Preparation of the floating HGMs-Ag3PO4 composites and the visible-light photocatalytic properties

The hollow glass microspheres (HGMs) with low density were used as the support to obtain the floating visible-light photocatalysts (HGMs–Ag3PO4) composites, in which silver orthophosphate (Ag3PO4) was dispersed on the surface of HGMs via silane coupling agents. The morphology, crystal structure and optical property of the HGMs-Ag3PO4 composites were﻿ characterized and the contents of Ag3PO4 dispersed on the surface of HGMs were quantified by atomic absorption spectrometry (AAS). Methylene blue (MB) was selected as the typical organic pollutants to evaluate the photocatalytic activity of HGMs-Ag3PO4 composite photocatalysts. The photo-catalytic rate of the floating composites was larger than the sinking composites and even pure Ag3PO4. Ag3PO4 particles could be dispersed well on the HGMs. The floating composites has a larger light contact area and the increased photo-catalytic activity sites even at the low Ag3PO4 loading. This work may open up a new insight for the floating photocatalyst to degrade organic dye on account of the low density and high visible-light responsiveness.


Introduction
In recent years, the experts and scholars have paid a wide range attention on environmental and energy issues such as non-compliance emissions of various textile industry dyestuffs, excessive consumption of traditional fossil fuels, and so on, which have threatened human health and ecosystem stability [1]. Because semiconductor photocatalysts can utilize solar energy to oxidize organic pollutants into carbon dioxide, water and other harmless substances [2], they are applied on environmental pollution remediation and solar energy conversion in an environmental friendly method.
As one of the most effective photo-catalysts, TiO 2 composites [3] have been widely used in the photodegradation of organic dye pollutants due to its low cost, stability, anti-corrosion non-toxic and energy saving. However, TiO 2 exhibits low quantum efficiency and low utilization of solar energy, for TiO 2 can only absorb the UV light (λ<380 nm) making up for only 4% of the sunlight and has a relatively wide band gap of 3.2 eV [4][5][6]. Hence, it is very necessary to have a research on efficient visible-light photo-catalysts with relatively narrow band gap. Yi [7] reported Ag 3 PO 4 had excellent performance on the evolution of O 2 from water splitting and remove of organic dyes under the visible-light exposure due to the capability to effectively separate photoexcited electrons and holes with suitable band gap of 2.45 eV, achieving a quantum yield of up to 90% (λ>420 nm), thus extending a new application of Ag 3 PO 4 in the field of photo-catalysis. The visible-light photo-catalytic degradation properties of Ag 3 PO 4 can be further enhanced by controlling its morphology [8], feature size [9], surface area [10] and low-index facets [11], by assembling with other semiconductors such as TiO 2 [6,12,13] , Fe 3 O 4 [14], AgX (X=Cl, Br, I) [15][16][17], In(OH) 3 [18], BiVO 4 [19], MOF [20], MoSe 2 [21], CeO 2 [22], N-Sr 2 Nb 2 O 7 [23], g-C 3 N 4 [24], polyaniline [25] and carbon based materials including graphene (GR) [26], graphene oxide (GO) [27], reduced graphene oxide (RGO) [28], carbon quantum dots (CQDs) [29]and carbon nanotubes (CNTs) [30] to construct a surface heterojunction, and the doping of atomic impurities [31]. The photo-catalytic activity performance of Ag 3 PO 4 -composites has been summarized into table 1. There are many typical organic pollutants to evaluate the photocatalytic activity of Ag 3 PO 4 -composites photocatalysts, such as using organic dyes (methylene blue-MB [8,13,24,30], rhodamine B-RhB [12,18,19], methyl orange-MO [22,29], etc), preservatives (methylparaben [16], phenol [20,25,28], etc), insecticides (imidacloprid [14], etc), antibiotic (tetracycline [26], etc). Compared with pure Ag 3 PO 4 photocatalyst, the Ag 3 PO 4 -composites have a certain degree of improvement in the photocatalytic activity on different organic pollutants. Nevertheless, in order to achieve high photocatalytic performance, Ag 3 PO 4 needs to account for a large proportion even in the Ag 3 PO 4 composite photocatalyst [20,24,25,30]. The high cost of raw materials AgNO 3 and the difficulty of separating as well as recovering Ag 3 PO 4 -based photo-catalysts from the reaction suspension limit the large-scale practical application of the photo-catalysts [32]. Therefore, the fabrication of high cost-performance Ag 3 PO 4 -based composites reducing the dosage of AgNO 3 and being recovered easily is of great significance.
Hollow glass microspheres (HGMs) are the hollow powdery borosilicate with a particle size of 10-100 μm with wall thickness of 1-2 μm. HGMs can be separated from the coal fly ash to make use of waste resources [33]. Due to the unique properties of low density, large surface areas, light-trapping effects, low thermal conductivity, high strength, non-toxic, non-combustible, thermal and sound insulating, high dispersion and good chemical stability, HGMs are widely applied in the synthesis of microwave absorbing materials [34], hydrogen storage materials [35], low density foams [36], deep sea buoyancy materials [37], catalyst carriers [13,38], biomaterial defect fillers [39] and so on.
The preparation of new photo-catalytic Ag 3 PO 4 -based composites using HGMs as the photo-catalyst carriers can compensate the limitations of Ag 3 PO 4 . On the one hand, Ag 3 PO 4 dispersed on the floating micronsized HGMs not only can be easily separated from the treatment solution, but also can improve the light contact chance and the photo-catalytic activity due to its floating property at the surface of the effluent. On the other hand, the surface of the HGMs loaded with a small amount of Ag 3 PO 4 , significantly reducing the dosage of raw materials AgNO 3 , can achieve a certain photo-catalytic performance because only the outer layer of photocatalyst could absorb visible-light in the practical application.
In this work, the floating visible-light-responsive HGMs-Ag 3 PO 4 composites were prepared by a facile ion exchange method using HGMs as a floating photo-catalyst carrier with a silane coupling agent as a bridge. The visible-light catalytic activities of the floating HGMs-Ag 3 PO 4 composites was investigated by the MB photodegradation efficiency under visible-light irradiation in contrast with pure Ag 3 PO 4 and the sinking HGMs-Ag 3 PO 4 composites. Furthermore, the possible photo-degradation mechanism of floating HGMs-Ag 3 PO 4 composites was proposed.

composites
The surface pretreatment of HGMs: 0.1 μl of silane coupling agent KH550 was dissolved in 12.5 ml of distilled water in a beaker, then 2.5 g HGMs were dispersed to the above solution under vigorous stirring. After magnetic stirring for 30 min, the mixture was vacuum filtered, then washed several times with deionized water and absolute ethanol. The floating products were dried at 60°C for 8 h. The pretreatment products were marked as HGMs-KH550. HGMs-Ag 3 PO 4 powders were prepared by the simple ion-exchange method. 2.0 g HGMs-KH550 was added to 100 ml of AgNO 3 aqueous solution with magnetic stirring at 280 r min −1 for 1.5 h. 100 ml Na 2 HPO 4 aqueous solution was added dropwise to the mixture slowly with magnetic stirring for 1 h. After the completion of the dropwise addition, the mixture was steadily stirred for 0.5 h and washed with distilled water to move any unreacted raw material. Last, the final product was vacuum filtered and dried at 60°C for 8 h. The HGMs-Ag 3 PO 4 composites with different Ag 3 PO 4 loadings were synthesized with adjusting the concentrations of AgNO 3 and Na 2 HPO 4 solutions. The AgNO 3 and Na 2 HPO 4 solutions with the same concentrations (0.0215, 0.0430, 0.0645 and 0.0860 mmol l −1 ) were used to fabricate the HGMs-Ag 3 PO 4 composites with different Ag 3 PO 4 theoretical loadings (0.045, 0.090, 0.135 and 0.180 g g −1 ), which were marked as HGMs1, HGMs2, HGMs3 and HGMs4, respectively. HGMs1, HGMs2 and HGMs3 floated in water, while HGMs4 deposited in water. Pure Ag 3 PO 4 was prepared in the same method without adding HGMs-KH550. HGMs0 was prepared as the same method as HGMs3 without amino modification.

Characterization
The morphology of the as-prepared photocatalysts were characterized by using a field emission scanning electron microscope (SEM, S-4800). The crystal phases of HGMs-Ag 3 PO 4 composites were determined by x-ray diffraction (XRD, D/MAX-2500) with CuKα radiation (λ=1.54 Å) in the range of 15-75°(2θ) with a scanning rate of 4°min −1 . The valence and content of silver were determined by X-ray photoelectron spectroscopy (XPS, ESCALAB 250 with monochromatized Mg KR x-ray as the source). The resulting binding energies were calibrated to the C 1 s (284.6 eV) peak. The specific surface area of HGMs, HGMs3, and Ag 3 PO 4 were measured with an Autosorb-1 specific surface area analyzer (Quantachrome Instruments, USA) using the BET method on the base of N 2 adsorption data. FT-IR spectra were measured on an Excalibur 3100 infrared spectrophotometer from KBr pellets as the sample matrix. The ultraviolet-visible diffuse reflectance spectra (Uv-vis DRS) were recorded with a UV-2450 spectrophotometer using BaSO 4 as the reference of diffuse reflection substance and converted to absorption spectra by the standard Kubelka-Munk method. UV-Vis absorption spectra was measured at room temperature on a UV-1800 (Mapada) spectrometer. When Ag 3 PO 4 on HGMs-Ag 3 PO 4 composites was dissolved with 1:499 (v/v) dilute HNO 3 solution, the Ag 3 PO 4 loading percentage of as-prepared composites was estimated by measuring Ag + contents with Atomic absorption spectroscopy (AAS, Thermo Scientific iCE3000).

Evaluation of photocatalytic properties
The visual-light catalytic activity of HGMs-Ag 3 PO 4 composites was evaluated by the photodegradation of MB organic dye pollutants under the simulated sunlight irradiation. In all catalytic activity of experiments, the asobtained samples (0.3 g) were put into MB solution (60 ml, 5 mg l −1 ), which was placed for 60 min in the dark to achieve adsorption-desorption equilibrium before the lamp turned on and then irradiated vertically by a 300W Xe arc lamp equipped with an ultraviolet cutoff filter (λ420 nm) to provide visible light without magnetically stirring to imitate the photodegradation of organic effluent at the natural environmental conditions. In the photo-catalytic testing, the liquid level of MB dye solution was 95 mm from the Xe arc lamp. With a 5-min interval, 3 ml suspension of MB was sampled, then filtered by a 0.22 μm Millipore filter to remove the residual catalysts. The concentration of MB was analyzed by UV-1800 at the wavelength of 663 nm.
To assess the stability of the HGMs-Ag 3 PO 4 composites, cycling runs of MB photodegradation were conducted. The floating composites were scooped out and dried at 60°C for 4h at the end of each circle. The photo-degradation rate of MB dye is fit from the line ln(C/C 0 )∼t. C is the concentration of MB at time t and C 0 is the initial MB concentration. On the basis of Beer-Lambert Law, C/C 0 is the ratio of the MB absorbance at time t to the initial absorbance. The concentrations of MB at time t were calculated by testing the maximum absorption intensity of the UV-visible spectrum at 663 nm.

Characterization of the composites
In figure 1, the preparation of HGMs-Ag 3 PO 4 powders were depicted. HGMs were modified by the silane coupling agent KH550 to introduce -NH 2 groups on HGMs. The valence electron orbit of the Ag + ion is 4d 10 5s 0 5p 0 , the sp-hybrid orbit of which can accept the lone pair electrons of -NH 2 from the silane coupling agent to form a complex. As a bridge, the silane coupling agent KH550 connects HGM surface and Ag + . PO 4 3− from the hydrolysis of HPO 4 2− reacts with Ag + to generate Ag 3 PO 4 on the surface of HGMs-KH550. As the reaction proceeded, the color of the floating HGMs-KH550 gradually varied from the initial white to yellow, illustrating the formation of Ag 3 PO 4 on the surface of HGMs-KH550.
The actual Ag 3 PO 4 loading on HGMs0, HGMs1, HGMs2, HGMs3 and HGMs4, were measured by AAS, which were listed in table 2. The only difference between HGMs0 and HGMs3 in the preparation process is that HGMs0 is not modified by amination. From table 2, we can see that only 0.4 wt% of the silane coupling agent can increase the catalyst loading rate by 52%. The results illustrated that trace KH550 had a significant effect on the loading of Ag 3 PO 4 and played a vital role as a bridge between Ag 3 PO 4 and HGMs. The measured Ag 3 PO 4 loading increased in the order of HGMs1, HGMs2, HGMs3 and HGMs4 with the increasing of the reactants. However, the actual Ag 3 PO 4 percent conversions decreased by HGMs1, HGMs2 and HGMs3, because the more Ag + could connect with NH 2 groups on HGMs and the HGMs-KH550 had limited NH 2 -active sites to absorb Ag + , resulting in the direct deposition of partial Ag 3 PO 4 at the bottom of the reaction vessel, rather than the attachments on HGMs. In addition, the actual Ag 3 PO 4 loading percentage of HGMs3 had dropped to 56.9%. If the more input of AgNO 3 and Na 2 HPO 4 , the loading percentage of Ag 3 PO 4 on floating HGMs-Ag 3 PO 4 composites will drop even lower and will generate more Ag 3 PO 4 deposition rather than loaded on HGMs. But for the deposited HGMs4, which is a sinking-type material including unloaded pure Ag 3 PO 4 and Ag 3 PO 4 loaded by hollow glass microspheres, the loading percentage of Ag 3 PO 4 increased with the increasing of the reactants, because Ag 3 PO 4 is already saturated with loads on HGMs-KH550 so that HGMs4 sinks and unloaded Ag 3 PO 4 sinks. This was very different from the floating HGMs1, 2 and 3.     3(b)-(d)). In addition, HGMs3 were treated for MB photo-degradation under the visible-light irradiation for an hour. This used HGMs3 (figure 3(e)) had a new obvious diffraction peak at 38.07°c ompared with the fresh HGMs3, which can be well-assigned to the (111) crystal face of metallic Ag (JCPDS NO. 04-0783), suggesting that a certain amount of Ag 3 PO 4 loaded on HGMs3 was reduced to Ag after treating MB photo-degradation [40]. And the color of HGMs3 become pale after being used for MB photo-degradation under the visible-light irradiation. This point can also be proved by x-ray photoelectron spectroscopy (XPS) shown as figure 4. It is found that the peaks of Ag 3d 5/2 and Ag 3d 3/2 are located at ∼368 and ∼374 eV both in figures 4(a) and (b), respectively. In figure 4(a), the XPS pattern of HGMs3 without the first photodegradation displays that the peaks of Ag 3d 5/2 and Ag 3d 3/2 were fitted at 373.98 and 368.07 eV, which can be attributed to Ag + ions in Ag 3 PO 4 [41]. And these peaks cannot continue to be divided and exhibited the smallest value of residual STD which degree of fitting was optimal. In other words, the silver elements in HGMs3 before the first photodegradation all existed in the form of Ag + . However, the Ag 3d 3/2 and Ag 3d 5/2 peaks of HGMs3 after the first photodegradation can be further divided into two different peaks at 375.14, 374.09 eV and 368.89, 368.02 eV in figure 4(b), respectively. According to the results reported by Zhang et al [41], the peaks at 374.09 and 368.02 eV can be attributed to Ag + ions in Ag 3 PO 4 , whereas the peaks at 375.14 and 368.89 eV are attributed to Ag 0 . The calculated contents of Ag 0 and Ag + are 13.07 mol% and 86.93 mol% by integrating peak area, respectively. This also means that 13% of Ag + is transformed into Ag 0 after one photodegradation of HGMs3.
Further evidence for successful loading of Ag 3 PO 4 comes from FT-IR. As can be seen from figures 5(a), (b), the difference in infrared spectra between the amino-modified and unmodified hollow glass microspheres is not  obvious. This may be due to the small amount of KH550 added, which is only 0.4% of the hollow glass microspheres mass fraction. 1065 cm −1 , 794 cm −1 and 463 cm −1 are Si-O symmetry stretching vibration, asymmetric stretching vibration and bending vibration absorption peak, respectively [42]. In figure 5(d) . We can see from figure 5(c) that HGMs3 has a more absorption peak than HGMs at 552 cm −1 , which is P-O stretching vibrations, proving that Ag 3 PO 4 exists on HGMs.
The optical absorption ( figure 6(a)) of HGMs-KH550, HGMs-Ag 3 PO 4 and pure Ag 3 PO 4 were determined with the ultraviolet-visible diffuse reflectance spectrum (UV-vis DRS). HGMs-KH550 had a certain absorption shorter than 400 nm and almost no absorption in the visible wavelength range. The band gaps of the as-prepared samples were estimated from the intercept of the tangent at the x-axis of the Tauc plot, which was (Ahν) 2 versus hv according to the formula (Ahv) 2 =hv-E g (A: absorbance; h: Planck constant; v: light frequency; E g : band gap energy) [25,26]. As shown in the inset image of figure 6(b), the E g of HGMs-KH550 was 4.59eV, which was an insulator, not a semiconductor. Pure Ag 3 PO 4 showed broad absorption of visible light at a wavelength less than 530 nm and the E g of Ag 3 PO 4 was estimated to be 2.36eV, consistent with the previous reported results [7,25]. At a wavelength<530 nm, the intensity order of visible light absorption is as follows: HGMs-KH550<HGMs1<HGMs2<Ag 3 PO 4 <HGMs4<HGMs3. HGMs3 had an absorption edge at  about 540 nm, slightly wider than the range of pure Ag 3 PO 4 . The crystal form of Ag 3 PO 4 loaded on microspheres is rhombic dodecahedrons, and the HGMs-Ag 3 PO 4 composites containing a certain amount of Ag 3 PO 4 could allow more multiple scattering of light [10]. The results also suggested HGMs3 might effectively utilize sunlight and enhance photocatalytic properties under visible light irritation. The band gaps of HGMs1, HGMs2, HGMs3, and HGMs4 are 2.28 eV, 2.29 eV, 2.37 eV and 2.35 eV, respectively. Especially, the E g of HGMs3, and HGMs4 were similar to Ag 3 PO 4 , which meant the HGMs-Ag 3 PO 4 composites showcased high optical responses under visible light irradiation, thus could be applied as a visible light active photocatalyst.

Photocatalytic properties
The effects of the Ag 3 PO 4 loading on photocatalytic properties were evaluated by determining the MB photodegradation rate under the irradiation of simulated visible light. The MB photo-degradation as a function of time was shown in figure 7(a). Prior to light irradiation, the MB solutions containing HGMs-KH550, the HGMs-Ag 3 PO 4 composites or pure Ag 3 PO 4 were placed in the dark for 60 min to achieve the adsorptiondesorption equilibrium. In the case of no light irradiation, the MB removal was ascribed to the adsorption rather the photo-degradation. MB absorbed by pure Ag 3 PO 4 and HGMs-KH550 is only about 15%, while about 54% of MB were adsorbed by HGMs3, 46% of MB adsorbed by HGMs2 or HGMs4, and 32% of MB adsorbed by HGMs1. For the floating HGMs-Ag 3 PO 4 composites, the amount of adsorption was as follows: HGMs3>HGMs2>HGMs1>HGMs-KH550. The more Ag 3 PO 4 was loaded on the floating microspheres, the more MB was adsorbed. The adsorption capability of HGMs-Ag 3 PO 4 composites was better than that of HGMs-KH550 support and pure Ag 3 PO 4 catalyst. Ag 3 PO 4 was dispersed on microspheres, increased the specific surface area of Ag 3 PO 4 and enlarged the contact area with MB solution, resulting in the enhanced adsorption capacity. The superfluous Ag 3 PO 4 particles for HGMs4 aggregated on the HGMs (figure 2(e)) and decreased the adsorption capacity.
When they were exposed under the irradiation of simulated visible light for an hour, the only a small amount of MB was degraded by HGMs-KH550. For the HGMs-Ag 3 PO 4 composites or pure Ag 3 PO 4 , as expected, and the MB concentration decreased much with the increasing of the exposed time. After an hour, the MB removal was as follows: HGMs3 (96%)>HGMs2 (91%)>HGMs4 (87%)>Ag 3 PO 4 (80%)>HGMs1 (79%). The BET surface area of HGMs was 0.760 m 2 g −1 and that of pure Ag 3 PO 4 was 1.299 m 2 g −1 , while the BET surface area of HGMs3 was 0.779 m 2 g −1 , little higher than that of raw HGMs, which may have promoted the adsorption of MB on HGMs-Ag 3 PO 4 composites. For the floating photo-catalysts, the more Ag 3 PO 4 content, the better the photo-catalytic performance of photo-catalysts. In addition, the Ag 3 PO 4 content of floating HGMs2 and HGMs3 was less than pure Ag 3 PO 4 , but the catalytic properties were better than pure Ag 3 PO 4 , which could be attributed to the better adsorption performance of HGMs-Ag 3 PO 4 composites and the larger light contact area.
Since the MB removal was very different during the adsorption process for HGMs-KH550, the HGMs-Ag 3 PO 4 composites or pure Ag 3 PO 4 , the photo-catalytic degradation rate of MB over HGMs-Ag 3 PO 4 composites and pure Ag 3 PO 4 was concerned rather than the MB removal in the photo-catalytic process. When the concentration of dyes is very low, the Langmuir-Hinshelwood model is suitable for the kinetics photocatalytic degradation, which expression is υ=− dC dt = + . k KC KC 1 r In this formula, υ, k r , K and C are the reaction rate, reaction rate constant, adsorption coefficient and the reactant concentration, respectively. When C is very small, ln C C 0 =kt is obtained, k is the apparent constant (first order rate). The photo-catalytic degradation of MB over Ag 3 PO 4 and HGMs-Ag 3 PO 4 composites followed the pseudo-first-order kinetics, as shown in figure 7(b). ln(C 0 /C) versus t is a straight line and the slope of this fitted line is the photo-catalytic rate, recorded as k. In order to simulate the quiescent state of rivers and lakes in practical application, no stirring was used in the photodegradation. The apparent rate constants (k) for floating HGMs3 and HGMs2 were 0.0444 and 0.0297 min −1 , which are larger than those for sinking HGMs4 (k=0.0229 min −1 ) and pure Ag 3 PO 4 (k=0.0242 min −1 ). HGMs3, which had the best photocatalytic performance herein, was actually loaded with 0.0768 g Ag 3 PO 4 per gram of sample (see table 2). Compared with the same mass of pure Ag 3 PO 4 , the mass of Ag 3 PO 4 contained in HGMs3 is only 7.68% (≈8 wt% of the composite shown in table 1), but its photocatalytic efficiency (k=0.0444 min −1 ) is nearly twice that of pure Ag 3 PO 4 (k=0.0242 min −1 , see figure 7(b)). However, as shown in table 1, to enhance the photocatalytic activity of Ag 3 PO 4 -composite, the proportion of Ag 3 PO 4 often exceeds 50% [18,20,24,25,30]. Therefore, the superiority of the HGMs-Ag 3 PO 4 composite is obvious.
In views of the loading of Ag 3 PO 4 catalyst, the floating HGMs-Ag 3 PO 4 composites as the photo-catalyst could not only greatly decrease the input of Ag 3 PO 4 and the cost, but also obviously enhance the photodegradation efficiency. Moreover, it is easy to salvage and recover in practical applications, indicating that the floating HGMs-Ag 3 PO 4 composites can work as an efficient and promising visible-light photo-catalyst.
In order to investigate the effect of the initial MB concentration on the photo-catalytic properties of floating HGMs3 and sinking HGMs4, the degradation rates of MB solution with three different initial concentrations were measured under visible-light irradiation demonstrated in figure 8. When the initial MB solutions increased from 5 to 20 mg l −1 , the MB removal efficiency of floating HGMs3 and sinking HGMs4 decreased much, but the high removal efficiency could be reached after a longer contact time. It was consistent with the results reported in previous works [44]. With the increasing of initial MB solutions, the apparent rate constants (k) of the floating HGMs3 for three initial MB solutions of 5, 10 and 20 mg l −1 (in figure 8(a)) were respectively 0.0444, 0.0119 and 0.0037 min −1 , while the values of the sinking HGMs4 were only 0.0229, 0.0069 and 0.0016 min −1 . These results illustrated that the floating HGMs3 has larger light contact area and better photo-catalytic performance than the sinking HGMs4.
The cycling photo-degradation tests were carried out to investigate the stability of floating HGMs-Ag 3 PO 4 composites under visible-light irradiation, as exhibited in figure 9(a). After each cycle, the amount of MB adsorbed on HGMs3 decreased gradually, the apparent rate of photo-degradation reaction gradually slowed down. This result is due to the fact that Ag 3 PO 4 was reduced to Ag (Ag + +e -→Ag 0 ) under light exposure [32]. As described in figures 3(e) and 4, silver component was proven in the HGMs-Ag 3 PO 4 composites. The photocatalytic efficiency of the sample HGMs3 after five cycles was poor, but we can use it as a catalyst for chemical reduction to degrade 4-NP (p-Nitrophenol) and MO (Methyl Orange), as shown in figure 9(b). With the addition of excess reducing agent sodium borohydride (NaBH 4 ), the degree of degradation of 0.1 mmol l −1 4-NP and MO is very low, but after adding the 0.25 g l −1 5-cycled HGMs3, both 4-NP and MO can be almost fully degraded within 10 min. In this process, the reducing agent NaBH 4 plays two roles: in-situ reduction of the remaining Ag 3 PO 4 on the 5-cycled HGMs3 to Ag 0 and reduction of 4-NP and MO. In other words, 5-cycled HGMs3 can also be used as a catalyst to degrade organic pollutants and further exert its effectiveness.

Possible photo-catalytic mechanism
The following reasons may account for the improved photo-catalytic properties of floating HGMs-Ag 3 PO 4 composites. In the first place, with hollow glass microspheres as the carrier of Ag 3 PO 4 , Ag 3 PO 4 dispersed on the surface of HGMs not only can improve the light contact chance to enlarge the active surface area of the composites, but also enhance the lattice defects to generate new active centers and improve the performance of Ag 3 PO 4 [45]. Secondly, the as-prepared HGMs-Ag 3 PO 4 composites can be floating on the treatment solution, enlarging the light contact area and improving the photo-catalytic efficiency. Thirdly, Ag 3 PO 4 particles loaded on the surface of HGMs showed rhombic dodecahedrons using facile ion exchange method, which had higher surface energy values for {110} facet [11] and provide more catalytic active sites than other crystal forms, leading to improved adsorption capacity of dye molecules and better photo-catalytic properties [46]. In addition, Ag 3 PO 4 itself has excellent photo-catalytic properties, mainly due to the high dispersion of the conduction band (CB, +0.45 V versus NHE) and the large negative charge induced effect of PO 4 3− in Ag 3 PO 4 , resulting in the separation of photo-generated electrons and holes [47]. As depicted in figure 10, the electrons on the valence band (VB) of Ag 3 PO 4 absorb the photon energy larger than the bandgap of Ag 3 PO 4 (2.45V versus NHE) to jump to the conduction band, leaving holes with oxidative properties at the original valence band position [48]. These photo-generated holes could oxidize H 2 O into strongly oxidative ·OH radicals and the electrons on the CB of Ag 3 PO 4 combine with the absorbed O 2 to produce ·O 2 − . These reactive species ·OH, h + , and ·O 2 − play a conducive role in the oxidative degradation of organic pollutants MB [49,50]. Four methyl groups of MB can facilitate attack on MB by electrophilic species (·OH or h + or ·O 2 − ) in the demethylation process and be advantageous for accelerating the photocatalytic oxidative degradation of MB [51,52]. However, photogenerated electrons have a negative effect at the same time in photocatalytic degradation of MB. With the extension of visible-light illumination time, Ag replaces more active sites of Ag 3 PO 4 for Ag 3 PO 4 is easily reduced to Ag by photogenerated electrons under light, thus the amount of photo-generated holes is declined and the  photocatalytic performance of Ag 3 PO 4 is decreased without any sacrificial agent [53]. Therefore, the further investigation was necessary to improve the stability of Ag 3 PO 4 .

Conclusions
In summary, Ag 3 PO 4 was attached to the surface of the inorganic material HGMs via a facile ion exchange process to prepare the floating HGMs-Ag 3 PO 4 composites using silane coupling agent as a bridge. The photocatalytic properties were assessed by the MB photo-degradation under the visible-light irradiation. The photocatalytic rate of the floating HGMs-Ag 3 PO 4 composites was much larger than that of pure Ag 3 PO 4 and the sinking HGMs-Ag 3 PO 4 composites. The floating samples could float on the treatment solution and enlarge the contact area significantly between the photo-catalyst and the light, which is beneficial to enhance the efficiency of photo-degradation. The floating HGMs-Ag 3 PO 4 composites, a promising photocatalyst that provides a potential approach for large-scale practical application on the removal of organic dye contaminants, which greatly cut down the cost and enhance the photo-catalytic properties under the visible-light irradiation.