XRD analysis was employed to detect phase evolution of the products during the variation of hydrothermal temperatures (Fig. 1). It is shown that sample-100 (obtained at 100℃) can be well indexed as hexagonal (NH₄) _xWO_{3- y} (h-(NH₄) _xWO_{3- y}) with the lattice parameters of a = b = 0.7392 nm, c = 0.7512 nm, α = β = 90° and γ = 120° (JCP-DS Card 42-0452). This sample was observed to be blue, indicating the appearance of W⁵⁺. When the reaction temperature increased up to 120℃, a tiny XRD peak centered at 18.1° appeared (not shown). Further increasing reaction temperature up to 140℃ and 180℃, this XRD peak is strengthened in intensity. Simultaneously, the XRD peak centered at ~24° gradually divided into three peaks, characterizing the appearance of orthorhombic WO₃·1/3H₂O phase (o-WO₃·1/3H₂O, JCPDS Card 54-1012, a= 1.2547 nm, b= 0.7737 nm, c= 0.7345 nm, α = β = γ = 90°) in products. This XRD results indicated that reaction temperature played a crucial role in controlling the phase composition of the final products. In our system, h-(NH₄) _xWO_{3- y} was firstly synthesized at 100℃, which partially transformed into o-WO₃·1/3H₂O phase at high reaction temperatures (≥ 120℃), resulting in samples with mixed phases. Additionally, it is worthy to point out that the XRD peaks of h-(NH₄) _xWO_{3- y} are obviously broaden than those of o-WO₃·1/3H₂O, implying that the former crystallite should possess a much finer size.

Figure 2 shows typical SEM images for sample-100, -140, and -180. As shown in Fig. 2(a), the sample obtained at 100℃ (sample-100) consisted of micro-scopic particles, which were further built by uniform nano-rods with lengths of ~100 nm. No other morphology has been detected in this product. For samples obtained at 140℃ (sample-140), in a stark contrast, a few micro-rods with lengths of ~2 μm appeared besides the uniform nano-rods (Fig. 2(c)-(d)). High-magnification SEM images illustrated that the micro-rods possessed orthorhombic end faces (pointed by a red arrow in Fig. 2(d)). Moreover, these orthorhombic micro-rods became larger in both volume-percentage and sizes with a higher reaction temperature of 180℃ (sample-180, Fig. 2(e)-(f)). This morphology changes of the orthorhombic micro-rods can reasonably explain the XRD signal evolution of o-WO₃·1/3H₂O phase in Fig. 1. Therefore, we assigned the orthorhombic micro-rods and tiny nano-rods in our samples to o-WO_{3-- x}·1/3H₂O and h-(NH₄) _xWO_{3- y}, respectively.

The morphological and phase features of all samples had been further confirmed by TEM observation (Fig. 3). Low-magnification TEM images clearly showed that sample-100 comprised of short nano-rods with a uniform diameter of ~9 nm and lengths of 20~100 nm. The continuous lattice fringes in TEM images suggested that each nano-rod is a single h-(NH₄) _xWO_{3- y} crystallite with a [001] growth direction. Analogous to SEM results, orthorhombic micro-rods were detected in sample-140 and sample-180, besides the tiny nano-rods. High- magnifi-cation TEM images and selected area electron diffraction (SAED) patterns indicated that nano-rods in sample-140 and sample-180 were still [001]- direction grown h-(NH₄) _xWO_{3- y} crystallites; On the other hand, the orthorhombic micro-rods were found to be o-WO₃·1/3H₂O crystallites grown along the [001] direction (Fig. 3(f)). This TEM results were in good line with the XRD and SEM data.

Fig. 1

	Figure Option ViewDownloadNew Window
	Fig. 1 XRD patterns of sample-100, -140 and -180

Fig. 2

	Figure Option ViewDownloadNew Window
	Fig. 2 SEM images of sample-100 (a, b), -140 (b, d) and -180 (e, f)

Fig. 3

	Figure Option ViewDownloadNew Window
	Fig. 3 TEM images of sample-100 (a, b), -140 (c, d) and -180 (e, f)

Figure 4 shows the thermo-gravimetry (TG) curves of sample-100, -140 and -180 recorded in air at a heating rate of 10 ℃/min. As shown in Fig. 4, all samples manifested three mass evolution processes located at 100-273, 273-480 and 480-900 ℃, respectively. The end product after TG measurement is confirmed to be monoclinic WO₃ by XRD technique. These mass evolution could be ascribed to the desorption of adsorbed H₂O, the release of NH₄⁺/crystal H₂O, and the elimination of oxygen deficiency^{[ 21]}. It is shown that the volumes of absorbed H₂O (0.94wt%-1.04wt%) are comparable in three samples. Whereas, the volume of NH₄⁺/crystal H₂O (from 1.31wt% to 2.07wt%), as well as oxygen deficiency (0.33wt% to 1.05wt%), increased obviously in samples with reaction temperature elevation. In our samples, the oxygen deficiency was introduced by the addition of reducing agent N₂H₄·2HCl in the precursor solution. Oxygen deficiency has been proved to be beneficial for the improvement of NIR shielding ability by different researchers^{[ 8, 22]}.

Figure 5 shows the FTIR spectra of sample-100, -140 and -180. These spectra were very similar in profiles, implying the appearance of analogous functional groups in these samples. Generally, the absorption bands centered at 1401 and 3200 cm^-1 were indicative of bending ( δNH) and stretching ( υNH) vibration of NH₄⁺ ions^{[ 21]}. With reaction temperature elevated, absorption peaks of NH₄⁺ ions became weak, most possibly due to the reduction of NH₄⁺ ions in products. Combining with the TG, XRD and FTIR data, we believed that the h-(NH₄) _xWO_{3- y} products became oxygen deficient with reaction temperature increase. Consequently, some of the intercalated NH₄⁺ ions in h-(NH₄) _xWO_{3- y} were gradually replaced by intercalated H₂O in a charge balance way (pay attention that both oxygen vacancies and NH₄⁺ were positively charged). Finally, the h-(NH₄) _xWO_{3- y} transformed into oxygen deficient o-WO₃·1/3H₂O.

Fig. 4

	Figure Option ViewDownloadNew Window
	Fig. 4 TG plots of sample-100, -140 and -180 recorded in an air flow at a heating rate of 10 ℃/min

Fig. 5

	Figure Option ViewDownloadNew Window
	Fig. 5 FTIR spectra of sample-100, -140 and -180 ( υ represents stretching vibration and δ represents bending vibration)

On the basis of XRD, SEM, TEM, TG and FTIR results, a mechanism has been extracted to explain the transformation from h-(NH₄) _xWO_{3- y} to o-WO₃·1/3H₂O, as shown in Fig. 6 as well as Equation (2):

(2)

Where h and o represented hexagonal and orthorhombic structure type phases. By heating up to 100℃, the tungsten precursor in hydrothermal solution decomposed instantaneously to form plenty of monomers, and thus leading to a highly supersaturated initial reaction system. Therefore, a nucleation process was followed and large number of nucleus appeared. The growth of nucleus consumed the monomers quickly, restraining the further growing of the particles. As a result, numerous tiny h-(NH₄) _xWO_{3- y} nano-rods were formed (Fig. 6(a)). Consequently, some of the nano-rods began to grow by oriented attachment along the radial directions of nano-rods, under the driving forces of surface energy reduction (Fig. 6(b)). At this stage, no characteristic peak of o-WO₃·1/3H₂O phase was observed in the XRD pattern. When the reaction temperature increased up to 140℃ and 180℃, large, orientedly-attached h-(NH₄) _xWO_{3- y} rods transformed into o-WO₃·1/3H₂O ones (Fig. 6(c)). The o-WO₃·1/3H₂O rods were observed to further attached into bigger ones during following reactions (Fig. 2(c)-(f), 3(c)-(e) and 6(d)-(e)), finally resulting in samples consisted of tiny h-(NH₄) _xWO_{3- y} nano-rods and extremely large o-WO₃·1/3H₂O micro-rods (Fig. 3(e) and 6(e)).

In both crystal structures of h-(NH₄) _xWO_{3- y} and o-WO₃·1/3H₂O, distorted [WO₆] octahedra are connected in the a-, b-, and c-direction, thereby forming a three- dim-en-sional structure, in which, NH₄⁺ and H₂O randomly occupies the hexagonal channels of h-(NH₄) _xWO_3-y and dodecahedron of o-WO₃·1/3H₂O^{[ 23]}. According to these theoretical models, one can find that the hexagonal framework of [WO₆] octahedra could be transformed into orthorhombic one by simply shifting the neighbored a- b planes by 1/2 a.

In this study, the direct TEM images, which were capable to exhibit the phase transition process from h-(NH₄) _xWO_{3- y} to o-WO₃·1/3H₂O, were failed to obtain. Nevertheless, noticing the dramatically difference of sizes (and thus specific surface areas) between o-WO₃·1/3H₂O and h-(NH₄) _xWO_{3- y} crystallites, we speculated that finite size effects may play a crucial role during the phase transformation. Generally, occurrence of phase transfor-ma-tions was triggered by Gibbs energy decrease. However, the surface energy variation between two phases should also be taken in account for systems with high specific surface area^{[ 24]}. For our samples, the specific surface energy of h-(NH₄) _xWO_{3- y} is possibly lower than that of o-WO₃·1/3H₂O. Therefore, two energy evolution, surface energy increase (Δ E_s) and Gibbs energy decrease (Δ E_g), co-determined the occurrence of the phase transformation. With the decrease of crystallite size, Δ E_s becomes the dominating fact and thus makes the transformation from h-(NH₄) _xWO_{3- y} to o-WO₃·1/3H₂O energetically unfavorable, resulting in products comprised of small h-(NH₄) _xWO_{3- y} and large o-WO₃·1/3H₂O crystallites. Of course, more detailed mechanism should be further investigated.

Fig. 6

Figure Option ViewDownloadNew Window

Fig. 6 Scheme illustration of the transformation mechanism from h-(NH4) xWO3- y to o-WO3·1/3H2O (f) and corresponding TEM proofs (a-e). Panels (b) and (d) are the enlarged images of (a) and (c), respectively, and they clearly reveal the undergoing oriented attachment process between primary rods. This process has also been schematically illustrated in the framed regions of red circles in (a) and (c)

Figure 7 shows the transmittance spectra of a bare fused silica glass substrate, sample-100, -140 and -180. The T_lum, T_sol, and T_infr (summarized in Table 1) were calculated with transmittance spectra and equation (1). As expected, the fused silica glass substrate was highly transparent for both visible and near infrared light (transmittance ~91%). For all our samples, the T_lum were still noteworthy (56.1%- 70.3%). The relatively poor T_lum of sample-180 could be ascribed to the light scattering effects by its large crystallites. The transmittance of all samples decreased rapidly in wavelength ranges of 350-450 nm, due to the intrinsic absorption of tungsten oxides^{[ 9, 22, 25]}.

Significantly, sample-100, -140 and -180 exhibited dra-matic different T_infr of 25.5%, 62.7% and 66.8%, respec-tively. It is widely reported that the conductive h-(NH₄) _xWO_{3- y} nano-particles could introduce a selective absorption of NIR light by localized surface Plasmon resonance^{[ 8, 10]}. The loss of NIR shielding ability with reaction temperature elevation could be attributed to the increase of the large o-WO₃∙1/3H₂O phase, which offer weak or no LSPR.

Fig. 7

	Figure Option ViewDownloadNew Window
	Fig. 7 Transmittance spectra of a bare fused silica glass substrate, sample-100, -140 and -180

Table 1

Table 1

Table 1 Optical parameters of a bare fused silica glass substrate, sample -100, -140 and -180 Sample Transmittance / % Tlum Tsol Tinfr Silica substrate 92.8 93.0 93.4 Sample-100 67.6 45.9 25.5 Sample-140 70.3 65.9 62.7 Sample-180 56.1 60.5 66.8

Table 1 Optical parameters of a bare fused silica glass substrate, sample -100, -140 and -180