Depressed haze and enhanced solar modulation capability for VO₂-based composite films with distinct size effects

Abstract

For decades, scientists have devoted themselves to developing vanadium dioxide (VO₂) films for thermochromic smart windows, but there remains a major challenge to their large-scale implementation owing to their limited solar modulation ability. Here, we show a new method for improving their solar modulation ability via increasing optical switching in the visible region upon phase transition. VO₂-based composite films that exhibited reduced haze (haze_{555 nm} down from 46.5% to 2.2% @RT), improved luminous transmittance (up from 9.6% to 38.5% @RT) and enhanced solar modulation ability (ΔT_sol up from 8.1% to 16.9% @RT) have been fabricated by a convenient and controllable approach. With this method, the mean sizes of VO₂ particles can be easily tuned from 150 nm to 40 nm by varying the preparation parameters. Via a combination of the optical transmittance spectrum and Mie calculations, we first observed the behavior of optical switching in the visible region upon phase transition, which is size-dependent. Moreover, distinctive size effects on variations in the band gap energy and specific interband transitions were studied based on the absorption spectra of VO₂ composite films.

---

1. Introduction

Owing to its near-room-temperature metal–insulator transition (MIT), vanadium dioxide (VO₂) has attracted much interest over the last several decades since it was first discovered by Morin.¹ At around 68 °C (for undoped bulk VO₂), the MIT phenomenon is accompanied by a structural transition from a high-temperature tetragonal rutile (R) phase to a low-temperature monoclinic (M₁) phase. The high-symmetry tetragonal structure formed above the critical temperature is based on adjacent VO₆ octahedra sharing edges along the c-axis with uniform V–V bond distances. Upon cooling through the phase transition, a slight distortion of the VO₆ octahedra gives rise to alternate long and short V–V bonds that form adjacent pairs of V atoms.^2,3 Abrupt changes in optical transmittance and electrical conductivity can be observed across this phase transition. For instance, below the critical temperature, the insulating state of vanadium dioxide mostly appears transparent in the infrared region. However, above the critical temperature, the optical transmission of VO₂ in the near infrared region (NIR) exhibits a sharp decrease due to the transformation into the metal phase.⁴ As it possesses such unique properties, VO₂ has been regarded as an exclusive and potential material for fabricating intelligent devices such as energy-saving smart windows, optical storage media and Mott field effect transistors.^5–8

Regarding the most promising application, namely, for energy-efficient smart windows, there are several substantial barriers to achieving low-cost and large-scale industrial production. Typically, VO₂ is processed directly in the form of dense films by means of sputtering,⁹ vapor deposition,¹⁰ ion implantation¹¹ and solution-based processes.¹² All these methods require elevated temperatures of at least 400–500 °C for the crystallization and growth of VO₂, so that coatings can only be deposited on heat-resistant substrates. Furthermore, precise control of the atmosphere is difficult in large chambers, where a unique stoichiometry of V–O is generated.

One of the most effective approaches to overcoming these problems is to prepare VO₂ (M₁) powders beforehand^13–15 and then produce films by coating a dispersion composed of VO₂ (M₁) particles and a polymer matrix on a substrate. Various morphologies and sizes of VO₂ nanostructures have been synthesized by wet chemistry methods.^16–20 Moreover, control of the growth of high-quality VO₂-based nanoparticles in combination with multifunctionality has been achieved recently. Li et al.²¹ reported a one-step hydrothermal synthesis, which was assisted by seeding with TiO₂, of Mo-doped VO₂ (M)/TiO₂ composite nanocrystals. Gao et al.²² crystallised mesoporous TiO₂ (A)–VO₂ (M/R) nanocomposite films with self-cleaning and excellent thermochromic properties. Jin et al.²³ reported the synthesis of novel VO₂ (M)/SnO₂ heterostructured nanorods, i.e., nanosized SnO₂ particles were successfully grown on the surface of VO₂ nanorods without aggregation. The band gap of the VO₂ crystals was widened from 0.75 to 1.7 eV. A composite film based on these novel VO₂ (M)/SnO₂ heterostructured nanorods achieved a high visible transmittance of up to 35.7% and an IR modulation of 56% at 2500 nm. A two-step method for producing a VO₂-based composite film could separate the heat treatment required for the crystallization of the VO₂ (M) phase from that used in the process for coating the film,^24–26 which provides great possibilities for the application of VO₂ in smart windows. Furthermore, the films prepared by this method can easily be deposited on non-specific substrates, for instance, ordinary glass, plastics²⁷ and even textiles, which could widen the range of applications of VO₂. Thirdly, the two-step method shed light on the large-scale commercialization of VO₂ smart coatings. Based on experimental results and theoretical estimations, substantial energy-saving effects of VO₂-based composite coatings in smart windows applications and the management of solar heat have been anticipated by several groups.^28,29 In its report, the International Energy Council's predictions have estimated that with ∼2 billion m² of windows worldwide coated with smart coatings, energy savings in the building and automotive fields would be equivalent to CO₂ reductions of about 100 million tons.

However, traditional VO₂ fillers are often on a micron scale in size, such as those investigated by Valmalette et al.,³⁰ which will give rise to substantial light scattering, resulting in a severe deterioration in visual transparency, namely, the generation of haze.^31,32 A reduction in haze is necessary for VO₂-based composite films owing to the fact that casement windows should be very clear and transparent for capturing visual information. There are two main reasons for the appearance of haze caused by scattering, which include a mismatch in refractive indices (RI) between the nanoparticles and the polymer matrix, as well as the large dimensions of the internal fillers in composite films. The mismatch in RI can be compensated for by reducing the particle size to significantly below the wavelength of visible light.³³ Theoretical studies have indicated that the loss of light intensity induced by scattering can be negligible if the particle size is reduced to below 100 nm.³⁴ In addition, for applications in smart windows, VO₂ films have two major drawbacks, which consist of their low luminous transmittance when considerable thermochromism is achieved and insufficient regulation of solar energy owing to limited switching in the NIR region upon the MIT.^5,35–38 In fact, theoretical calculations performed by Xie et al.³⁸ have shown that dilute composites with VO₂ nanoparticles embedded in dielectric hosts exhibit improved luminous transmittance and modulation of solar energy, which suggests a possible practical application of this material in smart windows. Recently, widening the band gap between the O 2p π and π* orbitals in VO₂ by means of doping with Mg or F has been proved to be an effective strategy for achieving improved luminous transmittance, as well as a colourless appearance.^39,40 This is because strong intraband and interband absorption in the short-wavelength region is responsible for low visible transmittance. However, the doping-based process led to a deterioration in the solar modulation efficiency. In contrast to the dominant transformation in the NIR, little attention has been paid to optical switching in the visible region, which comprises nearly half of solar radiant heat.⁴² Based on optics calculations, Xu et al. have demonstrated that the changes in luminous transmittance upon optical switching were thickness-dependent, which resulted from interference effects caused by the difference in refractive indices between M₁ and R-phase VO₂ in the visible region.⁴³ Nevertheless, increases in optical switching upon the MIT that led to enhancements in solar modulation efficiency were always at the expense of a reduction in luminous transmittance for both the metallic and the semiconductor states of VO₂.^12,43

Here, we report a convenient and controllable method for the fabrication of VO₂-based composite films, in which the particle size can easily be tuned from tens to more than 100 nanometers. In contrast to traditional methods, the films can easily be deposited on non-specific substrates by a simple and inexpensive coating process. Above all, the as-obtained VO₂ films exhibited reduced haze and improved luminous transmittance in combination with enhanced solar modulation ability. This superior performance could be accomplished by controlling the VO₂ particle size in the composite films via varying the preparation parameters. In this paper, we first investigated the behavior of optical switching in the visible region, which is size-dependent. This can be attributed to the difference in light scattering in the VO₂ film between the high-temperature R phase and the low-temperature M₁ phase. Subsequently, we demonstrated that owing to the decrease in particle size the reduction in absorption is also responsible for the improvement in luminous transmittance. Also, distinctive size effects on variations in the band gap energy and specific interband transitions were studied.

2. Experimental

2.1. Synthesis

There are several kinds of crystalline phases in the vanadium dioxide system, which mainly include monoclinic VO₂ (B), monoclinic rutile-type VO₂ (M₁) and tetragonal rutile-type VO₂ (R), which can be interconverted into each other under certain conditions. In addition, according to a survey, an additional antiferromagnetic Mott insulating M₂ phase can be stabilized by doping or applying uniaxial high pressure in the [110] direction of the R phase of pure VO₂.³

In this paper, initially VO₂ (B) nanobelts were synthesized under hydrothermal conditions using NH₄VO₃ and oxalic acid as precursors. All chemicals were purchased from Sinopharm Chemical Reagent Co., Ltd and were used without further purification. In a typical procedure, 0.5 g NH₄VO₃ was added to 0.1 M oxalic acid dihydrate under mild stirring. The resulting mixture was continuously stirred until a clear blue solution formed, and then it was transferred into a Teflon-lined autoclave with a stainless-steel shell. The autoclave was kept at 180 °C for 48 h and then allowed to cool to room temperature. The dark blue precipitate, namely, VO₂ (B), was collected by centrifugation and washed several times with deionized water and anhydrous alcohol. Subsequently, VO₂ (B) powders were transformed into high-temperature VO₂ (R) by an annealing process at 500 °C for 30 min under an argon atmosphere, which yielded VO₂ (M₁) nanoparticles upon cooling to room temperature. The as-prepared VO₂ (M₁) nanoparticles were dispersed ultrasonically in butyl acetate for 30 min in the presence of an appropriate amount of polyelectrolyte (BYK, which was bought from Daiquan Technology Co.), and the solid content of VO₂ in the starting suspension was 5%. Then, the blending solution was processed by grinding in a planetary ball mill (QM-3SP4, Nanjing Nanda Instrument Plant; the material of both the ball mill jar and the beads was zirconia, and the diameter of the ball mill beads was about 0.4 mm) at 600 rpm for periods of 1 h, 4 h, 8 h and 12 h. Finally, thermochromic composite films were obtained by spin-coating the VO₂/polymer composite dispersion on ordinary glass.

2.2. Characterization

X-Ray diffraction (XRD) patterns were collected using a powder diffractometer operated at 40 kV and 40 mA with Cu Kα radiation (λ = 0.154 nm). Scanning electron microscopy (SEM) was performed using a Hitachi S-4800 FESEM microscope with an Oxford energy-dispersive X-ray detector. Optical transmittance and absorption spectra were recorded with a UV-vis-NIR spectrophotometer (Lambda 750) with an attachment of a controlled temperature sensor in contact with the films. Specifically, the spectrophotometer was equipped with an integrating sphere to measure the total transmittance. The range of temperatures was set at between 20 °C and 100 °C. The film thickness was determined using a stylus profiler measuring system (Bruker, DektakXT).

3. Results

3.1. Morphology and structure of VO₂ nanoparticles

With respect to the experimental protocol, firstly, metastable VO₂ (B) was prepared using NH₄VO₃ and oxalic acid as precursors by a hydrothermal method. Subsequently, an annealing process transformed VO₂ (B) into the high-temperature rutile VO₂ (R) phase, which yielded monoclinic VO₂ (M₁) via cooling to room temperature and undergoing the metal–insulator phase transition. Fig. 1a shows an SEM image of the as-prepared hydrothermal product, which exhibits a 1D nanobelt structure. The length of the belts varies from 500 to 1500 nm and their width ranges from 100 to 200 nm. In fact, VO₂ nanostructures obtained directly in hydrothermal conditions have a strong preferential growth orientation along the [110] direction, according to previous research into one-dimensional (1D) nanowires, nanobelts, nanorods and nanotubes.^44–47

Fig. 1 SEM images of metastable VO₂ (B) nanobelts obtained by hydrothermal treatment at 180 °C for 48 h (a) and thermochromic VO₂ (M1) nanostructures obtained after annealing at 500 °C for 30 min (c); the corresponding experimental XRD patterns and standard patterns from JCPDS cards of VO₂ (B) (b) and VO₂ (M) (d), respectively.

As shown by the XRD pattern in Fig. 1b, the peaks of the as-obtained product resulting from hydrothermal treatment at 180 °C for 48 h match well with those of the standard JCPDS card no. 65-7960 and are indexed to monoclinic VO₂ (B) with the C_2m space group. Acting as one of the metastable phases of vanadium dioxide, the B phase is generally an intermediate state of the R phase or is often isolated as the major product of hydrothermal syntheses, which is of great interest owing to its layered structure and promising properties as a cathode material for Li-ion batteries.^48,49 The transformation from the VO₂ (B) phase into the VO₂ (R) phase was deemed to be an order–disorder phase transition by Leroux, who believed that VO₂ (B) easily accommodates a high density of oxygen vacancies via crystallographic shear.^44,50 However, above a critical threshold, the tetragonal rutile structure became advantageous. Upon annealing, the metastable monoclinic VO₂ (B) phase was transformed into tetragonal VO₂ (R), which yielded monoclinic VO₂ (M₁) upon cooling to room temperature. The XRD pattern in Fig. 1d confirms the formation of the monoclinic VO₂ (M₁) phase (JCPDS card No. 72-0514), and the relative intensities of the peaks are strong, which suggests high crystallinity. Nevertheless, thermal processing in argon has a pronounced effect on the morphology of vanadium dioxide.⁵¹ As illustrated in Fig. 1c, the aforementioned nanobelts changed into irregular oblate spheroids or cylinder-like structures after annealing. These spheroid bodies were partly interconnected or isolated and their widths were comparable to those of the VO₂ (B) nanobelts that are shown in Fig. 1a. Indeed, previous studies have deduced that recrystallization, accumulation and melting would occur during the annealing process. Consequently, the morphology of the VO₂ polymorph was modified when it was transformed from the B phase into the R and M₁ phases. It is proposed that a low precursor concentration, a high hydrothermal temperature and a long reaction time are beneficial for maintaining the structure of nanobelts.⁵²

3.2. Morphology and structure of VO₂ composite films

An effective and convenient ball milling method was utilized for preparing colloidal suspensions containing VO₂ (M₁) nanoparticles, to which a certain amount of polyelectrolyte was added to maintain high dispersity of the nanoparticles. Then, VO₂-based composite films were deposited onto ordinary glass by a simple spin-coating process. The XRD patterns of VO₂ films prepared using various milling times are shown in Fig. 2. All of the peaks are consistent with the monoclinic VO₂ (M₁) phase. In comparison with the above mentioned VO₂ (M₁) nanoparticles, the intensities of all the peaks of the composite films decreased, which is attributed to the introduction of the amorphous polymer. Furthermore, with an increase in the milling time, the peak intensity decreased and the full width at half maximum (FWHM) of the peaks increased. This means that a longer ball milling time will destroy the crystallinity of VO₂ particles, as well as reducing their grain size.⁵³ Owing to the presence of the insulating polymer matrix, it is difficult to immediately determine the size of VO₂ particles in the composite films. Therefore, the composite films were processed by annealing at 250 °C for 1 h under the protection of an argon atmosphere to reduce the hindrance that arose due to the presence of the polymer. Fig. 3a–d show SEM images of processed films prepared using different ball milling times of 1 h, 4 h, 8 h and 12 h, respectively. Obviously, it is observed that the particle sizes in the films decreased with an increase in the milling time, accompanied by the fact that the morphology tended to become homogeneous and quasi-spherical.

Fig. 2 XRD patterns of VO₂ composite films prepared using various milling times.

Fig. 3 SEM images of processed VO₂ composite films prepared using various milling times: 1 h (a), 4 h (b), 8 h (c) and 12 h (d). The insets show the corresponding estimated particle size distributions.

As listed in Table 1, the mean particle sizes of samples prepared by milling for 1 h, 4 h, 8 h, and 12 h were 150 ± 50 nm, 100 ± 50 nm, 60 ± 20 nm and 45 ± 5 nm, respectively, which were determined by randomly measuring the dimensions of 100 grains in a given micrograph (the insets of Fig. 3). Nevertheless, on the basis of research conducted by Shi et al.,⁵³ the mean particle size at first decreased and then increased owing to the reaggregation of particles as the milling time was increased. In fact, owing to the enormous surface energy of nanoparticles, polymer modifiers are needed for absorption or covalent bonding on the surface of the particles.^21,22 Hence, the polyelectrolyte played an important role in preventing VO₂ nanoparticles from reaggregating in our experiment. Here, we developed a novel and easy method for controlling the particle size in VO₂ films, in which particle sizes of as small as 40 nm can be obtained.

Table 1 Mean particle sizes, thicknesses and haze values at 555 nm of VO₂ composite films prepared using various milling times

Sample	BM time (h)	Mean size (nm)	H555 nm (%)	Thickness (nm)
I	1	150	46.5	759
II	4	100	22.3	648
III	8	60	8	793
IV	12	40	2.2	800

---

3.3. Thermochromic property

Accompanied by a small distortion of the VO₆ octahedra in the VO₂ lattice, VO₂ (M₁) undergoes a fully reversible metal–insulator phase transition at about 68 °C, in which the low-temperature monoclinic phase VO₂ (M₁) is transparent and the high-temperature rutile phase VO₂ (R) is opaque in the near infrared region (NIR).⁴ We employed temperature-dependent optical transmittance spectra to observe the thermochromic properties of VO₂ composite films. Typical optical transmittance spectra (from 250 to 2500 nm) derived using various milling times of 1 h, 4 h, 8 h, and 12 h are shown in Fig. 4a, c, e, and g, respectively. Unsurprisingly, we clearly observe differences between the transmittances recorded at 25 °C (blue curves) and 100 °C (red curves), especially in the NIR, where the differences at 2000 nm are as large as 57.7% (1 h), 54.6% (4 h), 50.1% (8 h) and 48.7% (12 h), respectively. It is observed that the thermochromic properties in the NIR were suppressed in the samples obtained using a longer milling time (Fig. 4b and 5b), which can be rationalized from the fact that, as was previously mentioned, the crystallinity of VO₂ particles was damaged due to prolonged milling.⁵³

Fig. 4 Optical transmittance spectra from 250 nm to 2500 nm recorded at 25 °C (blue curves) and 100 °C (red curves) and the corresponding thermal hysteresis loops at 2000 nm for VO₂ composite films obtained by milling for 1 h (a and b), 4 h (c and d), 8 h (e and f) and 12 h (g and h), respectively. The gray areas indicate normalized values of the solar spectral irradiance (φ_sol). The insets of (b), (d), (f) and (h) show the corresponding differentials of the hysteresis loops. The red curves represent the heating process and the blue curves represent the cooling process.

Fig. 5 Phase transition temperatures upon heating and cooling and width of hysteresis loop versus milling time (a); transition temperature and ΔT at 2000 nm versus milling time (b).

To quantify the potential for applications in smart windows, we calculated the integrated luminous/visible transmittance (T_{lum/vis, 380–780 nm}), integrated solar transmittance (T_{sol, 250–2500 nm}) and visible/solar modulation efficiency (ΔT_{vis, 380–780 nm}/ΔT_{sol, 250–2500 nm}) based on the transmittance spectra. The values were obtained from the following equations:

ΔT_vis/sol = ΔT_vis/sol(25 °C) − ΔT_vis/sol(100 °C) (3)

where T(λ) denotes the optical transmittance at a wavelength of λ, φ_lum is the standard luminous efficiency function for photopic vision⁵⁴ and φ_vis,sol is the solar irradiance spectrum for air mass 1.5, which corresponds to the sun standing 37° above the horizon.⁵⁵

In general, solar modulation efficiency increases at the expense of a reduction in luminous transmittance. However, as shown in Table 2, the samples with smaller particle sizes exhibited improved integrated luminous transmittances (T_lum) of 9.6%, 25.3%, 34.8% and 38.5%, in combination with increased solar modulation efficiencies of as high as 8.1%, 12.3%, 15.6%, and 16.9%, respectively. Our optimized VO₂ composite films prepared by milling for 12 h exhibited high luminous transmittance (T_lum = 38.5%) and superior solar modulation efficiency (ΔT_sol = 16.9%). This result is interesting because the value of ΔT_sol (16.9%) is much higher than the values of 4.4% for a single-layered VO₂ film⁵⁶ and 12.1% for a five-layered film containing TiO₂/VO₂/TiO₂/VO₂/TiO₂ prepared by a sputtering method,⁵⁷ as well as 13.6% for a composite film produced by casting a suspension of VO₂@SiO₂ nanoparticles.¹⁹ Furthermore, it is comparable to the theoretical value (ΔT_sol ≈ 16.7%), which was calculated by Li et al.⁵⁸ for a model of randomly oriented spheroids dispersed in a dielectric medium. The size effect of VO₂ particles may contribute to the improvements in T_lum and ΔT_sol.

Table 2 Integrated luminous/solar transmittance, visible solar transmittance and total solar modulation efficiency of as-prepared VO₂ composite films with different milling times: I (1 h), II (4 h), III (8 h) and IV (12 h)

Sample	Tlum (%)	Tvis (%)		ΔTvis (%)	Tsol (%)		ΔTsol (%)
	25 °C	25 °C	100 °C		25 °C	100 °C
I	9.6	11.4	16	−4.6	22.9	14.8	8.1
II	25.3	25	26	−1	35.9	23.6	12.3
III	34.8	32.3	28.6	3.7	40.8	25.2	15.6
IV	38.5	34.8	28.8	6	42.1	25.2	16.9

---

Tuning the phase transition temperature and hysteresis loops by scaling VO₂ particles to nanoscale dimensions has drawn extensive attention.^59,60 This illustrates that size effects are not only present in the optical properties but also in the features of the phase transition itself. Fig. 4b, d, f and h show the hysteresis loops obtained using various milling times of 1 h, 4 h, 8 h and 12 h, respectively. The insets show the corresponding differentials of the variable-temperature processes. It is observed that the films prepared using longer milling times with a smaller particle size exhibit higher transition temperatures upon heating and lower temperatures upon cooling (Table 3 and Fig. 5a), which thus leads to a progressive widening of the hysteresis loops. Indeed, the origins of the transition temperature and hysteresis are closely related to the density of defects in films, which can serve as nucleation sites for phase transformation. The decreased volume of VO₂ nanoparticles with a lower number of effective defects reduces the probability of triggering the structural phase transition at a certain temperature.⁶¹ Therefore, the width of the hysteresis loop gradually increases for the films containing smaller VO₂ nanoparticles, as a result of a requirement for overheating or overcooling to overcome the higher barrier to nucleation. As observed in Fig. 4b, d, f and h, the hysteresis loops between the heating and cooling processes became more symmetrical with an increase in the milling time. According to an investigation by Petit et al., disoriented crystallites in films led to wider and asymmetrical hysteresis loops, whereas a (011) texture was beneficial for narrow and symmetrical hysteresis loops.⁶² In our experiment, the films prepared by milling for 1 h displayed ragged and poorly orientated particles, which generated low-symmetry hysteresis loops. In contrast, the homogeneous particles in films that were produced by an adequate period of milling gave rise to more symmetrical hysteresis loops. Actually, the appearance of two peaks in the inset of Fig. 4d is also attributed to polydispersity in the size of nanoparticles in the film, as shown in Fig. 3b.³⁷

Table 3 Optical transmittances at 660 nm and phase transition temperatures of samples with different milling times: I (1 h) and IV (12 h)

Sample	T (%) at 660 nm		ΔT (%)	Tc,h (°C)	Tc,c (°C)	Tc (°C)	ΔT (°C)
	30 °C	100 °C		Heating	Cooling	Critical
I	16.3	23.8	−7.5	66	62	64	4
IV	45.2	40.1	5.1	82	55	68.5	27

---

Actually, there is slight switching of optical transmittance in the visible-light region through the MIT, as well as for our samples. According to the solar spectral irradiance (gray areas in Fig. 4), radiant heat in the visible region (380–780 nm) has a high power intensity, whereas this is lower in the NIR region.⁴² Hence, slight switching in the visible region can give rise to major alterations in solar throughput upon the MIT. With an increase in the milling time, the greatly improved change in ΔT_vis (from −4.6% to 6%) was primarily responsible for the increase in ΔT_sol (from 8.1% to 16.9%), which was summarized in Table 2. In a previous work, Xu et al. demonstrated that the changes in luminous transmittance upon optical switching were thickness-dependent, which resulted from interference effects caused by the difference in refractive indices between the M₁ and R phases of VO₂ in the visible region.⁴³ Accordingly, some researchers modified the luminous transmittance upon the MIT by varying the thickness of VO₂ film.^12,36,43 However, in our experiment, the thicknesses of films obtained using various milling times are comparable (Table 1). Here, we first observed that the alterations in the visible region upon the MIT are size-dependent. As shown in Table 2, for films with large particle sizes (prepared by milling for 1 h and 4 h) the visible transmittance at 25 °C (T_vis = 11.4% and 25%, respectively) is lower than that at 100 °C (T_vis = 16% and 26%, respectively), and the reverse will occur for films with small particle dimensions. This optical switching is very distinct for the samples obtained by milling for 1 h and 12 h, of which the optical transmittance spectra (from 250 to 2500 nm) recorded from 25 °C to 100 °C are displayed in Fig. 6a and b, respectively. As expected, both samples exhibited decreased optical transmittance in the NIR at 100 °C, where VO₂ films are in the metallic phase. Extraordinarily, as the temperature increased, the transmittance in the visible region increased steadily for the sample with large particle sizes (Fig. 6a), which is in contrast with a decrease in visible transmittance for the sample consisting of small particles (Fig. 6b). In reality, the scattering states of VO₂ between the M₁ and R phases are diverse in the visible region. Lopez et al. confirmed that VO₂ nanoparticles in the metallic phase displayed a lower scattering efficiency than those in the insulating phase, in particular in the broad blue-green peak of the visible region.⁶¹ We performed Mie calculations⁶³ based on the optical constants of the M₁ and R phases of VO₂.⁴³ Fig. 7b shows various scattering coefficients calculated for spherical VO₂ particles with different diameters of 150 nm, 100 nm, 60 nm and 40 nm dispersed in a transparent polymer, where the solid lines and dotted lines represent the M₁ and R phases, respectively. It is obvious that the differences in scattering between the M₁ and R phases of VO₂ are diverse for the various particle sizes in the visible region. For VO₂ particles with a diameter of 150 nm, the scattering state of the M₁ phase is stronger than that of the R phase. Nevertheless, the differences between the M₁ and R phases decrease as the particle sizes become smaller. In particular, for particles with a diameter of 40 nm, nearly identical scattering coefficients appear in all wavelength ranges. Accordingly, for VO₂ films with a mean particle size of 150 nm (achieved by milling for 1 h), the scattering gradually decreased because the M₁ phase was gradually transformed into the R phase as the temperature rose, which resulted in an improvement in luminous transmittance. It has been demonstrated that the visible transmittance at low temperature for films thicker than 100 nm exceeds that at high temperature.^12,36,43 Hence, the transmittance of films obtained by milling for 12 h normally decreased in the visible region upon the MIT owing to the slight change in scattering for VO₂ particles with a diameter of 40 nm.

Fig. 6 Optical transmittance spectra from 250 nm to 2500 nm recorded from 25 °C to 100 °C for VO₂ composite films obtained by milling for 1 h (a) and 12 h (b); the corresponding thermal hysteresis loops at 660 nm (c and d).

Fig. 7 Haze values for samples with various milling times at different wavelengths (a) and calculated scattering coefficients for various mean particle sizes (b).

We first presented the hysteresis loops for a wavelength of 660 nm. Fig. 6c and d show that the transmittance monotonically increased or decreased upon heating or cooling, respectively, from which we can deduce that the scattering at 660 nm also changed monotonically. For our samples, we find no evidence for a three-state hysteresis loop for scattering, which was discovered with strictly periodic arrays of VO₂ nanoparticles.⁶¹ Distinctive size effects upon the MIT could also be perceived from the hysteresis loops at 660 nm. In other words, the reduction in particle size in the VO₂ film brought about higher transition temperatures on heating and lower temperatures on cooling, which led to increasingly wider hysteresis loops (Fig. 6c and d and Table 3).

4. Discussion

In order to provide an explanation of luminous transmittance, the absorption spectra of VO₂ films were investigated from 250 to 2500 nm at room temperature, as shown in Fig. 8a. It was found that the changes in absorption ability were correlated with the particle size in the films.³⁷ For instance, the sample prepared by milling for 1 h, of which the particles were the largest among these samples, exhibited the strongest absorption. In contrast, a reduction in absorption appeared for samples with particles of lesser dimensions. In particular in the visible region (as observed from the inset of Fig. 8a), such a phenomenon of size-dependent absorption became more evident, which largely determined the luminous transmittance of the semiconducting VO₂ films. It is known that optical absorption is closely related to the inherent band gap.^39–41 The absorption coefficient α can be derived using the Beer–Lambert law:

α = (1/d)ln[(1 − R)/T] (4)

Fig. 8 Absorption spectra of samples obtained using various milling times at different wavelengths (a); the relationships between (αhν)^1/2 and hν for VO₂ films prepared using various milling times (the inset shows the intrinsic or interband gap energy) (b); and schematic of band structure of semiconducting VO₂ (c).

Optical band gaps (denoted by E_g) are determined by the following equation:

(αhν)^m = A(hν − E_g) (5)

where A is a constant and the exponent m, depending on the nature of the optical transition, is 1/2, 1/3, 2 and 2/3 for indirect-allowed, indirect-forbidden, direct-allowed, and direct-forbidden optical transitions, respectively.³⁹ Linear extrapolation of a plot of (αhν)^m vs. hν near the band gap gives E_g as the intercept with the axis at α = 0. Assuming that the transitions are indirect-allowed, Fig. 8b shows the relationship between (αhν)² and hν for semiconducting VO₂ films with varying milling times of 1 h, 4 h, 8 h and 12 h. As shown in Fig. 8a, we can identify two primary absorption edges in the regions marked A and B, which correspond to two band gaps at around 2.5 eV and 0.67 eV. According to Goodenough's investigation⁶⁴ into monoclinic VO₂ formed below the critical temperature, the d_∥ band splits into two bands and the π* band rises above the Fermi level. The π* band is located between the lower occupied d_∥ band and the upper unoccupied d_∥ band. The intrinsic band gap energy (E_g) is determined from the gap between the lower occupied d_∥ band and the π* band (Fig. 8c). Values in the vicinity of 0.67 eV were assigned to the intrinsic band gap.^65,66 It is well known that the low luminous transmittance of VO₂ mainly originates from intense absorption in the visible region (marked A in Fig. 8a). The electrons in VO₂ can strongly absorb photons with the appropriate energy. Gavini et al.⁶⁶ regarded the threshold energy of approximately 2.5 eV as a specific interband gap (E₀) associated with the transition between the O 2p π and π* levels (Fig. 8c). Hence, the luminous transmittance of VO₂ film is dominated by the interband gap.^39–41 Extraordinary, distinctive size effects on both interband gaps can be observed from our samples, as shown in the inset of Fig. 8b. In other words, the smaller size of VO₂ particles in films prepared using a longer milling time brings about a widening of the interband gap from 2.33 to 2.88 eV, leading to a blue shift as well as a steepening of the absorption edge, which resulted in a significant decrease in luminous absorbance; in other words, an enhancement in luminous transmittance.^37,67 Therefore, it is possible that the alteration in the interband gap is a consequence of size effects.

Besides high luminous transmittance, high transparency is also a requisite for the practical application of VO₂ films in smart windows. The integration of VO₂ nanoparticles into polymers will bring about light scattering resulting in a severe deterioration in visual transparency, namely, the generation of haze.^31,32 As mentioned previously, there are two main reasons for the appearance of haze caused by scattering, which include the mismatch in RI and the large size of nanoparticles in composite films. It has been shown that the mismatch in RI can be compensated for by decreasing the particle size to significantly below the wavelength of visible light.^33,34 Also, theoretical studies have indicated that in the case of thinner films the loss of light intensity induced by scattering can be negligible if the particle size is reduced to below 100 nm.^33,34

A standard metric for quantifying a film's light scattering ability is the haze factor:

H(haze) = T_d/T_t × 100% = (T_t − T_s)/T_t × 100% (6)

where T_d is the diffuse light transmittance, T_s is the specular transmittance and T_t is the total transmittance, which is the sum of T_d and T_s.⁶⁸ Fig. 7a shows the haze values for various VO₂ composite films measured over the wavelength range between 380 nm and 2500 nm. As the milling time increased from 1 h to 12 h, the haze of the films decreased, in particular in the visible-light region, whereas the haze in the NIR was quite slight. This is because scattering by small particles generates unique resonances owing to electronic confinement.⁶¹ As listed in Table 1, the haze at 555 nm, where the human eye is most sensitive,⁵⁴ was reduced from 46.5% to 2.2%, which is suitable for practical applications in smart windows. Acting as effective scattering sites, the mean particle sizes in VO₂ films decreased from 150 nm to 40 nm with an increase in the milling time (Table 1). In the visible-light region, theoretical scattering coefficients, which were determined by Mie calculations, are shown in Fig. 7b. It can be seen that the scattering coefficient was significantly decreased with a reduction in the particle size from 150 nm to 40 nm. Therefore, the appearance of a reduction in haze can be explained, because the haze is primarily derived from light scattering. As can be seen, the calculated results are qualitatively consistent with our experimental observations.

5. Conclusions

In this study, we report a convenient and controllable method for the fabrication of VO₂-based composite films, in which the sizes of VO₂ particles can easily be tuned by optimizing the parameters of ball milling. The haze values at 555 nm of composite films were reduced from 46.5% to 2.2% as the mean size of VO₂ particles decreased from 150 nm to 40 nm, which is responsible for the improvement in luminous transmittance (from 9.6% to 38.5% @RT). Moreover, the behavior of optical switching in the visible region also displays a distinct size effect. The VO₂-based composite film exhibited reduced haze and improved luminous transmittance in combination with superior solar modulation ability (ΔT_sol up from 8.1% to 16.9% @RT), which is suitable for practical applications in smart windows.

Acknowledgements

This work was supported by the Science and Technology Planning Project of Guangdong Province, China (Grant No. 2013B050800006), the External Cooperation Program of BIC, Chinese Academy of Sciences (Grant No. 182344KYSB20130006), and the National Natural Science Foundation of China (Grant No. 51572049, 51562005) Guangxi Natural Science Foundation of China (Grant No. 2015GXNSFFA139002).

References

1. F. J. Morin, Phys. Rev. Lett., 1959, 3, 34.
2. M. M. Qazilbash, M. Brehm, B. G. Chae, P. C. Ho, G. O. Andreev, B. J. Kim, S. J. Yun, A. V. Balatsky, M. B. Maple, F. Keilmann, H. T. Kim and D. N. Basov, Science, 2007, 318, 1750.
3. L. Whittaker, C. J. Patridge and S. Banerjee, J. Phys. Chem. Lett., 2011, 2, 745.
4. W. T. Liu, J. Cao, W. Fan, Z. Hao, M. C. Martin, Y. R. Shen, J. Wu and F. Wang, Nano Lett., 2011, 11, 466.
5. L. S. Long, H. Ye, Y. F. Gao and R. Q. Zou, Appl. Energy, 2014, 136, 89.
6. Y. F. Gao, H. J. Luo, Z. T. Zhang, L. T. Kang, Z. Chen, J. Du, M. Kanehira and C. X. Cao, Nano Energy, 2012, 1, 221.
7. C. Z. Wu, F. Feng and Y. Xie, Chem. Soc. Rev., 2013, 42, 5157.
8. W. W. Li, J. J. Zhu, J. R. Liang, Z. G. Hu, J. Liu, H. D. Chen and J. H. Chu, J. Phys. Chem. C, 2011, 115, 23558.
9. M. A. Sobhan, R. T. Kivaisi, B. Stjerna and C. G. Granqvist, Sol. Cells, 1996, 44, 451.
10. D. Vernardou, M. E. Pemble and D. W. Sheel, Chem. Vap. Deposition, 2007, 13, 158.
11. R. Lopez, L. A. Boatner, T. E. Haynes, R. F. Haglund and L. C. Feldman, Appl. Phys. Lett., 2001, 79, 3161.
12. Z. T. Zhang, Y. F. Gao, Z. Chen, J. Du, C. X. Cao, L. T. Kang and H. J. Luo, Langmuir, 2010, 26, 10738.
13. X. G. Tan, T. Yao, R. Long, Z. H. Sun, Y. J. Feng, H. Cheng, X. Yuan, W. Q. Zhang, Q. H. Liu, C. Z. Wu, Y. Xie and S. Q. Wei, Sci. Rep., 2012, 2, 466.
14. C. X. Cao, Y. F. Gao and H. J. Luo, J. Phys. Chem. C, 2008, 112, 18810.
15. L. Whittaker, C. Jaye, Z. Fu, D. A. Fischer and S. Banerjee, J. Am. Chem. Soc., 2009, 131, 8884.
16. H. Kim, Y. Kim, K. Soo Kim, H. Young Jeong, A.-R. Jang, S. Ho Han, D. Ho Yoon, K. S. Suh, H. S. Shin, T. Y. Kim and W. S. Yang, ACS Nano, 2013, 7, 5769.
17. D. Li, M. Li, J. Pan, Y. Luo, H. Wu, Y. Zhang and G. Li, ACS Appl. Mater. Interfaces, 2014, 6, 6555.
18. H. Zhang, X. Xiao, X. Lu, G. Chai, Y. Sun, Y. Zhan and G. Xu, J. Alloys Compd., 2015, 636, 106.
19. X. Xiao, H. Zhang, G. Chai, Y. Sun, T. Yang, H. Cheng, L. Chen, L. Miao and G. Xu, Mater. Res. Bull., 2014, 51, 6.
20. G. Thirumala Rao, B. Babu, R. Joyce Stella, V. Pushpa Manjari, C. V. Reddy, J. Shim and R. V. S. S. N. Ravikumar, J. Mol. Struct., 2015, 108, 254.
21. D. Li, M. Li, J. Pan, Y. Luo, H. Wu, Y. Zhang and G. Li, ACS Appl. Mater. Interfaces, 2014, 6, 6555.
22. Z. Chen, C. Cao, S. Chen, H. Luo and Y. Gao, J. Mater. Chem. A, 2014, 2, 11874.
23. W. Li, S. Ji, K. Qian and P. Jin, J. Colloid Interface Sci., 2015, 456, 166.
24. I. G. Madida, A. Simo, B. Sone, A. Maity, J. B. Kana Kana, A. Gibaud, G. Merad, F. T. Thema and M. Maaza, Sol. Energy, 2014, 107, 758.
25. I. G. Madiba, A. Simo, B. Sone, L. Kotsedi and M. Maaza, Mater. Today, 2015, 2, 4006.
26. J. Puetz, N. Al-Dahoudi and M. A. Aegerter, Adv. Eng. Mater., 2004, 6, 733.
27. H. Althues, J. Henle and S. Kaskel, Chem. Soc. Rev., 2007, 36, 1454.
28. F. Favoino, M. Overend and Q. Jin, Appl. Energy, 2015, 156, 1.
29. Y. F. Gao, S. B. Wang, H. J. Luo, L. Dai, C. X. Cao, Y. L. Liu, Z. Chen and M. Kanehira, Energy Environ. Sci., 2012, 5, 6104.
30. J. C. Valmalette and J. R. Gavarri, Sol. Energy Mater. Sol. Cells, 1994, 33, 135.
31. D. Domine, F. J. Haug, C. Battaglia and C. Ballif, J. Appl. Phys., 2010, 107, 044504.
32. S. Mehra, M. G. Christoforo, P. Peumans and A. Salleo, Nanoscale, 2013, 5, 4400.
33. Y. Q. Li, S. Y. Fu, Y. Yang and Y. W. Mai, Chem. Mater., 2008, 20, 2637.
34. K. Laaksonen, S. Y. Li, S. R. Puisto, N. K. J. Rostedt, T. A. Nissila and C. G. Granqvist, Sol. Energy Mater. Sol. CellsSolar Cells, 2014, 130, 132.
35. S. Y. Li, G. A. Niklasson and C. G. Granqvist, J. Appl. Phys., 2010, 108, 063525.
36. L. T. Kang, Y. F. Gao, H. J. Luo, Z. Chen, J. Du and Z. T. Zhang, ACS Appl. Mater. Interfaces, 2011, 3, 135.
37. L. T. Kang, Y. F. Gao, Z. T. Zhang, J. Du, C. X. Cao, Z. Chen and H. J. Luo, J. Phys. Chem. C, 2010, 114, 1901.
38. M. Zhou, J. Bao, M. S. Tao, R. Zhu, Y. T. Lin, X. D. Zhang and Y. Xie, Chem. Commun., 2013, 49, 6021.
39. S. L. Hu, S. Y. Li, R. Ahuja, C. G. Granqvist, K. Hermansson, G. A. Niklasson and R. H. Scheicher, Appl. Phys. Lett., 2012, 101, 201902.
40. L. Dai, S. Chen, J. J. Liu, Y. F. Gao, J. D. Zhou, Z. Chen, C. X. Cao, H. J. Luo and M. Kanehira, Phys. Chem. Chem. Phys., 2013, 15, 11723.
41. J. D. Zhou, Y. F. Gao, Y. L. Liu, Z. Chen, L. Dai, C. X. Cao, H. J. Luo, M. Kanahira, C. Sun and L. M. Yan, Phys. Chem. Chem. Phys., 2013, 15, 7505.
42. C. G. Granqvist, Adv. Mater., 2003, 15, 1789.
43. G. Xu, P. Jin, M. Tazawa and K. Yoshimura, Jpn. J. Appl. Phys., 2004, 43, 186.
44. L. Whittaker, T. L. Wu, C. J. Patridge, S. Ganapathy and S. Banerjee, J. Mater. Chem., 2011, 21, 5580.
45. C. J. Patridge, L. Whittaker, B. Ravel and S. Banerjee, J. Phys. Chem. C, 2012, 116, 3728.
46. Z. Gui, R. Fan, W. Q. Mo, X. H. Chen, L. Yang, S. Y. Zhang, Y. Hu, Z. Z. Wang and W. C. Fan, Chem. Mater., 2002, 14, 5053.
47. X. Chen, X. M. Sun and Y. D. Li, Inorg. Chem., 2002, 41, 4524.
48. L. Q. Mai, Y. H. Gu, C. H. Han, P. C. Zhang, L. Xu, W. L. Guo and Y. Dai, Nano Lett., 2009, 9, 826.
49. C. Tsang and A. Manthiram, J. Electrochem. Soc., 1997, 144, 520.
50. C. Leroux and G. Nihoul, Phys. Rev. B: Condens. Matter Mater. Phys., 1998, 57, 5111.
51. S. A. Corr, M. Grossman, Y. F. Shi, K. R. Heier, G. D. Stucky and R. Seshadri, J. Mater. Chem., 2009, 19, 4362.
52. X. D. Xiao, H. L. Cheng, G. P. Dong, Y. G. Yu, L. H. Chen, L. Miao and G. Xu, CrystEngComm, 2013, 15, 1095.
53. J. Q. Shi, S. X. Zhou, B. You and L. M. Wu, Sol. Energy Mater. Sol. Cells, 2007, 91, 1856.
54. G. Wyszecki and W. S. Stiles, Quantitative Data and Formulae, Wiley, New York, 2nd edn, 2000.
55. ASTM G173-03 Standard Tables of Reference Solar Spectral Irradiances: Direct Normal and Hemispherical on a 37 Tilted Surface; Annual Book of ASTM Standards, American Society for Testing and Materials: Philadelphia, PA, USA, 2008 http://rredc.nrel.gov/solar/spectra/am1.5.
56. P. Jin, G. Xu, M. Tazawa and K. Yoshimura, Jpn. J. Appl. Phys., Part 2, 2002, 41, L278.
57. N. R. Mlyuka, G. A. Niklasson and C. G. Granqvist, Sol. Energy Mater. Sol. Cells, 2009, 93, 1685.
58. S. Y. Li, G. A. Niklasson and C. G. Granqvist, J. Appl. Phys., 2011, 109, 113515.
59. Z. D. Lu, C. G. Li and Y. D. Yin, J. Mater. Chem., 2011, 21, 14776.
60. Y. F. Sun, S. S. Jiang, W. T. Bi, R. Long, X. G. Tan, C. Z. Wu, S. Q. Wei and Y. Xie, Nanoscale, 2011, 3, 4394.
61. R. Lopez, L. C. Feldman and R. F. Haglund, Phys. Rev. Lett., 2004, 93, 177403.
62. C. Petit, J. M. Frigerio and M. Goldmann, J. Phys.: Condens. Matter, 1999, 11, 3259.
63. E. Huber and M. Frost, J. Water Supply: Res. Technol.--AQUA, 1998, 47, 87.
64. J. B. Goodenough, J. Solid State Chem., 1971, 3, 490.
65. S. Shin, Phys. Rev. B: Condens. Matter Mater. Phys., 1990, 41, 4993.
66. A. Gavini and C. C. Y. Kwan, Phys. Rev. B: Solid State, 1972, 5, 3138.
67. H. K. Kim, H. You, R. P. Chiarello, H. L. M. Chang, T. J. Zhang and D. J. Lam, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 12900.
68. ASTM, D1003–11 Standard Test Method for Haze and Luminous Transmittance of Transparent Plastics.

---