Sol–gel synthesis of doped Cu₂V₂O₇ fine particles showing giant negative thermal expansion

Controlling thermal expansion has become an important issue in modern industry, and there is a strong demand for negative thermal expansion (NTE) materials that have excellent thermal expansion control capabilities.^1–5 In particular, control of thermal expansion at the micrometer-level in local areas has become increasingly important, especially in electronic devices,⁶ and requires submicron-sized thermal expansion inhibitors.

In recent years, materials such as manganese nitrides,^7–10 bismuth–nickel oxides,^11–14 and lead–vanadium oxides^15,16 have been found to exhibit a large NTE due to a volume change accompanying the phase transition. Such phase-change type NTE materials exhibit large negative linear expansion coefficients α. However, the drawback is that their operating temperature (T) ranges are narrow, resulting in limited use. Hence, attention has been focused on materials that have NTE mechanisms arising from microstructural effects.^17–20 The microstructural effects are a result of the anisotropic thermal deformation of the crystal lattice and have fewer restrictions on the operating T range. By doping Cu₂V₂O₇^21–25 with Zn, the obtained β-Cu_1.8Zn_0.2V₂O₇ showed a large NTE with the linear thermal expansion coefficient α = −14.4 ppm/K over a wide T range of 100 K–700 K.²⁶ However, in the case of NTE resulting from microstructural effects, as in the present system, there is a risk that the microstructures relevant to NTE might be destroyed if pulverized. In order to reduce the particle size, it is necessary to “make it small from the beginning.”

For this reason, we focused on the sol–gel method^27–29 as an alternative to the conventional solid-state reaction method. The sol–gel method is suitable for producing solid solutions with uniform composition since it starts from an aqueous solution in which the elements are uniformly mixed at the atomic level. Furthermore, obtaining a solid phase from a solution suppresses the growth of crystal grains and facilitates the production of fine particles having a shape close to that of a true sphere. In a previous study,³⁰ β-Cu_1.8Zn_0.2V₂O₇ was synthesized into fine particles by a spray-dry method that produces a solid phase from a solution, similar to the sol–gel method. In this spray-dry synthesis, NTE characteristics comparable to those of the bulk sintered bodies with median particle sizes of 2.7 µm were achieved. In the present study, we achieved NTE characteristics comparable to those of the bulk sintered bodies using the sol–gel method. The fine particles produced were about 1 µm in size, smaller than those mentioned in previous studies using the spray-dry method. In addition, we succeeded in making substitutions of rare earth elements for Cu, which had not been successful in the conventional solid-state reaction or spray-dry methods.

α-Cu₂V₂O₇, β-Cu_1.8Zn_0.2V₂O₇, and a rare-earth-doped system were prepared by the sol–gel method.²⁷ The sol–gel method is a synthesis method for obtaining a solid substance from a sol state in which fine particles, such as colloids, are dispersed in a gel solution having no fluidity. Powders of Cu(NO₃)₂·3H₂O, Yb(NO₃)₂·4H₂O, Sm(NO₃)₂·6H₂O, Ce(NO₃)·2H₂O, ZnO, and V₂O₅ (with purities of 99.9%) were weighed at an appropriate molar ratio. 2 g of the combined powders was mixed together with 100 ml distilled water and 6 g of citric acid (C₆H₈O₇) and was dissolved by stirring at room temperature (295 K) for 3 h to obtain a green solution. To this solution, 5 g of polyethylene glycol (polymerization degree: 500 000) was added, and the mixture was stirred at 80 °C for 30 min. Thereafter, the mixture was rapidly cooled to 20 °C using ice in order to form a gel. The gel was heated in air at 400 °C for 5 h to decompose the citric acid and then heated in air at 650 °C for 10 h to obtain the desired ceramic fine particles. For comparison, samples were also prepared by the spray-dry method³⁰ and the conventional solid-state reaction method.²⁶ For the spray-dry method, sintering was performed at 670 °C for 10 min. For the conventional method, sintering was performed at 670 °C for 10 h. All obtained samples were identified by powder x-ray diffraction (XRD) (Rigaku; RINT2000) using Cu Kα radiation at 295 K. The variable temperature XRD was measured using a synchrotron radiation (λ = 0.6521 Å) powder x-ray diffractometer in the BL5S2 of the Aichi Synchrotron Radiation Center; the results were used to derive the crystallographic parameters. The particle size was evaluated by measuring the particle size distribution using a laser diffraction/scattering type particle size distribution device (Horiba; LA-950V2) and by observing the surface with a scanning electron microscope (SEM) (Keyence; VE-7800). In the particle size distribution measurement, 0.2% by weight of sodium hexametaphosphate was added to prevent aggregation. In order to verify the thermal expansion suppressing ability of the ceramic fine particles, the obtained powder was composited with a one-component epoxy resin (Konishi; E39) by a rotational molding method.¹³ Thermal expansion was evaluated in the range of 100 K–400 K by a strain gauge method (Kyowa; type KFL).⁹ 

Figure 1 shows the results of XRD for the samples prepared by the sol–gel, spray-dry, and conventional methods. A small peak due to impurities of Cu₃V₂O₈ was observed in the spray-dried sample; however, no impurities were observed in the conventional and sol–gel samples within the range of this measurement. In each case, the peaks could be indexed to the monoclinic C2/c space group. Supplementary material Fig. S1 shows an enlarged view (2θ = 11.6°–12.4°) of the variable temperature XRD at 300 K–400 K obtained by synchrotron radiation. Comparing the spray-dry and sol–gel methods, it can be confirmed that there are no changes in peak position, shape, or shift with respect to temperature. The crystallographic parameters were obtained by Rietveld analysis³¹ of the synchrotron radiation XRD results (supplementary material Fig. S2). Comparing the lattice constants at 300 K (Table I), there is no difference in value between the three preparations. Figure 2 shows (a) the lattice constants and (b) changes in unit cell volume Δv/v in the T range of 100 K–700 K. Previous research has shown that there is no difference in crystallographic parameters between the conventional and spray-dry methods.³⁰ In addition, the same crystallographic material can be obtained even when made by the sol–gel method.

From the graph of the variable temperature XRD (supplementary material Fig. S1), it was confirmed that the sample prepared by the sol–gel method had a wider peak width than that of the spray-dry method. The coherence length was estimated from the shape of the (200) peaks (2θ to 24°) using Scherrer’s equation. The coherence length was derived by fitting the peaks with a Gaussian function. The crystallographic coherence length is an index for the crystallinity of single crystal grains forming the material structure and is different from the actual particle size of the ceramic fine particles discussed in the next paragraph. It should be noted here that the coherence length determined by Scherrer’s equation gives a lower limit on the size of a crystal grain and the actual crystal grain may be larger. The value obtained for the spray-dry method was 54.6 ± 1.1 nm, whereas the value obtained for the sol–gel method was smaller, at 40.5 ± 0.9 nm. This difference will be discussed later.

Figure 3 shows the results of the particle size distribution evaluation by the laser diffraction/scattering method. Figure 3(a) shows the volume frequency particle size distribution, and Fig. 3(b) shows the addition integral. The median diameter of the volume frequency (hereinafter referred to as the median diameter D₅₀) was 2.7 μm for the spray-dry method,³⁰ but it was only 1.5 μm for the sol–gel method. The particle size D₉₀ at a volume frequency of 90% was 4.9 μm for the spray-dry method³⁰ and 3.5 μm for the sol–gel method. This suggests that the particles produced via the sol–gel method have a narrow and uniform particle size distribution. Variations in grain size and shape can be confirmed by the surface SEM photographs (supplementary material Fig. S3).

In order to verify the ability of the obtained ceramic particles to suppress thermal expansion, a composite material with an epoxy resin was prepared. Since this material displays NTE as a result of a material structure effect consisting of crystal grains showing anisotropic thermal deformation and voids, crystallographic methods such as powder XRD cannot be used for evaluation of NTE. It is also difficult to directly evaluate the thermal expansion of each fine particle using a dilatometer. Therefore, a method of indirectly evaluating the thermal expansion of the fine particles, through the measurement of the thermal expansion of the epoxy resin in which the fine particles are dispersed, was adopted.

Supplementary material Fig. S4 shows the linear thermal expansion of the 50%-β-Cu_1.8Zn_0.2V₂O₇/epoxy composite material using the fine particle sample prepared in this study, in addition to the pure epoxy resin and the β-Cu_1.8Zn_0.2V₂O₇ sintered body.³⁰ Figure 4(a) shows an enlarged view of the linear thermal expansion of the resin composite material. Figure 4(b) shows an SEM image of the composite. The diameter and shape of the filler, as well as its homogeneous distribution, are confirmed from the image. The volume fraction was determined from the measured weights of the materials, with the specific gravities of β-Cu_1.8Zn_0.2V₂O₇ and the epoxy resin being 3.86 and 1.15, respectively. At the time of calculation, the porosity p of the epoxy resin was taken as 0. The porosity of β-Cu_1.8Zn_0.2V₂O₇ was assumed to be p = 0.2, based on the example of Ca₂Ru_0.92Fe_0.08O₄,³² another microstructural-type NTE material, and preliminary calculations made using the Archimedes method. The details are described in the previous study.³⁰ For comparison, the prediction for the composite thermal expansion (rule of mixture; ROM),² for which the contributions of the spray-dried sintered body and the epoxy resin were weighted at a volume ratio of 50:50, is shown by the dashed line in Fig. 4. Significantly, the linear thermal expansion of the composites was similar, regardless of the fabrication method. In many NTE materials, including Ca₂RuO₄, which is categorized under the microstructural type, the thermal expansion of the resin composite material is slightly closer to that of the filler material than ROM.^13,32–37 This is because the thermal expansion of an inorganic material having a large elastic modulus (the NTE material) is more strongly reflected by the elastic interaction through the interface.² The thermal expansion of the composite material containing β-Cu_1.8Zn_0.2V₂O₇ microparticles was similar to ROM, but it was significantly closer to that of the filler. This suggests that the material structure is maintained even with the fine particles dispersed in the resin and that NTE close to that of the bulk sintered body used for calculating the ROM is maintained. Therefore, it can be concluded that particles with a size of about 1 μm produced by the sol–gel method have the same thermal expansion suppression ability as particles produced by the spray-dry and conventional methods.

Next, we will discuss the effects of rare-earth doping on the thermal expansion of α-Cu₂V₂O₇. In this study, rare earth substitutions for Cu were realized by the sol–gel method. Rare earth substitutions have not been successful because of the poor reactivity of the raw materials in the conventional solid–state reaction method and the low solubility of the elements in aqueous citric acid solution in the spray-dry method. Supplementary material Fig. S5 shows the results of the powder XRD using Cu Kα radiation. When doped with 5% Yb, Sm, or Ce, the peaks could be indexed to the orthorhombic Fdd2 α phase, and a peak shift due to the rare-earth doping was observed [supplementary material Fig. S5(b)]. There were no traces of impurities found. In addition, the bulk thermal expansion characteristics were measured (Fig. 5). NTE was lower for the pure α-Cu₂V₂O₇ prepared by the sol–gel method (solid circle) than that which was prepared by the conventional method (open square). However, the NTE characteristics were improved by the substitution of rare earth elements, and the same NTE (α = −5.3 ppm/K) as α-Cu₂V₂O₇ prepared by the conventional method²⁶ was obtained with the substitution of 5% Ce. The bending in the linear thermal expansion curves, which is particularly noticeable for the Sm-doped sample, might be a sign of a structural phase transition. The difference in the ionic radius between the transitional elements and the lanthanides induces internal stress, which in turn possibly causes a structural phase transition.²⁷ Further investigation is necessary.

Ceramic particles of about half the size of those obtained by the spray-dry method were produced by the sol–gel method. The holding time at the maximum temperature is 10 h for the sol–gel method, whereas it is only 10 min for the spray-dry method. In general, it is thought that the longer the heating time, the more the grain growth proceeds; therefore, although considering that the latter has a maximum temperature 20 °C higher (650 °C for the sol–gel method but 670 °C for the spray-dry method), it seems unreasonable to explain the difference in particle size by the difference in sintering time. Rather, the smaller particle size produced by the sol–gel method is attributed to the presence of nitrate in the raw materials during the citric acid aqueous solution stage, yielding a smaller proportion of β-Cu_1.8Zn_0.2V₂O₇ components at the end. The reduction in the size of the ceramic fine particles corresponds to a reduction in the number of crystal grains, a reduction in the size of the crystal grains, or both. To determine this, it is necessary to examine the internal structure of the ceramic fine particles in detail.

Considering this difference in particle size, the difference in coherence length when prepared by the sol–gel method is slight, but the value is smaller than that of the spray-dry method. The coherence length corresponds to the size of the region where the crystal orientations are aligned, that is, the size of the region that can be regarded as single-crystalline. This does not necessarily correspond to the particle size of the powder sample used in the diffraction experiment. If the crystallinity is poor, the coherence length may be smaller than the size of the grains. Conversely, no matter how good the crystallinity, the coherence length does not exceed the size of the grain. In the spray-dry method, the hold time at the maximum temperature is extremely short, 10 min, so it should be expected that the sol–gel method would be better in terms of crystallinity. Even so, the shorter coherence length of the sol–gel method may be indicative of the difference in particle size rather than a difference in crystallinity.

In the sol–gel method, small proportions of the materials are allowed to react to produce β-Cu_1.8Zn_0.2V₂O₇ in order to obtain a small grain size with good crystallinity. Conversely, in the spray-dry method, the ratio of the materials is large, resulting in larger grain sizes, but divided into several domains. The reason for the smaller ceramic fine particles obtained by the sol–gel method discussed in the previous paragraph may be because the single crystal grains constituting the ceramic fine particles are smaller to begin with. This result does not necessarily mean that the sol–gel method is generally superior to the spray-dry method for producing fine particles. It is known that the particle size and the crystallinity depend on various parameters within a method for obtaining a solid through an aqueous solution phase such as the sol–gel, spray-dry, or hydrothermal methods. In particular, the particle size highly depends on the pH of the solution.^38,39 For β-Cu_1.8Zn_0.2V₂O₇, citric acid has been used in both the sol–gel and the spray-dry methods, without the pH being optimized. Furthermore, the particle size depends also on the heat treatment and synthesis temperatures for removing organic substances.^28,29 In the future, in addition to consideration of raw materials, such as coordination complexes, polymers, and acids, optimization of the pH of the solution, the decomposition temperature of organic compounds, and the maximum temperature for the main sintering will be required.

In contrast to the sol–gel method, in a previous study,³⁰ the particle size could be controlled to about 1 μm through pulverization with a ball mill; however, the electronic properties, crystallographic properties, or both may be altered. In the case of fine particles, it is undesirable for the function of NTE to be lost by way of destruction of the material structure. In addition to this, ball-mill grinding tends to produce particles with irregular sizes or shapes and necessitates a sizing process. For these reasons, it is difficult to obtain NTE fine particles via mechanical pulverization, and it is meaningful to manufacture fine particles using a sol–gel method that can form a material structure on a small scale from the beginning and concentrate the particle size distribution in a narrow range.

It is said that the anisotropic thermal deformation of the unit cell that generates the NTE of Cu₂V₂O₇ and related vanadium oxides is due to the orbital degrees of freedom of the Cu²⁺ (3d)₉ electrons.^23,25 Therefore, controlling the Cu 3d electronic state is useful for deepening the understanding of this phenomenon and achieving higher functionality.⁴⁰ In particular, if rare earth elements are doped into the system, then it is expected that the magnetic interaction between Cu and rare earth elements can be utilized to produce a wider variety of physical properties. In this experiment, although the performance of the doped system did not exceed that of the pure system, the bulk NTE was clearly changed in accordance with the rare earth substitution. Future detailed crystallographic analysis of rare-earth substituted systems will help elucidate the role of crystallographic parameters in microstructural effects.

By the sol–gel method, the same level of thermal expansion suppression capability as before was achieved with β-Cu_1.8Zn_0.2V₂O₇ ceramic fine particles of about 1 μm in size. The sol–gel method was also shown to be effective for rare-earth doping of Cu₂V₂O₇. The present results open the way to industrial applications of Cu₂V₂O₇-based materials as thermal expansion inhibitors for local areas.

See the supplementary material for temperature-dependent XRD profiles of β-Cu_1.8Zn_0.2V₂O₇ (Fig. S1), Rietveld refinement of XRD data of β-Cu_1.8Zn_0.2V₂O₇ (Fig. S2), microscopic images of β-Cu_1.8Zn_0.2V₂O₇ (Fig. S3), linear thermal expansion of the composites (Fig. S4), and XRD profiles of rare-earth-doped α-Cu₂V₂O₇ (Fig. S5).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

The authors are grateful to K. Otsuka, M. Mitamura, M. Ozeki, and Y. Mizuno for their help with the experiments. This work was financially supported by the Grants-in-Aid for Scientific Research (Grant Nos. JP17H02763, JP18H01351, and JP19H05625) from MEXT, Japan. Synchrotron powder x-ray diffraction experiments were conducted at the BL5S2 of Aichi Synchrotron Radiation Center, Aichi Science and Technology Foundation, Aichi, Japan (Proposal Nos. 201903064, 201904078, and 201905016).