1 Introduction

Ferroelectric materials may be divided into two classes: classical or relaxor ferroelectrics 〚1〛. Typically, relaxors have at least one crystallographic site that is occupied by two or more ions. In addition to the usual applications for classical ferroelectrics, relaxors are of great interest for dielectrics in capacitors and actuators. Most used relaxors are lead-based perovskite ceramics such as PbMg_1/3Nb_2/3O₃ (PMN) and derived compounds 〚2〛. However, these compositions have the obvious disadvantages associated with the volatility and toxicity of PbO. Therefore, much current research is directed towards more environmentally friendly Pb-free relaxor materials. In this way, a large number of relaxor lead-free BaTiO₃ derived ceramics with perovskite structure was recently prepared and characterised 〚3〛.

Likewise lead-free relaxor compositions belonging to tetragonal tungsten bronze (TTB) could be also of interest. Fig. 1 shows a schematic representation of the TTB structure projection along a 001 direction. For a general formulation A₂BC₂M₅X₁₅ (X = O, F), large cations (e.g. Na⁺, K⁺, Sr²⁺, Ba²⁺, Pb²⁺, La³⁺, Bi³⁺...) occupy the 15 CN (A) and the 12 CN (B) sites, small cations like Li⁺ are in the 9 CN (C) sites and small and highly charged cations (e.g. Nb⁵⁺, Ta⁵⁺...) are in the octahedral (M) site.

Previous authors had announced the solid-state solution Sr_2.5(1–x)Ba_2.5xNb₅O₁₅ to be relaxor 〚4–9〛. The present work is devoted to other lead-free relaxor compositions selected among a great number of TTB-type compositions.

2 Experimental

The oxides and the oxyfluorides were prepared by solid state reaction at temperatures between 900 and 1100 °C for 15 h from intimate and ground mixings of starting oxides, fluorides and carbonates in stoichiometric amounts. After new intimate and ground mixings, powders obtained were pressed under 100 MPa into disks of 8 mm diameter and about 1 mm thickness. Depending on the compositions, heat treatments were then performed at temperatures between 1150 and 1450 °C in air for the oxides and in platinum sealed tube under dry oxygen for the oxyfluorides, in order to avoid hydrolysis at high temperature.

Room temperature powder X-ray diffraction patterns were recorded on a Philips diffractometer, using CuKα radiation (λ = 1.5406 Å), in the angular range 5 ≤ 2θ ≤ 60°. This made it possible to verify that the samples were single phase. The diameter shrinkages ΔΦ/Φ were systematically determined as (Φ_init – Φ_fin)/Φ_init.. Their values were in the range 0.13–0.17.

Dielectric measurements were performed on the ceramic disks after deposition of gold or platinum electrodes on the circular faces. The real ε^′_r and imaginary ε^′′_r relative permittivities were determined under helium as a function of both temperature (77–900 K) and frequency (10²–2 × 10⁵ Hz) using a Wayne–Kerr 6425 component analyser.

3 Results

In addition to barium–strontium niobates, a relaxor behaviour was recently demonstrated in various other TTB compositions including niobates, tantalates, niobo-tantalates and niobo-titanates (Table 1). As an example, Fig. 2 shows the typical temperature dependence of ε^′_r for a relaxor ceramic with composition Sr₂K(Nb_0.775Ta_0.225)O₁₅: the temperature T_m of ε^′_r maximum is shifted towards higher values for increasing frequencies.

Of all the lead-free compounds studied, it is with one of the compositions containing Bi³⁺, another 6(sp)² lone pair cation, that the highest value of T_m was obtained: T_m (BaBiNb₅O₁₅) = 319 K at 10³ Hz. In addition, independently of the value of T_m, the bismuth compositions allow the highest values of ΔT_m: ΔT_m = 80 and 73 K for the potassium and barium niobates respectively. Such a result is in good agreement with that obtained for lead-free compositions derived from BaTiO₃ 〚10〛.

Some solid solutions situated between classical ferroelectric and relaxor ones have also been studied. Such is the case with A₂B(Nb_1–xTa_x)₅O₁₅. The relaxor behaviour occurs for the highest values of x. Fig. 3 shows, as an example, the variations of T_C (0 ≤ x < 0.16) and T_m (0.16 ≤ x ≤ 1) for ceramics with composition Sr₂K(Nb_1–xTa_x)₅O₁₅. Both values decrease when the tantalum rate increases.

This is also the case with the Sr₂KNb₅O₁₅/SrK₂Nb₅O₁₄F and Ba₂NaNb₅O₁₅/BaNa₂Nb₅O₁₄F systems; for the compositions close to the oxide, the behaviour is ferroelectric, while it becomes relaxor when the fluorine rate and thus the sodium or potassium rate are sufficiently high. Fig. 4 shows as an example the variations of T_C (0 ≤ x ≤ 0.25) and T_m (0.20 ≤ x ≤ 1) for the barium/sodium system. It is interesting to notice that both relaxor and classical ferroelectric phases coexist for compositions (0.20 ≤ x ≤ 0.25) giving thus the transition sequence: relaxor–classical ferroelectric–paraelectric, on heating. Here also both values of T_C and T_m decrease when x increases. The frequency dependence of the permittivities of a ceramic with a composition BaNa₂Nb₅O₁₄F clearly displays a frequency relaxation at 77 K, while it disappears at 300 K, i.e. for T > T_m (T_m = 140 K at 10³ Hz) (Fig. 5).

4 Discussion

Some compositions had been previously prepared and characterised. They had been announced to be ferroelectric, measurements having been realised only at one frequency (10³ Hz) and not at various frequencies as it has to be the case for relaxor property studies. It is the automation of dielectric measurements performed with an impedance analyser that allows us to obtain such actual results.

As for the perovskites, the relaxor effect appears in many TTB-type compounds when at least two ions occupy the same crystallographic site. The co-existence of cations seems as favourable in the A site as in the M site. In addition, at least one cation ferroelectrically active (Ti⁴⁺, Nb⁵⁺, Ta⁵⁺) must be present in the octahedral site.

The relaxor effect can be due to a cationic disorder in the A site, whatever the coupled substitution ensuring the electrical neutrality; this is the case with BaLaNb₅O₁₅, when Ba²⁺ and La³⁺ occupy only the A site. Ba₂NaNb₅O₁₅, in which the same two Ba²⁺ cations occupy the A site, is a classical ferroelectric; on the contrary, BaNa₂Nb₅O₁₄F is a relaxor one. The crystal structure determination of this last compound is in good agreement with the dielectric properties, i.e. the A-site is statistically occupied by equal quantities of Ba²⁺ and Na⁺, thereby giving a cationic disorder in this A-site while the B-site is filled by Na⁺ cations 〚11〛. In addition, there is here coexistence of F^– and O^2– in the anionic sites.

Concerning the Ta–Nb substitution, the replacement of Nb⁵⁺, highly polarisable, by Ta⁵⁺ transforms the macroscopic polarisation into a local one. This results in a relaxor state as the long-range order is not induced by a local dipolar order. The higher the associated cationic order in the A and B sites, the higher the Ta–Nb substitution rate leading to relaxor behaviour (Table 1); e.g. x = 0.16 for Sr₂K(Nb_1–xTa_x)₅O₁₅ and x = 0.65 for Ba₂Na(Nb_1–xTa_x)₅O₁₅. In fact, there is no ambiguity regarding the disordered distribution of the two small Sr²⁺ cations and the single large K⁺ one, nor regarding the ordered distribution of the two large Ba²⁺ cations and the single small Na⁺ ones, since both occur in two large A sites and in one smaller B one.

For oxyfluorides, F^––O^2– substitution is coupled with a cationic one in order to ensure the electric neutrality. The origin of the relaxor effect comes from cationic or/and anionic substitutions when there are two different cations in the same crystallographic site, e.g. Ba_2–xNa_1+xNb₅O_15–xF_x, the A site being filled by both Ba²⁺ and Na⁺ for x > 0. However, it seems that it can be attributed to the only anionic F^––O^2– substitution in the lack of different cations in the same site, e.g. K₃LiNb₅O₁₄F: due to their very different sizes, the K⁺ and the Li⁺ cations occupy only the (A + B) sites and C site respectively. It is also the case of Ba_2.25□_0.75Nb₅O_14.5F_0.5, where A and B sites are occupied by only Ba²⁺. In these two compositions, the relaxor effect is induced by a dilution of the oxygen network, leading to a local polarisation that breaks the long-range order usually at the origin of classical ferroelectric properties.

5 Conclusion

In TTB-type compositions, the relaxor behaviour appears when at least two ions occupy the same crystallographic site. The Mⁿ⁺ cations in the octahedral site have to be ferroelectrically active (e.g. Ti⁴⁺, Nb⁵⁺ or Ta⁵⁺). In any case, local polarisation leads to a relaxor effect; on the contrary a macroscopic one gives a classical ferroelectric behaviour. The relaxor effect is due to either a cationic or an anionic disorder in the A or B sites, or to a dilution of the ferroelectric character by cationic or anionic substitutions. Some of these materials could prove valuable (dielectrics for capacitors and actuators), because they are environment-friendly.