. Int., 2009, 35, 2761–2765.
22. F. M. F. De Groot, J. C. Fuggle, B. T. Thole and G. A. Sawatzky, Phys. Rev. B, 1990, 41, 928–937.
23. D. H. Kim, H. J. Lee, G. Kim, Y. S. Koo, J. H. Jung, H. J. Shin, J. Y. Kim and J. S. Kang, Phys. Rev. B: Condens. Matter Mater. Phys., 2009, 79, 2–5.
24. F. M. F. De Groot, J. C. Fuggle, B. T. Thole and G. A. Sawatzky, Phys. Rev. B, 1990, 42, 5459–5468.
25. B. Poumellec, J. F. Marucco and B. Touzelin, Phys. Rev. B, 1987, 35, 2284–2294.
26. T. Yamamoto, T. Mizoguchi and I. Tanaka, Phys. Rev. B: Condens. Matter Mater. Phys., 2005, 71, 3–6.
27. N. Jiang, D. Su and J. C. H. Spence, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76, 1–9.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra21458d

---