Defective Bismuth Oxide as Effective Adsorbent for Arsenic Removal from Water and Wastewater

In this work, we report solid-state synthetized defective Bi2O3 containing Bi(V) sites as effective and recyclable arsenic adsorbent materials. Bi2O3 was extensively characterized, and structure-related adsorption processes are reported. Both As(V) and As(III) species-adsorption processes were investigated in a wide range of concentrations, pH values, and times. The effect of several competing ions was also tested together with the adsorbent recyclability.


Introduction
Water shortage is one of the most concerning threats envisaged for the future, given the growing population and the changing climate [1]. Enhancing the availability of clean water by efficiently removing natural or anthropic contaminants is therefore of paramount importance. Wastewater management is another essential link in reducing the degradation of the environment and improving the sustainability of productive systems [2].
Some of the most dangerous and difficult-to-avoid pollutants affecting water quality are the potentially toxic elements (PTE), ubiquitous in both developed [3,4] and developing countries [5,6]. Among PTE, arsenic is one of the most diffused and dangerous contaminants [7], with varying toxicity levels according to its chemical speciation. In general, the inorganic species, arsenite and arsenate, are considered more toxic than the organic ones [8]. Furthermore, arsenite is more toxic and more mobile in the environment than arsenate [9], which poses further challenges for its removal from water compared to arsenate, requiring, e.g., preliminary oxidation steps in water treatment plants. High levels of As (above 10 µg L −1 ) in groundwater used for drinking and irrigation are not infrequent in southeast Asia, where endemic As poisoning has caused long-term health effects and a public health emergency [5][6][7].
Nowadays, a number of different technologies are available for arsenic removal from groundwater and wastewater, including reverse osmosis [10], ion exchange [11], precipitation [12] and adsorption [13]. Among these, the techniques based on As adsorption are preferred because of their efficiency, economic feasibility and wide adsorbent availability [14] (e.g., oxides and oxy(hydr)oxides of Fe, Al, Ti, and Mn [15,16]; double layered hydroxides [17]; and carbon- [18][19][20] and silica-modified [21,22] surfaces). The key points to be considered for the choice of the adsorbent include its adsorption efficiency and the economic and environmental sustainability of its production and use. To this last point, the potential for adsorbent regeneration deserves great attention since the plants for contaminant removal may generate great amounts of solid waste as contaminant-saturated media, which is hazardous for humans and the environment and requires high costs for disposal [23]. For these reasons, continuing the research aimed at developing new highperformance adsorbents for water treatment is essential. Research has been focusing lately on innovative substrates for the removal of contaminants from water, with bismuth-based materials standing out due to being safe, non-toxic and non-carcinogenic [24]. Bi-based compounds have a number of biomedical applications [25] and display great potential for photocatalytic degradation of organic contaminants [26] and for the removal of different anionic species often found in groundwater and wastewater (e.g., PO 4 3− , NO 3 − , SO 4 2− , Cl − , and F − ) [27]. These materials have been proposed as possible candidates for As adsorption only recently [28,29], with promising results depending on material composition and synthesis method. Wang et al. [29] tested As(III) and As(V) adsorption on Bi-based substrates with promising results, and Zhu et al. [30] described an innovative approach based on bismuth doped biochar for effective arsenic removal showing the promising properties of bismuth based material for the removal of aqueous inorganic pollutants. In this research, the authors suggested an adsorption mechanism based on both surface interactions and electrochemical processes.
In this work, we report the use of surface defective bismuth oxide as an effective adsorbent to remove arsenite (As(III)) and arsenate (As(V)) species from water solutions while considering the influence of pH, contact time, and competitive ions and testing the potential for adsorbent regeneration, together with an extensive characterization of the bismuth oxide material.

Synthesis and Characterization of Bismuth Oxide (Bi 2 O 3 )
Bismuth oxide (Bi 2 O 3 ) was synthesized through solid-state reaction by direct heating of bismuth nitrate (Bi(NO 3 ) 3 ·5H 2 O, 99% purity, Sigma Aldrich, St. Louis, MO, USA) at 150 • C for 30 min for dehydration, followed by 2 h at 250 • C. The material was subsequently annealed at 550 • C for 2 h and finally cooled gradually until reaching room temperature.
X-ray diffraction (XRD) patterns of the synthesized Bi 2 O 3 were measured with a Pan'Analytical X'Pert Pro diffractometer (Pan'Analytical, Almelo, The Netherlands) equipped with a Cu Kα source. The spectra were recorded between 5-90 • 2θ, with a step size of 0.05 • /s and a count time of 8 s/step. Diffraction patterns were indexed using Match!3™ software and Powder Data File database (P.D.F. 2000, International Centre of Diffraction Data, Newtown Square, PA, USA).
High-resolution field emission scanning electron microscopy (FE-SEM) was used to determine the surface morphology of Bi 2 O 3 using a Zeiss Supra-40 microscope (Zeiss, Oberkochen, Germany).
The surface area, pore volume, and pore diameter of the Bi 2 O 3 were determined by analysis of data points obtained for N 2 adsorption at −196 • C using a TriStar II analyzer (Micromeritics Instrument Corporation, Norcross, GA, USA) and calculated using the Brunauer-Emmett-Teller (BET) model.
The electrophoretic mobility of Bi 2 O 3 equilibrated for 24 h in 0.01 M KCl at pH values between 3 and 10 (adjusted with 0.01 or 0.1 M HNO 3 or KOH) was measured by Laser Doppler Velocimetry coupled with Photon Correlation Spectroscopy (DELSA 440, Beckman Coulter Inc., Brea, CA, USA), and the zeta potential (ζ) was calculated using the Smoluchowski equation [31]. The plot of the ζ potential vs. pH (not shown) provided a PZC for the Bi 2 O 3 at pH 8.5, in line with the value range reviewed by Ranjan et al. [27] for Bi-based adsorbents.
The surficial chemical composition of the Bi 2 O 3 was investigated by X-ray photoelectron spectroscopy (XPS) with a PHI 5000 Versaprobe spectrometer (Physical Electronics, Chanhassen, MN, USA) equipped with monochromatic Al Kα X-ray source and operating at 1486.6 eV energy, 15 kV voltage, and 1 mA anode current.
All adsorption experiments were carried out in duplicate using 40 mg of Bi 2 O 3 washed with 10 mL 0.01 M HNO 3 for 1 h, then with doubly deionized water until the conductivity was below 10 µS m −1 , indicating the removal of salts from the synthesis. The Bi 2 O 3 was resuspended in 5 mL of deionized water, sonicated for 15 min (Transsonic T460, Camlab Ltd., Over, UK), then equilibrated with As(III) or As(V) solutions for a final volume of 10 mL on a rotating shaker at 40 rpm and 25 ± 3 • C. The suspensions were subsequently centrifuged for 5 min at 3000 rpm and filtered through a 0.20 µm nylon filter, and the supernatant was analyzed for As(III) or As(V) by the molybdenum blue method proposed by Huang and Fujii [33] unless otherwise stated.
The amount of adsorbed As, Q a (µmol g −1 Bi 2 O 3 ) was calculated using the following equation: where C 0 (µmol L −1 ) is the initial concentration of As(III) or As(V), C e (µmol L −1 ) is the equilibrium concentration, V (L) is the volume of solution, and m (g) is the mass of Bi 2 O 3 .
Adsorption isotherm experiments were conducted at initial As(III) and As(V) concentrations raging between 1.33 and 800 µmol L −1 and reaction time of 24 h. The data were fitted to the linear and non-linear Langmuir and Freundlich models, and to the two site Langmuir model, as presented in Table S1 of the Supplementary Material.
Adsorption kinetics experiments were conducted at time intervals ranging between 5 min and 48 h and initial As concentration of 533 µmol L −1 . Data were fitted to the pseudo-first-order, the pseudo-second-order, and the intra-particle diffusion models, in the forms presented in Table S2.
The competitive effect of coexisting anions on the adsorption of As was studied by equilibrating Bi 2 O 3 with 533 µmol L −1 As(III) or As(V) solution in the presence of 0.01 M or 0.1 M Cl − , NO 3 − , SO 4 2− or PO 4 3−, and 0.01 M SiO 3 2− for 24 h, and As was determined by hydride generation (HG) coupled with AAS (Perkin-Elmer 4100 equipped with a FIAS 400 hydride generator; Perkin-Elmer Inc., Waltham, MA, USA). The reusability of Bi 2 O 3 as an adsorbent was evaluated by performing four cycles of As adsorption/desorption. Thus, after equilibration with 533 µmol L −1 As(III) or As(V) solution for 24 h, the Bi 2 O 3 was separated by centrifugation and washed three times with 0.1 M KNO 3 to remove the arsenic-containing solution. Thereafter, the adsorbent was re-suspended in 10 mL 0.1 M NaOH solution and shaken for 30 min to displace adsorbed As. The supernatants were separated and analyzed for As(III) or As(V), and the Bi 2 O 3 was dried at 40 • C, resuspended at the initial pH values, and reused.

Characterization of the Bi 2 O 3
Solid-state synthesis is a well-established method to produce size-controlled materials using a mixture of several precursors [34]. In this work, the synthesis of Bi 2 O 3 was performed with a solid-state approach using Bi(NO 3 (Figure 1) [35]. This indicated a predominance of the α-phase, according to ICDD database, with a coefficient of similarity of up to 82%, revealing structural and surface defects in the region from 30 • 2θ to 50 • 2θ, as shown in Figure 1.

Characterization of the Bi2O3
Solid-state synthesis is a well-established method to produce size-controlled materials using a mixture of several precursors [34]. In this work, the synthesis of Bi2O3 was performed with a solid-state approach using Bi(NO3)3·5H2O as the single precursor. The XRD analysis of the obtained Bi2O3 showed peaks at 2θ = 19.8°, 21 (Figure 1) [35]. This indicated a predominance of the α-phase, according to ICDD database, with a coefficient of similarity of up to 82%, revealing structural and surface defects in the region from 30° 2θ to 50° 2θ, as shown in Figure 1. The FE-SEM analysis showed that the surface of Bi2O3 was smooth with a diffuse porous structure ( Figure 2). Plate-like particles form slit-shaped macro-and mesopores, which may have important implications in the adsorption of As(III) and As(V) from solution. The FE-SEM analysis showed that the surface of Bi 2 O 3 was smooth with a diffuse porous structure ( Figure 2). Plate-like particles form slit-shaped macro-and mesopores, which may have important implications in the adsorption of As(III) and As(V) from solution. The Bi2O3 had a relatively low surface area of 0.83 m 2 g −1 and an average pore diameter of approximately 23.8 nm, which may be attributed to the high temperature employed in the synthesis, previously shown to cause the collapse of micropores, resulting in the formation of mesopores [36]. The shape of N2 sorption/desorption isotherms ( Figure S2) suggests that the synthesized Bi2O3 is characterized by a wide distribution of pore sizes due to plate-like particles, in agreement with the FE-SEM observations. employed in the synthesis, previously shown to cause the collapse of micropores, resulting in the formation of mesopores [36]. The shape of N 2 sorption/desorption isotherms ( Figure S2) suggests that the synthesized Bi 2 O 3 is characterized by a wide distribution of pore sizes due to plate-like particles, in agreement with the FE-SEM observations.
The surface defects of the synthesized Bi 2 O 3 were further investigated by XPS spectrometry, as reported in Figure 3. The Bi2O3 had a relatively low surface area of 0.83 m 2 g −1 and an average pore diameter of approximately 23.8 nm, which may be attributed to the high temperature employed in the synthesis, previously shown to cause the collapse of micropores, resulting in the formation of mesopores [36]. The shape of N2 sorption/desorption isotherms ( Figure S2) suggests that the synthesized Bi2O3 is characterized by a wide distribution of pore sizes due to plate-like particles, in agreement with the FE-SEM observations.
The surface defects of the synthesized Bi2O3 were further investigated by XPS spectrometry, as reported in Figure 3.  [37], bismuth oxides are characterized by two monomodal signal peaks at around 157.9 eV and 163.2 eV. The synthetized Bi2O3 showed an additional component, probably due to the higher oxidation state of the bismuth species. These species could be induced by the degradation of nitrate groups leading to overoxidation of Bi(III) to high unstable Bi(V) [38,39].
According to this hypothesis, oxygen signal showed three components while pure bismuth oxide showed only at 530 eV with one shoulder at around 529 eV due to hydroxylic functionalities [40]. The additional oxygen peak could be associated with Bi(V) sites included in Bi2O3 crystalline structure. Furthermore, no appreciable signals in the region of nitrogen (390-410 eV) were detected ( Figure S3), supporting the presence of Bi(V) instead of bismuth sub-nitrates species.  [37], bismuth oxides are characterized by two monomodal signal peaks at around 157.9 eV and 163.2 eV. The synthetized Bi 2 O 3 showed an additional component, probably due to the higher oxidation state of the bismuth species. These species could be induced by the degradation of nitrate groups leading to overoxidation of Bi(III) to high unstable Bi(V) [38,39].
According to this hypothesis, oxygen signal showed three components while pure bismuth oxide showed only at 530 eV with one shoulder at around 529 eV due to hydroxylic functionalities [40]. The additional oxygen peak could be associated with Bi(V) sites included in Bi 2 O 3 crystalline structure. Furthermore, no appreciable signals in the region of nitrogen (390-410 eV) were detected ( Figure S3), supporting the presence of Bi(V) instead of bismuth sub-nitrates species.

Adsorption Isotherm Studies
The adsorption of As was strong at low initial concentrations, with approximately 98% of As(III) and As(V) being adsorbed at C 0 between 13.3 and 53.4 µmol L −1 (Figure 4). The adsorption efficiency decreased at higher C 0 to around 14%, representing 27.5 µmol As(III) and 26.2 µmol As(V) g −1 Bi 2 O 3 . Despite their generally similar adsorption, As(III) saturated the surface of Bi 2 O 3 at a lower C 0 than As(V) (400 vs. 800 µmol L −1 ).

Adsorption Isotherm Studies
The adsorption of As was strong at low initial concentrations, with approximately 98% of As(III) and As(V) being adsorbed at C0 between 13.3 and 53.4 µmol L −1 (Figure 4). The adsorption efficiency decreased at higher C0 to around 14%, representing 27.5 µmol As(III) and 26.2 µmol As(V) g −1 Bi2O3. Despite their generally similar adsorption, As(III) saturated the surface of Bi2O3 at a lower C0 than As(V) (400 vs. 800 µmol L −1 ). We plotted the experimental data of As(III) and As(V) adsorption on Bi 2 O 3 in nonlinearized Langmuir, Freundlich, and two-site Langmuir isotherms (Figure 4), which may help understand the mechanisms controlling the partition of the adsorbate between the liquid and the solid phases at equilibrium. The linearized form of the Langmuir and Freundlich equations are presented in Figure S4. The Langmuir model is valid for monolayer adsorption on a surface with finite number of identical sites [41], while Freundlich model can be applied to non-ideal adsorption on heterogeneous surfaces and to multilayer sorption [42]. The two-site Langmuir model describes the adsorption of a species on two or more distinct types of sites, which can be described with their own Langmuir expression [43].
The non-linear Langmuir model did not yield a good fitting of the data, as suggested by the difference between the calculated and the experimental values for As(III) and As(V) adsorption ( Figure 4a) and by the low R 2 (Table 1). In most adsorption studies, Langmuir (as well as Freundlich) model is applied in linearized form, since this allows simpler calculation of the equation constants, often yielding apparently good descriptions of the adsorption data. Indeed, the linearized form ( Figure S4a) had a high R 2 also in our case (0.996 and 0.998 for As(III) and As(V), respectively), and a Q max close to experimental values ( Table 1), suggesting that As molecules adsorb on the Bi 2 O 3 surface forming a monolayer. However, linearization of the Langmuir equation is known to bias the regression analysis toward fitting the low C e values better than the high C e values [44]. The regression analysis may therefore be strongly affected by small errors in the low aqueous concentration range because of the reciprocal form of the equation. Moreover, the adsorbing sites of the defectrich material are likely heterogeneous, binding As with different strengths and possibly through more than a single mechanism, hence no mechanistic inferences can be drawn solely from the good fitting of the linearized Langmuir equation. Table 1. Isotherm parameters obtained for non-linear and linear Langmuir and Freundlich models and for the two-site Langmuir model for the adsorption of As(III) and As(V) onto Bi 2 O 3 .

Non-Linear Model
Linear Model As ( As(III) 16 Q max , Q max1 , and Q max2 (µmol g −1 )-maximum amount of analyte that can bind to the Bi 2 O 3 as a monolayer or on type 1 and type 2 sites, respectively; K L , K L1 , and K L2 (L mol −1 )-Langmuir affinity constants of the monolayer and the type 1 and type 2 sites. K F -Freundlich constant, representing adsorption capacity at unitarian concentration at equilibrium. n-empirical constant, indicating the adsorption intensity of the system. R 2 -coefficient of determination.
The non-linear Freundlich model was a better fit for the experimental data compared to the non-linear Langmuir (Figure 4a,b), giving R 2 of 0.957 and 0.904 for As(III) and As(V), respectively ( Table 1), suggesting that their adsorption mechanism on Bi 2 O 3 may be controlled by multilayer adsorption on heterogeneous sites. The value of 1/n gives an indication on the affinity between adsorbent and adsorbate, and n between 1 and 10 suggests favourable adsorption. In our case, the high n values for As(III) and As(V) ( Table 1) indicated that their adsorption on Bi 2 O 3 is favourable and helped describe the fast increase in Q a with increasing C e at low C 0 values, which gave rise to a highly curved isotherm (Figure 4b). Despite the similar adsorption of As(III) and As(V), the linearized form of the Freundlich model was not suitable to describe As(V) data ( Figure S4b), giving R 2 of 0.767, while the fitting of As(III) adsorption gave an R 2 of 0.927 (Table 1). On the other hand, the adsorption of As(V) on Bi-impregnated biochar and Bi-impregnated aluminium oxide was well described by the linear Langmuir and linear Freundlich models, respectively [28,30], indicating that the preparation of the adsorbent plays an important role in the adsorption mechanisms.
The two-site Langmuir model assumes that the adsorbent has two types of adsorption sites with different adsorption energies and complies with the three assumptions of the Langmuir model: (i) the adsorption of the molecules forms a monolayer; (ii) the adsorption energy is uniform and remains unchanged; and (iii) there is no interaction between adsorbate molecules [43,45]. Our data fit the two-line Langmuir model (Figure 4c) better than the non-linearized forms of one-site Langmuir and Freundlich equations (Figure 4a,b), giving R 2 values of 0.987 and 0.920 for the adsorption of As(III) and As(V), respectively. The sum between Q max1 and Q max2 was higher than the Q max calculated with the non-linear one-site Langmuir equation (Table 1) and was closer to experimental values, suggesting that the two-site Langmuir may be more appropriate for the description of As adsorption on Bi 2 O 3 . While the values of K L1 are comparable for the two As forms, the higher K L2 for As(V) compared to As(III) suggests that type II binding sites are more favourable for the adsorption of As(V) than of As(III).
When the adsorption capacity of Bi 2 O 3 is calculated in terms of µmol m −2 from the sum of the two adsorbed layers predicted by the two-sites Langmuir model, the result is of about 36 µmol m −2 for As(III) and 34 µmol m −2 for As(V), which is in the order of magnitude found by Wang et al. [29] for δ-Bi 2 O 3 , and on average much higher than other metal-oxide based adsorbents (typically 2 to 5 µmol m −2 for Fe-(hydr)oxides at the most favourable pH) [46]. The adsorbed concentrations of As(III) and As(V) were much higher than 1.11 µmol g −1 obtained for As(V) adsorption onto a hydrous bismuth oxide [47], although smaller than the concentrations adsorbed onto a δ-Bi 2 O 3 with SSA of 8.99 m 2 g −1 (170.8 and 49.3 µmol m −2 , respectively) [29]. This very high concentration of adsorbed As was nearly irreversibly retained within the solid [29], hampering its regeneration and reusability, differently from the adsorbent tested in this study ( § 3.2.4). For obtaining adsorbents with larger specific surface area, Bi-doped materials, such as biochar [30] or Al oxides [28], were prepared, reaching SSA values of 190.4 and 130.65 m 2 g 1 respectively, with 1.1 and 2.7 µmol m −2 of As(III) adsorbed. The adsorption capacity of these Bi-impregnated substrates seemed to be controlled by the amount of Bi incorporated rather than by the specific surface [30].
The high adsorption density for surface unit displayed by bismuth oxides in this study and in [29] increases the interest in developing Bi-based adsorbents for water treatment. It could suggest that specific single layer adsorption might not be the sole adsorption mechanism, in agreement with the better fitting of the adsorption data by Freundlich than Langmuir model in the non linearized form (Figure 4), and by the better performance of two-site than one-site Langmuir equation (Table 1).

Adsorption Kinetic Studies
Adsorption kinetic experiments showed that As adsorption on Bi 2 O 3 proceeded rapidly at first, reaching 94% of the total As(III) and As(V) concentration in 4 and 2 h, respectively, then increased slowly until 48 h (Figure 5a). The pseudo-first order model was adequate to describe the adsorption of As(III) for the first 1 h and of As(V) for the first 2 h of adsorption, but beyond that, the data deviated from linearity (Figure 5b). This model gave low R 2 of 0.732 and 0.770 for As(III) and As(V), respectively, and we therefore divided the two curves, and the calculated parameters are presented in Table 2. However, the inconsistencies between experimental and calculated values of q e indicated that this model is not the most suitable to describe the adsorption of As on Bi 2 O 3 . The pseudo-secondorder kinetics model on the other hand was able to describe the adsorption more precisely, yielding a linear plot for the entire range of adsorption time (Figure 5c), high R 2 values, and q e similar to the experimental values, providing the best description of As(III) and As(V) adsorption kinetics among the tested models (Table 2). This suggests that the rate-limiting step for the adsorption of As(III) and As(V) may be controlled by the same mechanism and only slightly changes with time.
2 h of adsorption, but beyond that, the data deviated from linearity (Figure 5b). This model gave low R 2 of 0.732 and 0.770 for As(III) and As(V), respectively, and we therefore divided the two curves, and the calculated parameters are presented in Table 2. However, the inconsistencies between experimental and calculated values of qe indicated that this model is not the most suitable to describe the adsorption of As on Bi2O3. The pseudosecond-order kinetics model on the other hand was able to describe the adsorption more precisely, yielding a linear plot for the entire range of adsorption time (Figure 5c), high R 2 values, and qe similar to the experimental values, providing the best description of As(III) and As(V) adsorption kinetics among the tested models (Table 2). This suggests that the rate-limiting step for the adsorption of As(III) and As(V) may be controlled by the same mechanism and only slightly changes with time.   k 1 , k 2 (min −1 ), k (g/µmol/min), k i1 , k i2 (µmol/g/min 0.5 )-adsorption rate constants. q e1 , q e2 , q e (µmol g −1 )concentration of As(III) or As(V) adsorbed on Bi 2 O 3 at equilibrium. t (min)-time. R 2 -coefficient of determination.
In order to verify the importance of intraparticle diffusion as the possible limiting step of As adsorption on Bi 2 O 3 , the data were fitted to the Weber-Morris intraparticle diffusion model [48]. This model assumes that if the plot shows a straight line passing through origin, then the adsorption process is governed by intraparticle diffusion. Our data generated plots exhibiting multiple straight lines that did not pass through the origin (Figure 5d), indicating that intraparticle diffusion is not the major rate limiting step in As multistage adsorption. The curve was thus divided into two distinct linear regions for As(III) and As(V) adsorption, which suggests that the adsorption of As on Bi 2 O 3 might not be completed in one step, because it is a rather complex process in which more than one mechanism dictates the sorption mechanisms. Similar observations were also made by Zhu et al. for the adsorption of As on Bi-impregnated Al oxide [28] and Bi-impregnated biochar [30], and by Srivastav and co-workers for the adsorption of fluoride [49] and nitrate [50] on hydrous bismuth oxide. The slightly lower rate constants calculated for As(V) than for As(III) ( Table 2) support the differences between the adsorption of the two species and the higher affinity of Bi 2 O 3 for As(III) compared to As(V).

Effect of Competing Anions
Competing ions had different effects on the adsorption of As(III) and As(V) on Bi 2 O 3 , depending on the species added and their ionic strength, as illustrated in Figure 6. The results showed that the adsorption of As(III) was not hindered by the coexistence of 0.01 M Cl − , NO 3 − and SO 4 2− . However, 0.01 M SiO 3 2− did reduce the amount of As(III) adsorbed on Bi 2 O 3 from 27.5 to 15.5 µmol g −1 (Figures 4 and 6a). This may be attributed to the similar chemistry between silicic acid (H 4 SiO 4 ) and arsenous acid (H 3 AsO 3 ), which have similar dissociation constants (pK a1 = 9.9 and 9.2, respectively [32]), enabling SiO The adsorption of As(V) was affected by the competing anions to a higher extent compared to As(III), even in the presence of 0.01 M Cl − , NO3 − , and SO4 2− , when adsorbed As(V) was reduced from 26.2 (Qmax) to approximately 18.4 µmol g −1 , while 0.01 M SiO3 2− decreased retained As(V) to 14.5 µmol g −1 (Figure 6a). These results may be attributed to the greater sensitivity of the negatively charged arsenate anions than the nearly undissociated arsenous acid to the formation of an electrostatic barrier by the adsorption of the competing anions. Indeed, the mass transfer of NO3 − on hydrous bismuth oxide, for example, was shown to occur in the first 40 min of adsorption [51], which is faster than the 2 h estimated for the mass transfer of As(V) on Bi2O3 (Figure 5d), suggesting that the adsorption of the negatively charged As(V) on Bi2O3 would be hindered in the presence of some anions because of electrostatic repulsion. However, raising the concentration of Cl − , NO3 −, and SO4 2− from 0.01 M to 0.1 M affected the amounts of adsorbed As(V) by only 10%, 5% and 14%, respectively (Figure 6b), while the amounts of adsorbed As(III) diminished more markedly, suggesting that As(V) adsorption was less affected by the increase in the background ionic strength compared to As(III). In a previous study conducted with δ-Bi2O3, Wang et al. [29] also observed a greater interference of competing anions on As(V) than on As(III) adsorption with concentrations of competitors as low as 0.1 and 1.0 mM. Thus, although the adsorption of As(V) was influenced even by low concentrations of Cl − , NO3 − , and SO4 2 , the decrease was not linear with the increase in the The adsorption of As(V) was affected by the competing anions to a higher extent compared to As(III), even in the presence of 0.01 M Cl − , NO 3 − , and SO 4 2− , when adsorbed As(V) was reduced from 26.2 (Q max ) to approximately 18.4 µmol g −1 , while 0.01 M SiO 3 2− decreased retained As(V) to 14.5 µmol g −1 (Figure 6a). These results may be attributed to the greater sensitivity of the negatively charged arsenate anions than the nearly undissociated arsenous acid to the formation of an electrostatic barrier by the adsorption of the competing anions. Indeed, the mass transfer of NO 3 − on hydrous bismuth oxide, for example, was shown to occur in the first 40 min of adsorption [51], which is faster than the 2 h estimated for the mass transfer of As(V) on Bi 2 O 3 (Figure 5d), suggesting that the adsorption of the negatively charged As(V) on Bi 2 O 3 would be hindered in the presence of some anions because of electrostatic repulsion. However, raising the concentration of Cl − , NO 3 −, and SO 4 2− from 0.01 M to 0.1 M affected the amounts of adsorbed As(V) by only 10%, 5% and 14%, respectively (Figure 6b), while the amounts of adsorbed As(III) diminished more markedly, suggesting that As(V) adsorption was less affected by the increase in the background ionic strength compared to As(III). In a previous study conducted with δ-Bi 2 O 3 , Wang et al. [29] also observed a greater interference of competing anions on As(V) than on As(III) adsorption with concentrations of competitors as low as 0.1 and 1.0 mM. Thus, although the adsorption of As(V) was influenced even by low concentrations of Cl − , NO 3 − , and SO 4 2 , the decrease was not linear with the increase in the concentration of the competing anions since As(V) bonding took place even in more concentrated saline solutions. On the other hand, As(III) adsorption decreased at the higher electrolyte concentration, but the final adsorbed amount remained greater than that of As(V) in all cases (Figure 6), indicating a better performance of Bi 2 O 3 in binding the more toxic and generally more difficult to remove As(III), even in a saline environment. The reduction in As adsorption onto Bi 2 O 3 with high saline concentration may be attributed to the interaction of the different dissolved inorganic species (Cl − , NO 3 − , SO 4 2− , SiO 3 2−, and PO 4 3− ) with the surfaces of the material [27], consequently competing for the active adsorption sites and deteriorating the capacity of Bi 2 O 3 to remove As.
The adsorption of As(III) and As(V) in the presence of 0.01 M and 0.1 M PO 4 3− was completely hampered (Figure 6), suggesting that Bi 2 O 3 had a greater affinity for PO 4 3− than for both As species. Even lower phosphate concentrations significantly decreased As adsorption on δ-Bi 2 O 3 [29], affecting As(V) adsorption to a larger extent compared to As(III). However, the effect of PO 4 3− on Bi-based compounds employed for the adsorption of As is not routinely determined [28,30,47], which may result in biased indications regarding the efficacy of this material for As removal from complex aqueous media.

Regenerability and Reusability of Bi 2 O 3
The possibility to regenerate and reuse the adsorbent greatly enhances its usefulness in water treatment processes. This implies a regeneration step in which all or some of the toxic substance adsorbed is separated from the material, which can be subsequently reused. We studied the removal of As from solution by four adsorption/desorption sequences and observed that the efficiency of Bi 2 O 3 , although decreasing to some extent with each regeneration cycle (Figure 7), remained satisfactory, particularly for As(III), up to the last tested step. The concentrations of adsorbed As(III) and As(V) in the second cycle were 71% and 54% of the amounts adsorbed in the first cycle, respectively. In the third and fourth cycle, the concentrations of As(III) adsorbed remained relatively constant, while As(V) adsorbed after the last regeneration cycle was approximately 30% of the initial amount. The reduction in As adsorption may be due to its gradual stabilization on the surface of Bi 2 O 3 , as indicated by the incomplete removal with NaOH (Figure 7), therefore rendering part of the active sites unavailable for subsequent As binding. Indeed, washing with 0.1 M KNO 3 and NaOH resulted in the removal of 6.5 µmol As(III) g −1 and 11.0 µmol As(V) g −1 in the first cycle (Figure 7), representing 21% and 39% of the adsorbed amounts, respectively, which may be constituted mainly by the most weakly-bound portion of As anions. The concentrations of As retained in each cycle, however, always exceeded those previously desorbed, which resulted in a linear accumulation of As(III) (R 2 = 1.000) and As(V) (R 2 = 0.969) on the surface of the material (Figure 7), indicating that the incomplete As removal was compensated by new adsorbing sites becoming available after the regeneration steps. Indeed, the decreasing of the As adsorbed at each cycle was paralleled by a reduction in the capacity of NaOH to displace bound As, as desorbed As(III) concentrations decreased from 6.5 to 1.2 µmol g −1 (Figure 7a), and As(V) decreased from 11.0 to 3.9 µmol g −1 (Figure 7b). Our results are in line with those obtained by Zhu et al. [28] for the amounts of As removed from Bi-impregnated Al 2 O 2 with 0.1 M HCl, which decreased with each regeneration cycle. Although a smaller fraction of the adsorbed As was removed at each regeneration step, the material continued to adsorb substantial amounts of both As specie. This resulted in a progressive build-up of As on the adsorbing substrate, which remained linear for As(III) after four regeneration cycles and started bending after the third one for As(V) (Figure 7). al. [28] for the amounts of As removed from Bi-impregnated Al2O2 with 0.1 M HCl, which decreased with each regeneration cycle. Although a smaller fraction of the adsorbed As was removed at each regeneration step, the material continued to adsorb substantial amounts of both As specie. This resulted in a progressive build-up of As on the adsorbing substrate, which remained linear for As(III) after four regeneration cycles and started bending after the third one for As(V) (Figure 7). The efficiency of Bi 2 O 3 to adsorb and retain As(III) was higher compared to As(V) throughout the four regeneration cycles, as suggested by the total concentrations of the two species (73.7 µmol g −1 vs. 43.1 µmol g −1 , Figure 7). This further confirms the higher affinity of Bi 2 O 3 for As(III) compared to As(V), observed in the adsorption studies and examining the competition with other ions (Figures 4 and 6). The building up of bound As, however, exceeding the Q max observed in the adsorption studies (Figure 4), suggests that the washing of the substrate with 0.1 M KNO 3 and desorption with 0.1 M NaOH may interact with the defective surface of the Bi 2 O 3 , consequently activating new sites available for As adsorption, or that other mechanisms different from adsorption, such as surface precipitation, might contribute to As retention.

In-Solution Behaviour of Defective Bi 2 O 3 and Implication for its Usage in Water Treatment
Bi 2 O 3 could be easy tailored through surface modifications both during and after its synthesis [26,52,53]. This property is very attractive for electrochemical applications but implies unneglectable consequences during contaminant adsorption in real groundwater and wastewater environments. When the yellow-colored defective Bi 2 O 3 is interacted with nitric acid, the solid color turned from light yellow to white. This was reasonably due to the surface modification induced by the formation of bismuth subnitrate sites.
This phenomenon leads to the uncontrolled variation of Bi 2 O 3 adsorption properties, its extent depending on the acid concentration, and time of interaction, thus making the surface properties, and hence its adsorption properties, dependent on the chemistry of the solution to be treated. Similarly, defects could be also induced by the washing with concentrated alkali, promoting the formation of hydroxyl functions on the surface of the material. An original defectiveness structure of oxygen sites in Bi 2 O 3 reticule further exploited the surface tailoring as reported by Gökagaç et al. [54].The surface modification of Bi 2 O 3 is reasonably accounted for by the adsorption performances after several cycles.
For both As(III) and As(V) species, the adsorption process took place very fast in the first 4 and 2 h, respectively, reaching a stable constant value after 16h. The initial fast adsorption of arsenic species might be attributed to the fine structures of Bi 2 O 3 combined with the surface modifications related to the disappearance of Bi(V) sites, supporting the hypothesis that they are related to a defective structure and not to a well defined phase.
The XPS spectra acquired after the fourth adsorption cycle (Figure 8) showed an appreciable simplification of bismuth and oxygen related signals. This could be one explanation for the general decrement in the adsorption of As(III) and As(V) with each regeneration cycle (Figure 7). However, the accumulation of As adsorbed may be attributed to the exposure of new binding sites with each cycle of desorption with NaOH and washing with KNO 3 . After both As(V) and As (III) adsorption, Bi 4f7/2 and Bi 4f5/2 were composed by only one component centered at 157.3 eV and 164.1 eV, respectively, while oxygen signal showed two components at 529. 3 eV and 528.1 eV. Curiously, the signals of adsorbed As(V) and As(III) were both composed of two components centered at 44.7 eV and 43.0 eV, corresponding to As(V) and As(III), respectively [53]. The redox of arsenic compounds during adsorption process is generally due to electrochemical active sites as in the case of Fe(II) or Mn related compounds [55,56], but in the case of Bi2O3 could be imputable to oxidative power of Bi(V) sites embedded into Bi2O3 [57] leading to oxidation of As(V).
This mechanism underwent a progressive deactivation caused by the high instability of Bi(V) sites together with the irreversibility of its reduction, leading to the progressive decrement of adsorption performances.
The combination of surface instability and defects insertion could represent a boost for the application of B2O3 as tool for treating high metal contaminated water wastestreams even if it is required a fine tuning for improving its duration.

Conclusions
Bi2O3 was synthesized by a solid-state reaction method that induced the formation of Bi(V) surface modifications in the defective structure of a highly reactive material. This new material was successfully tested for the removal of arsenite and arsenate from aqueous solutions. Both As species adsorbed rapidly, reaching equilibrium within 2 h for As(III) and 4 h for As(V), and the whole kinetics were well described by the pseudo- After both As(V) and As (III) adsorption, Bi 4f 7/2 and Bi 4f 5/2 were composed by only one component centered at 157.3 eV and 164.1 eV, respectively, while oxygen signal showed two components at 529. 3 eV and 528.1 eV. Curiously, the signals of adsorbed As(V) and As(III) were both composed of two components centered at 44.7 eV and 43.0 eV, corresponding to As(V) and As(III), respectively [53]. The redox of arsenic compounds during adsorption process is generally due to electrochemical active sites as in the case of Fe(II) or Mn related compounds [55,56], but in the case of Bi 2 O 3 could be imputable to oxidative power of Bi(V) sites embedded into Bi 2 O 3 [57] leading to oxidation of As(V).
This mechanism underwent a progressive deactivation caused by the high instability of Bi(V) sites together with the irreversibility of its reduction, leading to the progressive decrement of adsorption performances.
The combination of surface instability and defects insertion could represent a boost for the application of B 2 O 3 as tool for treating high metal contaminated water waste-streams even if it is required a fine tuning for improving its duration.

Conclusions
Bi 2 O 3 was synthesized by a solid-state reaction method that induced the formation of Bi(V) surface modifications in the defective structure of a highly reactive material. This new material was successfully tested for the removal of arsenite and arsenate from aqueous solutions. Both As species adsorbed rapidly, reaching equilibrium within 2 h for As(III) and 4 h for As(V), and the whole kinetics were well described by the pseudo-second order model, indicating that the rate-limiting step for the adsorption of both As species may be controlled by the same mechanism. The defective Bi 2 O 3 adsorbed up to 33.1 and 31.6 µmol m −2 As(III) and As(V), respectively, and the adsorption isotherms were best described by the two-sites Langmuir model. We showed that As adsorption onto Bi 2 O 3 is not hampered in the presence of chloride, nitrate, and sulfate but is moderately decreased by the presence of silicate and at high ionic strength. However, phosphate efficiently competed with both As species, impeding their removal from solution, which must be taken into consideration when using Bi-based substrates for the treatment of As-contaminated water or wastewater. This adsorbent can be regenerated and reutilized several times. It was still able to remove As from the solution after four cycles of regeneration, even if partial deactivation occurred. The aforementioned characteristics proved the suitability of Bi 2 O 3 as a reliable and effective adsorbent for As removal from water. The possibility to adsorb both As(III) and As(V) in comparable amounts, with similar kinetics and with no redox pre-treatments, on a substrate that can be regenerated and reused makes Bi 2 O 3 a promising and innovative material for the treatment of As-contaminated water.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/toxics9070158/s1, Figure S1. Effect of initial pH on the adsorption of As(III) and As(V) onto Bi 2 O 3 . Experimental conditions: initial As(III) and As(V) concentration-40 mg As L −1 ; adsorbent dose-4 g L −1 ; temperature-25 ± 3 • C; time-24 h, Figure S2. N 2 adsorption-desorption isotherms of synthesized Bi 2 O 3 , Figure S3. XPS spectra of Bi 2 O 3 after washing with HNO 3 0.1 M, Figure S4. Linear Langmuir (a) and Freundlich (b) isotherms for As(III) and As(V) adsorption onto Bi 2 O 3 (4 g/L) at pH 8 and 7 respectively, and 25 ± 3 • C, Table S1. Linear and non-linear forms of the Langmuir and Freundlich models, and to the two site Langmuir model used to fit the adsorption of As(III) or As(V) on Bi 2 O 3 , Table S2. Pseudo-first-order, pseudo-second-order, and intraparticle diffusion models used to fit the kinetic data of As(III) and As(V) adsorption on Bi 2 O 3 .

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to institutional and national data sharing restrictions.

Conflicts of Interest:
The authors declare no conflict of interest.