Exploring the Mechanisms of Selectivity for Metal(loid) Oxo-anion Removal During Water Treatment: a Review of Common Competing Oxo-anions and Tools for Quantifying Selective Adsorption

Abstract

Development of novel adsorbents often neglect the competitive adsorption between co-occurring oxo-anions, over-estimating realistic pollutant removal potentials, and overlook the need to improve selectivity of materials. This critical review focuses on adsorptive competition between commonly co-occurring oxo-anions in water and mechanistic approaches for the design and development of selective adsorbents. Five “target” metal(loid) oxo-anion pollutants (arsenate, arsenite, selenate, selenite, and chromate) were selected for study. Five “competing” co-occurring oxo-anions (phosphate, sulfate, bicarbonate, silicate, and nitrate) were selected due to their potential to compete with target oxo-anions for sorption sites resulting in decreased removal of the target oxo-anions. First, a comprehensive review of competition between target and competitor oxo-anions to sorb on commonly used, non-selective, metal hydr(oxide) materials is presented and the strength of competition between each target and competitive oxo-anion pair is classified. This is followed by a critical discussion of the different equations and models used to quantify selectivity. Next, five mechanisms that have been successfully utilized in the development of selective adsorbents are reviewed: variation in surface complexation, Lewis acid/base hardness, steric hindrance, electrostatic interactions, and hydrophobic interactions. For each mechanism, the oxo-anions, both target and competitors, are ranked in terms of adsorptive attraction and technologies that exploit this mechanism are reviewed. Third, given the significant effort to evaluate these systems empirically, the potential to use computational quantum techniques, such as density functional theory (DFT), for modeling and prediction is explored. Finally, areas within the field of selective adsorption requiring further research are detailed with guidance on priorities for screening and defining selective adsorbents.

1. Introduction

Exposure to toxic metal(loid) oxo-anions, such as arsenic, selenium, and chromium, through contaminated groundwater-sourced drinking water poses a threat to human and environmental health.¹ The scope of this challenge is large with approximately 40% of the United States population served by groundwater-sourced drinking water, through public supply or self-supplied (primarily from private domestic wells) systems.^2,3 Globally, arsenic contamination of drinking water has been reported in more than 70 countries, mainly in South and Southeastern Asia.⁴

Of the traditional water treatment technologies to remove oxo-anions, adsorbent processes are an attractive option for smaller-sized communities, which face high rates of water quality violations,⁵ while for larger-sized communities it is often more cost-effective to use chemical coagulants (e.g., iron, aluminum) that form floc and adsorb pollutants but form sludges that must be dewatered prior to landfill disposal. Advantages of adsorbents include their generally high removal efficiencies, feasible logistic use in decentralized field conditions, and regeneration potential.^1,5–9 However the low specificity of traditional non-selective adsorbent materials can lead to less effective removal of the target contaminants, contributing to failure in achieving regulatory limits.¹⁰ As many oxo-anions share similar chemical structure and behavior, their adsorptive mechanisms also tend to be alike, thus promoting competition for adsorption sites.¹¹ Additionally, lower concentrations of contaminants such as arsenic, selenium, and chromium, compared to background ions that compete for adsorption binding sites can result in decreased removal efficiency due to adsorption sites becoming saturated with competitive ions rather than target contaminants.^1,11

Selective adsorption, or the favorable removal of a target contaminant over competing and co-occurring background ions, overcomes competition by promoting target contaminant adsorption and/or hindering background oxo-anion adsorption.¹² Selective adsorbent material properties (surface area, pore size, surface charge, energy density of binding sites) can be developed by exploiting chemical behavioral differences, such as complexation strength, Lewis acid/base hardness, hydrophobicity, polarizability, size, and shape between target contaminants and competitive ions.¹¹ Depending the strength of competition, which is typically controlled by the degree of similarity in binding mechanisms and chemical structure,^13–16 multiple mechanisms may be required to realize selectivity.¹⁷ Likewise, competitive adsorption between bulk natural organic matter and target synthetic organic pollutants on activated carbon or other porous adsorbents has been well studied and numerous mechanistic and empirical models developed to aid in predictive pollutant removals or improvements in material designs.^18–21 However, there has been less focus on selectivity and competitive adsorption for co-occurring oxo-anions on metal oxide adsorbents, despite their widespread use and development for removing metallic oxo-anions from drinking waters.^1,22,23

Interference from competing ions greatly reduces removal effectiveness of the target contaminant increasing the requirements for sorbent material, and subsequently costs; as such, a tailored approach to water treatment that considers selectivity can increase material and resource efficiency.^24,25 Selectivity also results in a waste stream of reduced complexity due to the singular removal of a contamiant.²⁶ This could, in turn, decrease the energy and costs required for proper disposal.

In this critical review, we will classify the strength of competition between target/competitive oxo-anion pairs, review equations used to quantify selectivity, and discuss four mechanisms that have been successfully utilized in the development of selective adsorbents: variation in surface complexation, Lewis acid/base hardness, steric hindrance, and electrostatic interactions (Figure 1). For each mechanism, the oxo-anions are ranked in terms of adsorptive attraction and technologies that exploit this mechanism are discussed. Next, the potential to use computational quantum techniques to model and predict selective adsorption is explored. Finally, areas within the field of selective adsorption that require further research are detailed with guidance on priorities for designing, screening, and assessing selective adsorbents.

Figure 1.

Mechanisms for the selective adsorption of target oxo-anions over competitor oxo-anions. Green indicates target oxo-anion and red the competitor oxo-anion(s).

2. Target Oxo-anions and Competitors

This review will focus on three target drinking water contaminants of current interest to human health and the environment – arsenic, selenium, and chromium (Table 1) – and their most competitive oxo-anions – phosphate, sulfate, silicate, bicarbonate and nitrate (Table 1) under typical environmental conditions (pH ~6). This review will not focus on OH⁻, or pH, as a competitive anion since this circumstance can be adjusted during treatment conditions. Rather, this review is aimed at the inherent properties of sorbent materials that can be intentionally designed. It is important to note that the classifications of “target” and “competitor” are flexible as some of the competitor oxo-anions could also be considered “targets”. For example, phosphate and nitrate are also pollutants of environmental concern due to their role in eutrophication.^27,28 Additionally, this review will focus solely on targeting toxic metal(loid) oxo-anions. There are other contaminants of concern, such as uranyl and perchlorate, which warrant development of selective adsorption technologies; however, since uranyl is an oxo-cation and perchlorate is neither a metal nor a metalloid oxo-anion, these two contaminants fall outside the scope of this review.

Table 1.

Target and Competitor Oxo-anion pKa’s, Geometry, and Common Occurrence Range

Oxo-anion	pKa’s	Protonated Structure at pH 6	Oxo-anion Geometry	Common Occurrence Concentration Range in Groundwater	Refs.
Target Oxo-anions
Arsenate As(V) H2AsO4−	pKa1 = 2.2 pKa2 = 7.0 pKa3 = 11.6		Tetrahedral	Background: < 0.01 mg/L Affected waters: 0.08 to 5 mg/ L	29–31
Arsenite As(III) H3AsO3	pKa1 = 9.2		Trigonal pyramid	Background: < 0.01 mg/L Affected waters: 0.01 to 5 mg/ L	30,32
Selenate Se(VI) SeO42-	pKa1 = −3.0 pKa2 = 1.7		Tetrahedral	Background: < 2×10−5 mg/L Affected waters: 0.14 to 1.4 mg/L	33–35
Selenite Se(VI) HSeO3−	pKa1 = 2.6 pKa2 = 8.2		Trigonal pyramid	Background: < 2×10−5 mg/L Affected waters: 0.14 to 1.4 mg/L	33–35
Chromate Cr(VI) HCrO4−	pKa1 = −0.98 pKa2 = 6.5		Tetrahedral	Background: 0.002 to 0.014 mg/L Affected waters: 0.05 to 0.43 mg/L	36,37
Competitor Oxo-anions
Phosphate H2PO4−	pKa1 = 2.1 pKa2 = 7.2 pKa3 = 12.7		Tetrahedral	Background (total phosphorus): 0.05 to 2 mg/L	29,38–40
Sulfate SO42-	pKa1 = −3.0 pKa2 = 2.0		Tetrahedral	Background: 1 to 250 mg/L	41–43
Silicate H4SiO4	pKa1 = 9.8		Tetrahedral	Background: 1 to 30 mg/L	42 44
Bicarbonate HCO3−	pKa1 = 6 pKa2 = 10.3		Trigonal planar	Background: 150 to 450 mg/L	42,45
Nitrate NO3−	pKa1 = −1.3		Trigonal planar	Background: 0.2 to 10 mg/L	39,41,46

Arsenic, selenium, and chromium represent typical oxo-anion contaminants of concern with their multiple relevant redox states and known toxicity concerns. Arsenic frequently occurs as an oxo-anion in either +III (arsenite) or +V (arsenate) oxidation states, depending upon the redox conditions and dissolved oxygen concentration of a water.¹ Arsenite is more difficult to adsorb than arsenate as a result of its neutral charge in environmental conditions.^1,47,48 The WHO regulatory guideline for total arsenic concentration in drinking water is 10 ppb.³¹ Selenium exists in the environment predominantly as selenite (Se (IV)) and selenate (Se (VI)). Selenate is more difficult to remove due to its high solubility and weaker complexation with adsorbents.^49–52 While selenium is an essential nutrient at low doses, it is a toxin and potentially carcinogenic at higher doses.⁵³ The regulatory drinking water guideline set by the WHO for total selenium is 40 ppb, but because selenium also impacts aquatic ecosystems – its removal is often required from a diverse array of water sources.³¹ Chromium is predominantly found in the environment as Cr(III) (Cr₂O₃) or Cr(VI) (HCrO₄⁻). While Cr(III) has relatively low toxicity and is insoluble in most environmentally relevant conditions, Cr(VI) is extremely toxic and highly mobile.^54,55 The WHO regulatory guideline for the total level of chromium (Cr(III) and Cr(VI)) allowed in drinking water is 50 ppb.³¹

While reductions in removal efficiency of target oxo-anions in the presence of competitive ions is well-known and varies by adsorbent material (Table 2), studies do not consistently report on strong competition pairs or effectively assess selectivity. For example, phosphate and silicate are generally strong competitors and should be considered priorities in designs of selective adsorbent. Bicarbonate competes moderately to weakly, never decreasing removal by more than 50%. The strength of competition by sulfate varies as it weakly competes with arsenate, arsenite, and selenite, but is a strong to moderate competitor for selenate and chromate. Nitrate is generally a weak competitor. Further, most studies examine the individual effects between target and competitive oxo-anions, but do not systematically consider the most relevant competition pairs. Additionally, under environmental conditions, multiple competitive oxo-anions will frequently be present, and the presence of more than one competitive oxo-anion tends to further decrease removal efficiency compared to singular competitors.⁵⁶

Table 2.

Strength of competitive adsorption on traditional adsorptive media by oxo-anions

Target	Competitor	Similar traits	Dissimilar traits	Adsorptive Media	Approximate Reduction in Removal Efficiency
Arsenate H2AsO4−	Phosphate H2PO4−	pKa,57,58LBH,59,60 geometry14,58	Size*,11,61,62 CD,11,57,63 hydration11,64,65	Fe oxide, hematite	50–70%66,67
				Ferrihydrite	90%68
				Goethite	32%69
				Fe-Mn oxide	53%70
				Fe-Ce oxide	47%71
				Fe-Ni oxide	40%72
				Zr-Fe hydroxide	57%73
				Al oxide	80–83%14,74
				Al-Mn oxide	44%75
				Nano-Ti oxide	10%76
				Cu oxide	51%77
	Sulfate SO42-	Geometry78	LBH,24,60,79 pKa, size*,11,61,80 CD, hydration	Fe (hydr)oxide, hematite	2–14%14,66,81
				Ferrihydrite	0%68
				Goethite	2%69
				Fe-Ni oxide	7%72
				Zr-Fe hydroxide	7%73
				Al oxide	15–30%14,74
				Al-Mn oxide	46%75
				Cu oxide	0%77
	Bicarbonate HCO3−	pKa, LBH59,60	Geometry,78 size*,11,61,62 CD, hydration	Fe oxide	13–64%66,67
				Fe-Mn oxide	12%70
				Fe-Ni oxide	20%72
				Zr-Fe hydroxide	33%73
				Al oxide	50%74
				Al-Mn oxide	31%75
	Silicate H4SiO4	Geometry78	pKa, LBH,60 size*,11,61 CD, hydration	Fe oxide	77–83%14,66
				Ferrihydrite	93%82
				Goethite	35%69
				Fe-Mn oxide	48%70
				Fe-Ni oxide	30%72
				Zr-Fe hydroxide	40%73
				Al oxide	63–86%14,74
				Al-Mn oxide	27%75
				Cu oxide	1%77
	Nitrate NO3−		pKa, LBH,60 CD,63 geometry,78 size*,11,61,62 hydration	Fe oxide	2%14
				Zr-Fe hydroxide	0%73
				Al oxide	0%14
Arsenite H3AsO3	Phosphate H2PO4−	Size*	pKa,83 LBH,60,84–86 geometry,78 CD, hydration	Fe hydroxide	71%56
				Ferrihydrite	62–75%68,87
				Goethite	70%69
				Fe-Mn oxide	48%−70%75,88
				Fe-Ni oxide	56%72
				Zr-Fe hydroxide	50%73
				Al-Mn oxide	76%75
				Cu oxide	44%77
	Sulfate SO42-	Size*	pKa, hydration,65,89 LBH,59,60,85,86 geometry,78 CD	Fe hyroxide	50%81
				Ferrihydrite	7–15%68,87
				Goethite	30%69
				Fe-Mn oxide	2%88
				Fe-Ni oxide	5%72
				Zr-Fe hydroxide	10%73
				Al-Mn oxide	16%75
				Cu oxide	7%77
	Bicarbonate HCO3−		pKa, LBH,59,60,85,86 geometry, size*,62 CD, hydration	Fe-Mn oxide	3–23%70,88
				Fe-Ni oxide	11%72
				Zr-Fe hydroxide	18%73
				Al-Mn oxide	24%75
	Silicate H4SiO4		pKa, LBH,59,60 geometry, size*62	Goethite	25%69
				Ferrihydrite	90%82
				Fe-Mn oxide	38%70
				Fe-Ni oxide	22%72
				Zr-Fe hydroxide	26%73
				Al-Mn oxide	34%75
				Cu oxide	6%77
	Nitrate NO3−		pKa, LBH,59,60,85,86 size*,62 geometry, CD, hydration	Zr-Fe hydroxide	0%73
Selenite HSeO3−	Phosphate H2PO4−	pKa, LBH,59 Size*, CD, hydration	Geometry	Nano-hematite	82%90
				Fe oxyhydroxide	83%91
				Fe-Mn (hydr)oxide	20–39%92,93
				Nano-Mn oxide	63%94
	Sulfate SO42-	LBH,59,95 size*62,80	pKa, CD, hydration,65,89, geometry	Fe oxide	6%90
				Fe oxyhydroxide	1%91
				Fe-Mn (hydr)oxide	0–3%92,93
				Nano-Mn oxide	30%94
	Bicarbonate HCO3−	pKa, LBH,59,95 geometry	Size*,62,80 CD, hydration	Nano-hematite	20%90
				Fe-Mn hydroxide	2%92
	Silicate H4SiO4		pKa, geometry, size*,62 CD, hydration	Nano-hematite	71%90
				Fe oxyhydroxide	39%91
				Fe-Mn hydroxide	28%92
	Nitrate NO3−	LBH,59,95 molecular geometry	pKa, size*62, CD, hydration	Nano-hematite	1%90
				Fe-Mn oxide	0%93
				Nano-Mn oxide	8%94
Selenate SeO42-	Phosphate H2PO4−	LBH,59 geometry	pKa, size*,11,62 CD, hydration	Fe-Mn oxide	85%93
				Nano-Mn oxide	84%94
				Nano-Mg oxide	20%96
	Sulfate SO42-	pKa, geometry97,98	LBH,16,60,99,100 size*,62,80 CD, hydration	Goethite	50%101
				Schwertmannite	70%102
				Zero valent Fe	63%93
				Basaluminite	52%102
				Nano-Mn oxide	84%94
				Nano-Mg oxide	10%96
	Nitrate NO3−	pKa	Geometry, size*,62 CD, hydration	Fe-Mn oxide	0%93
				Nano-Mn oxide	60%94
Chromate HCrO4−	Phosphate H2PO4−	pKa, hydration,65,103 size*,11,61,62,103 CD, geometry	LBH60	Fe-Ni oxide	30%104
				Al hydroxide	57%105
				Mn oxide	64%106
				Fe-Al hydroxide	80%107
	Sulfate SO42-	LBH,60,108 geometry, size*61,62,80	pKa,109 CD, hydration65,110	Fe-Ni oxide	2%104
				Fe-Mn oxide	8%111
				Al oxide	75%109
				Mn oxide	30%106
	Bicarbonate HCO3−	pKa	LBH,60 geometry, size*,61,62 CD, hydration	Mn oxide	42%106
	Silicate H4SiO4	Geometry	pKa, LBH,60 size,61,62 CD, hydration	Fe-Mn oxide	27%111
	Nitrate NO3−		pKa, LBH,60 geometry, size*,61,62 CD, hydration	Fe-Mn oxide	7%111
				Mn oxide	0%106

Note: LBH is Lewis Base Hardness, CD is charge density,

^*calculated as reported in Section 5 and Table 5. % removal reported at pH ~6.

3. Quantification of Selectivity

The need for a standardized quantitative definition of selective adsorption is critical in order to appropriately compare functional performance of commercially available or novel adsorbents. An effective selectivity metric should capture the antagonistic competition between the target and competitor oxo-anions as well as overall performance removal. This requires quantifying the adsorption affinity and removal efficiency for both target and competitor oxo-anions. The metric should also have a clear numerical threshold for selectivity. However, none of the selectivity metrics most commonly reported in the literature (Table 3) comprehensively capture all of these sorbent attributes.

Table 3.

Commonly Used Metrics to Quantify Selectivity

Eqn #	Relationship	Definition of Terms	Ref(s)
1	$\%R=\frac{(C_0-C_e)}{C_0}x100\%$	%R is percent removal, C0 is the initial concentration of the adsorbate (ppm) Ce is the equilibrium concentration of the adsorbate (ppm)	15,112
2	$q=\frac{v}{m}\left(C_0-C_e\right)$	$q$ is the equilibrium adsorption capacity for a target contaminant (mg/g), $v$ is the volume of adsorbate solution (mL), and $m$ is the mass of adsorbent (mg)	110
3	$\text{\%I}=q_t-q_{t+c}$	%I is percent interference, qt is the adsorption capacity for the target contaminant and qt+c is the adsorption capacity for the target contaminant with a competitive oxo-anion present.	113
4	$\mathit{\text{Adsorption capacity ratio}}\%=\frac{q_t}{q_{t+c}}x100\%$		114
5	${\propto }_{t/c}=\frac{q_t{C}_{e,c}}{q_c{C}_{e,t}}$	${\propto }_{t/c}$ is the binary separation factor for the adsorption of target oxoanion, t, in the presence of competitive oxo-anion, c, $C_{e,c}$ is the equilibrium concentration of the competitive oxoanion in the mixed ion system, and $C_{e,t}$ is the equilibrium concentration of the target oxoanion in the mixed ion system.	11,24,29,79,110,114–119
6	$K_d=\frac{C_0-C_e}{C_e}x\frac{v}{m}$	Kd is the distribution coefficient, C0 the initial concentration of the target oxo-anion, and Ce the equilibrium concentration	120–123
7	${\beta }_t=\frac{K_{d,t}}{K_{d,c}}$	${\beta }_t$ is the selectivity coefficient for the target anion, $K_{d,t}is$ the distribution coefficient for the target ion with a competitor present, and $K_{d,c}$ is the distribution coefficient for the competitor with the target ion present	124–127

The most commonly used metric for selectivity assesses changes in removal performance of the target contaminant when a competitive oxo-anion is introduced, defined as % removal (%R) (Table 3, Eqn.1). By comparing %R of the target contaminant with/without a competitive ion, if and how much a competitor decreases adsorptive performance for the target contaminant can be quantified. Some studies have defined a “tolerance” limit on an acceptable change in %R as a result of introduction of a competitive ion (e.g., the highest concentration of a competitive ion that can be introduced to the system while still allowing >90% recovery of the target oxo-anion).^128–130

In this review, %R was chosen as the metric for quantifying the competitive effect of oxo-anions (Table 2) as it was the most widely reported in the literature or could be readily calculated from the reported data. However, %R does not adequately quantify several key aspects for developing an effective selective adsorbent. Critically, %R does not evaluate how efficient the adsorbent is at removing the competitive oxo-anion. In order for an adsorbent to be truly selective, it must not only be effective in removing the target contaminant, but it also must be ineffective in removing competitive ions. This metric also does not consider the mass of adsorbent required to achieve selective removal, a key aspect of assessing the efficiency of an adsorbent, or have a threshold limit for selectivity.

Another commonly used metric is adsorption capacity, q. (Table 3, Eqn. 2). Similar to %R, strength of competition between target and competitive ions can be quantified by assessing adsorption capacity for the target ion with/without a competitive oxo-anion present, referred to as % interference (%I) (Table 3, Eqn. 3).^{113,131–133} Li et al. quantified the change in q as an adsorption capacity ratio (Table 3, Eqn. 4).¹¹⁴ An alternative is to model competitive adsorption using Ideal Adsorption Isotherm Theory (IAST). This approach, which takes parameters derived from two sets of Freundlich isotherm models, is currently more widely used for sorption of organics to porous adsorbents such as granular activated carbon.^134,135 However, IAST could be easily applied to selective and competitive oxo-anion sorption.¹³⁶

Measuring q, is slightly more robust than %R given that it does assess the mass of adsorbent required to remove the target contaminant, and thus the efficiency of the material. However, this metric is not capable of evaluating the strength of adsorption towards both target and competitor simultaneously, and thus, is inadequate in effectively characterizing selectivity. Both %I and the adsorption capacity ratio are variations on using q as a selectivity metric that attempt to mathematically define the strength of competition between the target and competitor. None of these selectivity metrics (e.g., %I, the adsorption capacity ratio, %R and q) provide a defined threshold value above which selectivity is demonstrated For example, an adsorbent could report >90% removal of a contaminant, but fail to decrease contaminant concentrations below MCLs depending on the starting concentration or mass of sorbent material.

An alternative metric that has emerged is the binary separation factor, sometimes referred to as the selectivity factor for target (t) and competitor (c) sorbates, ${\propto }_{t/c}$ (Table 3, Eqn. 5).^{11,24,29,79,110,114–119} A ${\propto }_{t/c}>1$ indicates that the target contaminant is preferred and the adsorbent is selective. ${\propto }_{t/c}$ is a particularly robust selectivity metric as it compares adsorption affinity towards both the target and competitive oxo-anions within the same system while also measuring efficiency of the material. However, ${\propto }_{t/c}$ does not effectively capture the ability of a sorbent to realize compliance with regulatory limits.

Another option is the distribution coefficient, K_d (Table 3, Eqn. 6).^120–123 The decrease in K_d of the target anion upon introduction of a competitive ion can be used to quantify selectivity towards the target anion.^120–123 However, similar to most of the above metrics (all except for ${\propto }_{t/c}$), K_d does not define a threshold limit above or below which selectivity has been demonstrated and it also alone does not evaluate removal of target and contaminant oxo-anions within the same system. Therefore, K_d alone cannot be used to effectively assess selectivity.

Some studies will also analyze K_d of the competitive ion with/without the target anion present. They will then compare the ratio of K_d of the target and competitor ions in the mixed oxo-anion system, defined as the selectivity coefficient, ${\beta }_t$ (Table 3, Eqn. 7).^124–127 ${\beta }_t$ is an effective selectivity metric as it quantifies the adsorption affinity for both oxo-anions, not just the target contaminant and additionally has a clear definition of selectivity where ${\beta }_t$ values above 1 demonstrate selectivity towards the target competitor. However, ${\beta }_t$ is mathematically equivalent to ${\propto }_{t/c}$, which was not previously recognized in the literature, and needs to be clarified in order to avoid confusion.

In order to compare the metrics listed in Table 3, a test system (Table S1) was evaluated and selectivity quantified through each of the equations discussed above. The results of this are reported in Table S2. Overall, ${\propto }_{t/c}$ and ${\beta }_t$ were found to be the most effective metrics to quantify selectivity due to their clear threshold limit for selectivity and that they evaluate adsorption affinity towards both the target and competitive oxo-anion.

While using ${\propto }_{t/c}$ or ${\beta }_t$ is a much more robust metric than %R or q, additional challenges remain. For example, there not yet a clear definition for the systems conditions to quantify ${\propto }_{t/c}$ or ${\beta }_t$. Initial concentrations of both the target and competitor oxo-anion must be agreed upon in addition to other critical considerations such as pH or presence of additional background electrolytes. To enable rigorous comparisons across sorbent materials, selectivity should be evaluated in standardized testing waters (e.g., NSF 53 Challenge Water).^137,138

An alternative approach to quantify selectivity is equilibrium surface complexation models (SCMs) that consider competition for binding sites in multi-ion systems such as the diffuse layer model (DLM) and charge distribution multi-site complexation (CD-MUSIC).^{136,139–141} Once parameterized with a subset of data, SCMs are able to predict the nature of competition or selectivity between multiple oxo-anions in a system in a wide range of conditions beyond those studied in limited isotherm tests.^69,136 However, SCMs require more skill to parameterize and information than more traditional quantitative methods such as those reported in Table 3 (e.g., binding site density, strength of binding sites, zero point of charge for adsorbent, etc.),¹³⁶ they lack a clear quantitative definition for selectivity, but they also are mechanistic in nature and could compliment techniques such as DFT.

Also lacking from the above metrics is consideration of adsorptive kinetics which can have a strong influence on competitive adsorption.^142,143 For example, silicate adsorbs more slowly than As(V) due to its surface polymerization.¹⁴² Thus, in a competitive system containing As(V) and silicate, the slower kinetics of silicate can lead to a decrease in both the kinetics of As(V) adsorption, as well as the quantity of arsenic adsorbed.¹⁴² Models such as the homogeneous diffusion model (HDSM) can be used to better understand differences in oxo-anion diffusion rates on various adsorptive materials by taking into account the concentration of solute in the bulk phase and the adsorption density gradient within individual adsorbent particles.^136,144 This knowledge could then be exploited in the design of selective adsorbents that promote faster diffusion of target oxo-anions over competitive oxo-anions, or that minimize the detrimental effects of competitive oxo-anions with slower kinetics.

4. Selectivity due to Differences in Surface Complexation

4.1. Surface Complexation Geometries

Metal (hydr)oxides form both inner-sphere and outer-sphere complexes with oxo-anions.^145,146 Outer-sphere complexes form between the fully coordinated shells of oxo-anions and metal (hydr)oxides.^147,148 The weak electrostatic attraction between the protonated surface hydroxyl groups (OH₂⁺) and the negatively charged oxo-anions is often facilitated through extensive hydrogen bond networks (Figure 2).¹⁴⁹ In comparison, the much stronger inner-sphere complexes form as a result of direct covalent bonding between the adsorbed oxo-anion and the sorbate.^147,148 The oxo-anion and sorbate first form a weak outer-sphere complex, then a ligand exchange between the surface hydroxyl groups and oxo-anions occurs, forming a stronger inner-sphere complex.^149–151, Inner-sphere complexes are most commonly monodentate (one bond between oxo-anion and the sorbate surface) or bidentate (two bonds between oxo-anion and sorbate) (Figure 2). Bidentate complexes can be further categorized as either mononuclear or binuclear, corresponding to the formation of edge-sharing complexes with a single surface atom, or corner-sharing complexes with two adjacent terminal atoms, respectively.

Figure 2.

Common coordination complex geometries between metal (hydr)oxides and oxo-anions: (a.) monodentate mononuclear complex; (b.) bidentate mononuclear edge-sharing complex; (c.) bidentate binuclear corner-sharing complex; (d.) outer-sphere complex.

The most commonly studied metal (hydr)oxides are hematite (α-Fe₂O₃), TiO₂, Al₂O₃, and goethite (α-FeOOH). Iron oxides (hematite and goethite) have shown greater adsorption strength than their aluminum counterparts.^13,14,152 The increased reactivity of iron oxides, especially hematite, is theorized to be due to the presence of reactive and highly exchangeable hydroxyl groups on the surface.^153,154 Titanium dioxide performs comparably to iron oxides in the ability to sorb oxo-anions.

Strong inner-sphere complexes typically occur with metal oxides that readily exchange surface hydroxyl groups (e.g. α-Fe₂O₃), and particularly with oxo-anions that have a strong affinity for the metal (hydr)oxide surface (e.g. HSeO₃⁻ H₂PO₄⁻, H₂AsO₄).¹⁵⁵ These high-affinity oxo-anions often form strong bidentate complexes regardless of the metal (hydr)oxide composition.^156–160 In comparison, “intermediate” oxo-anions (e.g. SO₄^2-, HCO₃⁻, SeO₄^2-, H₃AsO₃, HCrO₄⁻) have a lower affinity for oxide surfaces and can participate in both inner and outer-sphere complexation depending on the metal (hydr)oxide surface.^{145,155,156,158,161–163} Lastly, weakly adsorbing oxo-anions (e.g. NO₃⁻) have a low affinity for the surface of metal oxides and thus primarily form outer-sphere complexes.^14,164

Exploiting the surface chemistry of metal (hydr)oxides and other sorbents to target contaminants based on their mechanism of complexation will make it possible to design efficient, and possibly selective, sorption systems to remove target oxo-anion contaminants by maximizing their interaction with metal (hydr)oxide surfaces.

4.2. Facet-Dependent Adsorption Mechanisms

Current research has begun to correlate surface chemistry to adsorptive function by studying the role of different metal (hydr)oxides crystal facets on sorption mechanisms and capacity.^157,165,166 Facets are classified by their Miller indices {hkl}, where the curly brackets indicate an equivalent family of planes. This notation derives from the coordinates of the plane’s interception on the x, y, and z axes.

Recent research has suggested that adsorption capacity can be improved by synthesizing metal (hydr)oxides with more reactive exposed facets, which could be beneficial in sorbing contaminants with high and intermediate affinities. Additionally, exposing facets that promote certain complexation types (e.g. monodentate, bidentate, outer-sphere) may make it possible to selectively sorb contaminants with matching preferred geometries (Table 4).

Table 4.

Facet Preference for Target and Competitive Oxo-anions

Oxo-anion	Adsorbent Material	High-Affinity Facet(s)	Complexation Type	Low-Affinity Facet(s)	Complexation Type	Solution Chemistry	Reference(s)
Target
H2AsO4−	TiO2 Anatase	{001}	Not Studied	{101}	Not studied	DFT calculation	166
	α-MnO2	{100}	Bidentate/monodentate, more stable as bidentate	{110}	Bidentate/ monodentate, more stable as bidentate	DFT calculation	169
H3AsO3	TiO2 Anatase	{001}	Not Studied	{101}	Not studied	DFT calculation	166
	α-MnO2	{100}	Bidentate/ monodentate, more stable as bidentate	{110}	Bidentate/ monodentate, more stable as bidentate	DFT calculation	169
HSeO3−	Nano-hematite (α-Fe2O3)	{012}	Bidentate mononuclear	{110}	Bidentate binuclear	0.01 M KCl pH = 3.5	162
HCrO4−	Nano-hematite (α-Fe2O3)	{110}	Bidentate binuclear	{001}	Monodentate mononuclear	0.01 M NaCl pH = 2	168
Competitor
SO42-	Hematite	{012}	Bidentate	{001}	Monodentate	pH = 1	170

The majority of research on facet-dependent adsorption mechanisms focuses on hematite (α-Fe₂O₃) because it is a strong sorbent, ubiquitous in the natural environment, and thermodynamically stable. Given that different crystal facets have different surface oxygen configurations and availabilities, synthesizing hematite to expose crystal facets with a high density of singly coordinated oxygens could promote strong sorption of intermediate and high affinity oxo-anions.¹⁶⁷ For example, Catalano et al. showed that selenite and arsenic form strong, bidentate inner-sphere complexes preferentially at the singly-coordinated surface oxygen despite the availability of doubly-coordinated oxygen sites.¹⁵⁷

This variation in surface chemistry can therefore promote different sorption mechanisms on different facets. For example, it has been shown that that both selenite and chromate form binuclear, bidentate complexes on the {110} facet of hematite. Further, selenite forms mononuclear, bidentate complex on the {012} facet¹⁶² while chromate sorbs via monodentate, mononuclear geometry on the {001} facet.^167,168 Notably, the {012} facet showed a higher adsorption capacity than the {110} facet for selenite while the {110} facet had higher adsorption capacity for chromate than the {001} facet.^162,167,168 The enhanced reactivity on the singly-coordinated oxygen to form inner-sphere bidentate geometries indicates potential for utilizing controlled-synthesis methods to make iron oxide particles with highly reactive, and potentially highly selective, facets. This may promote the formation of strong bidentate geometries for high- and intermediate-affinity oxo-anions and thus resulting in increased sorption capacity.

These results broadly suggest that the mechanism and strength of sorption is dependent on the crystal facet, and thus surface functional groups, of the adsorbent material. With limited experimental and computational studies analyzing this effect on a variety of metal oxides, facet-dependent sorption is an understudied area. This is an important gap in the pursuit of developing selective and efficient metal (hydr)oxide sorbents. With recent research suggesting that different exposed facets promote different sorption mechanisms, a robust understanding of preferred complexation geometry for different oxo-anions could make it possible to selectively promote adsorption of target oxo-anions over competitors (Table 4). Future work establishing a comprehensive correlation between the crystallographic structure and sorptive function of metal (hydr)oxides, will elucidate favorable sorbent qualities and aid in the design of selective sorbent materials.

5. Selectivity due to Steric Interactions

Steric hindrance as it relates to adsorption can be described as interference in adsorption due to the molecular structure (size and molecular geometry) of either the adsorbate or the adsorption site. In reviewing the literature there are numerous reports of oxo-anion size reported as ionic radius, hydrated ionic radius, and molecular volume.^{11,61,62,80,171} However, many of these values differ significantly between sources, or are not calculated at the correct degree of protonation for environmental conditions. Therefore, in order to directly compare between target and competitive oxo-anions, we calculated the molecular volume of both target and competitor oxo-anions for their degree of protonation in environmental conditions (Table 5, SI). These calculations resulted in the following ranking of oxo-anions listed in order of decreasing molecular volume: H₂AsO₄⁻~SeO₄^2->H₄SiO₄>HCrO₄⁻>H₂PO₄⁻>SO₄^2->H₃AsO₃>HSeO₃⁻>HCO₃⁻>NO₃⁻. The molecular volume of silicate was calculated assuming a single tetrahedral molecule; however, it is known to polymerize frequently in the environment, so may be present at much higher molecular volumes. Comparing molecular geometry, arsenate, selenate, chromate, phosphate, sulfate, and silicate all have tetrahedral structure, while arsenite and selenite have trigonal pyramidal structure, and nitrate and bicarbonate a trigonal planar structure (Tables 1 and 2). These differences in molecular volume and geometry can potentially be exploited in the development of sterically selective adsorbents.

Table 5.

Molecular volume and Laplacian bond order (LBO) of target and competitor oxo-anions

Oxo-anion	Volume		LBO
	Bohr3	Å3	X-O1	X-O2	X-O3	X-O4
Target
H2AsO4−	690.9	102.4	0.338	0.338	0.232*	0.233*
H3AsO3	602.6	89.3	-	0.228*	0.242*	0.231*
HSeO3−	598.7	88.7	0.341	0.206*	0.357	-
SeO42-	690.9	102.4	0.324	0.324	0.324	0.324
HCrO4−	656	97.2	0.836	0.845	0.829	0.384*
Competitor
H2PO4−	635.3	94.1	0.751	0.774	0.415*	0.415*
SO42-	623	92.3	0.860	0.860	0.860	0.860
NO3−	424	62.8	0.863	0.865	0.866	-
H4SiO4	658.8	97.6	0.379*	0.379*	0.379*	0.379*
HCO3−	474.6	70.3	1.078	1.033	0.447*	-

Note:

^*values for the O atoms additionally bonded to H atoms

Several efforts aimed at developing a sterically selective adsorbent based on materials that contain an interlayer space where larger or more geometrically bulky molecules may be excluded, such as layered double hydroxides (LDH) and functionalized clay.¹⁷² However, multiple studies based on this approach found that selectivity for target oxo-anions is not reliably developed. For example, Mg-Al LDHs selected (measured as %I) the following: H₂PO₄⁻>H₂AsO₄⁻>HCO₃⁻~SO₄^2->NO₃⁻.^173,174 Similarly, using Fe-functionalized Akadama clay, phosphate and sulfate were preferred (measured via K_d) over chromate.¹²⁰ These materials may not reliably develop steric selectivity as the adsorption sites are not arranged to specifically prefer a given molecular geometry or size, and the interlayer space can shrink and contract.⁶²

An alternative approach is through cationic MOFs (Table 6) where by controlling the pore size and rigidity of the MOF, steric selectivity can be developed.^17,98,175 The rigidity of the MOF is critical in avoiding free rearrangement to accommodate anions of various size and molecular geometry, rather than sterically exclude them.^17,98 This type of steric selectivity is particularly strong as it does not singularly favor smaller anions, instead taking into account both size and molecular geometry to develop a high degree of anion shape recognition.^17,98 Cai et al. developed a Fe(III)-MOF which selects for As(V) over sulfate, nitrate, and bicarbonate (measured as %R), but did not test against phosphate.¹⁷⁵ Selectivity for chromate over nitrate, sulfate, and bicarbonate has also been developed using Ag(I)-,¹⁷⁶ Co(II)-,¹⁷⁷ Cd(II)-,¹⁷⁸ Zn(II)-,¹⁷⁷ and Zr(VI)-^112,122 cationic MOFs (Table 6). MOFs that select for As(III), Se(IV) or Se(VI) over other oxo-anions have not yet been developed.

Table 6.

Adsorbents selective for the removal of target oxo-anions

Target Oxo-anion	Technology Type	Adsorption Technology		Observed Selectivity	Selectivity Metric(s)	Mechanism(s) of Selectivity	References
H2AsO4−	PLE		Cu(II)-PLE	As(V) ~ P > C > S	$\propto$	Electrostatic, HSAB	24,79,117,179
			Zr(VI)-PLE	As(V) ~ P > C ~ S	%R	Electrostatic, HSAB	16,99
			ZV-PLE	As(V) > S > N > P	%R	Hydrophobicity, sterics	180
			Cu(II)-chitosan	As(V) > P	$\propto$, %R	Electrostatic, HSAB	11,133,181
	AE		WB-AE	S > As(V) > N	%I	Electrostatic, HSAB	25
			WB-AE	P > As(V) > S > N	%I	Electrostatic, hydrophobicity	57
	MOF		Fe(III)-MOF	As(V) > S ~ N > C	%R	Sterics	175
			SM-Zr(VI)-MOF	P ~ As(V) > Si > S > N	TL	Electrostatic	182
	LDH		Mg-Al LDH	P > As(V) > C ~ S > N	%I	Electrostatic, sterics	173,174
H3AsO3	MIP		MIP	As(III) > P > S > N	Kd, $\beta$	Electrostatic, sterics	126
			MIP	S > As(III) > Si ~ P > N	Kd, $\beta$	Electrostatic, sterics	183
HSeO3−	PLE		Cu(II)-PLE	Se(IV) > S	$\propto$	Electrostatic, HSAB	117
			Cu(II)-chitosan	Se(IV) > P	$\propto$, %R	Electrostatics, sterics	11
	MMSi		MMSi	Se(IV) > S ~ N > P	%R	Electrostatics	184
SeO42-	HBC		Mg(II)-HBC	S > Se(VI) > C	$\Delta G^{\circ }$	Hydrophobicity, sterics, electrostatic	98
HCrO4−	PLE		ZVE-PLE	Cr(VI) > S > N > P	%R	Hydrophobicity, sterics	180
	AE		SB-AE	Cr(VI) > S > N	$\propto$	Hydrophobicity, sterics, electrostatic	110
	MOF		Zr(VI)-MOF	Cr(VI) > S > N	K d , %R	Electrostatic, HSAB, sterics	112,122
			Co(II)-MOF	Cr(VI) > S ~ N	K d	Electrostatic, sterics	177
			Zn(II)-MOF	Cr(VI) > S ~ N	K d	Electrostatic, sterics	177
			Cd(II)-MOF	Cr(VI) > S > N	%R	Electrostatic, sterics	178
			Ag(I)-MOF	Cr(VI) > N > C	%R	Electrostatic, sterics	176
	SM-clay		SM-clay	Cr(VI) > N > S −	qr	Sterics	172
	MIP		MIP	Cr(VI) > S ~ N	Kd, $\beta$	Sterics	127
			MIP	Cr(VI) > S > N	Kd, $\beta$	Sterics	185
			MIP	Cr(VI) > S > P > N	Kd, $\beta$	Sterics	186,187
			NMO-MIP	Cr(VI) > N > S > P	TL	Electrostatic, sterics	188
	SM-MCWNT		SM-MCWNT	Cr(VI) > C ~ P ~ N	%R	Electrostatic	189
	SM-activated C		SM-activated C	Cr(VI) > P > S > N	%I	Sterics	190

Note: PLE is polymeric ligand exchanger, HSAB is hard-soft acid base principle, ZV is zero valent iron, SB-AE is a strong base anion exchanger, WB-AE is weak base anion exchanger, MOF is metal organic framework, LDH is layered double hydroxide, SM is surfactant modified, and MIP is molecularly imprinted polymer, MMSi is modified mesoporous silica, HBC is hydrogen bonding capsule, MCWNT is multi-walled carbon nanotube. $\propto$ is binary separation factor, %R is % removal, %I is % interference, TL is tolerance limit, K_d is distribution coefficient, $\beta$ is the selectivity coefficient, $\Delta G^{°}$ is energy of crystallization, qr is ratio of q the adsorption capacity.

Molecularly imprinted polymers (MIP)s are also effective at developing steric selectivity (Table 6). Similar to MOFs, MIPs contain a pore structure that can recognize potentially and select for the target oxo-anion through the macrocyclic affect.¹²⁶ Specific recognition cavities are developed by adsorbing the target oxo-anion, then adding a crosslinking agent to rigidly set the MIP structure to be complementary to the target oxo-anion.^127,191 The target oxo-anion is then desorbed, leaving pores within the MIP with functional groups rigidly set complementary to the target oxo-anion shape, size and coordination geometry.^127,191 Fang et al. successfully developed a MIP which selects for (measured as K_d and $\beta$) As(III) over phosphate, sulfate, and nitrate.¹²⁶ This is a particularly compelling technology as phosphate is a strong competitor with arsenic (Table 2). However, this technology has not yet been evaluated for competition between As(V) and phosphate. MIPs have also been used to selectively adsorb (measured as K_d and $\beta$) Cr(VI) over phosphate, nitrate, and sulfate.^{127,185–187,190} Qi et al. combined nanometal oxides with MIPs to design a hybrid MIP-functionalized-nanoparticle that selects for (measured as TL) HCrO₄⁻>NO₃⁻>SO₄^2->H₂PO₄⁻.¹⁸⁸ Sterically selective MIPs have not yet been developed that target As(V) and selenium.

Of the three technologies reviewed that utilize steric hindrance to develop selectivity, MIPs are particularly promising since they successfully removed As(III) and Cr(VI) in the presence of phosphate, a strong adsorptive competitor (Table 2).^{126,185–188,190} MOFs are highly selective towards arsenate and chromate, but studies have not yet tested their selectivity in the presence of strong competitors such as silicate and phosphate.

6. Selectivity driven by the Pearson Hard Soft Acid Base Principle (HSAB)

According to the Pearson Hard Soft Acid Base Principle (HSAB), Lewis bases (electron donors), and Lewis acids (electron acceptors) can be divided into three categories: hard, intermediate, and soft.^59,192 A hard Lewis base is an electron donor with low polarizability, high pK_a, high negative charge density, and inaccessible high energy empty orbitals.^192,193 A hard Lewis acid is an electron acceptor with low polarizability, small size, high positive charge density, and no easily excited outer electrons; soft acids/bases have inverse traits, and borderline have intermediate traits.¹⁹² More stable complexes are formed between acids/bases of similar hardness than between acids/bases of dissimilar hardness.⁵⁹

Applying the HSAB principle to oxo-anion remediation, both target and competitive oxo-anions are Lewis bases, and metal sorption sites are Lewis acids. Target oxo-anions arsenate, selenate, selenite, and chromate are all hard Lewis bases, while arsenite is a borderline Lewis base.^59,194 The competitive oxanions phosphate, silicate, sulfate, nitrate, and bicarbonate are also hard Lewis bases.^59,60,194 Trivalent metals Al³⁺ and Fe³⁺ are hard Lewis acids, and divalent metals Cu²⁺, Zn²⁺, Ni²⁺, and Fe²⁺ are borderline Lewis acids.⁵⁹ While most target and competitive oxo-anions are classified as hard Lewis bases, the relative hardness varies within this general classification. There is not yet a comprehensive classification of oxo-anions ranked according to their relative Lewis base hardness; however synthesis of the available studies suggests the following ranking: silicate > phosphate ~ arsenate > bicarbonate > sulfate ~ chromate > nitrate > arsenite.^24,60,79

Differences in Lewis acid/base hardness were utilized in the design selective polymeric ligand exchangers (PLE)²⁹ and MOFs (Table 6). Henry et al. developed a Cu(II)-PLE that selected (measured by $\propto$) for arsenate and phosphate over sulfate;¹⁷⁹ however, it failed to separate the most difficult competitive pair, arsenate and phosphate. Similar Cu(II)-PLEs were selective (measured by $\propto$) for arsenate and selenite over weakly competing bicarbonate and sulfate; however, strongly competitive phosphate was not evaluated.^{24,79,117,195} Zr(IV)-PLEs selected for (measured by %R) arsenate and phosphate over bicarbonate and sulfate.^16,99 MOFs are less studied than PLEs for HSAB driven selective adsorption. However, Rapti et al. found that a Zr(IV) MOF selected (measured by K_d) for chromate in the presence of only weakly competitive sulfate and nitrate.¹²²

Overall, PLEs and MOFs were effective in their selectivity for arsenate, selenate, and chromate over only relatively weak competitors: bicarbonate, sulfate, and nitrate (Table 6). As silicate is a harder Lewis base than most target oxo-anions,⁶⁰ theoretically, HSAB-selectivity could be developed using a weaker Lewis acid adsorption site. Overall though, the HSAB principle must be combined with other mechanisms of selective sorption such as electrostatic preference or steric hinderance in order to separate strong sorbing competitors such as arsenate and phosphate, due to their similar Lewis base hardness.^11,17,64,133

7. Electrostatic Selectivity

Electrostatic intermolecular interactions are mediated by forces that can be broadly categorized into van der Waals forces, ion-dipole forces, and hydrogen bonding. Van der Waals forces can be further separated into dipole-dipole, dipole-induced dipole, and London dispersion forces.¹⁹⁶ The type of intermolecular interactions that occur are dependent on the polarity of both adsorbent and adsorbate. If both the adsorbent and adsorbate are polar molecules, they can participate in dipole-dipole interactions.¹⁹⁶ For example, at environmental pH, As(III) is neutrally charged as H₃AsO₃, and has a permanent dipole moment due to its trigonal pyramid structure, so can participate in dipole-dipole interactions with polar adsorbents. As most target contaminants are negatively charged at environmental pH, they will participate in ion-dipole forces where anions and polar adsorbents are electrostatically attracted to each other. In ion-dipole interactions, the smaller the oxo-anion, the higher the charge density (ratio of charge/volume) and the stronger the interaction with the dipole.^98,110,196 Therefore, the following charge density ranking is expected (from highest to lowest) based on the molecular volumes calculated in Table 5 and the protonation expected in environmental conditions: SO₄^2->SeO₄^2->NO₃⁻>HCO₃⁻>HSeO₃⁻>H₂PO₄⁻>HCrO₄⁻>H₂AsO₄⁻>H₃AsO₃>H₄SiO₄.

In order to systematically probe electrostatic differences between oxo-anions, we calculated the Laplacian bond order (LBO) values (Table 5) and Electron Localization Functions (ELF) for each of the target and competitive oxo-anions (Fig.S1–S4). LBO values measure the local charge concentration or depletion within an oxo-anion,¹⁹⁷ with positive LBO values corresponding to locally depleted electron density. This parameter correlates directly with the bond dissociation energy, bond polarity and bond vibrational energy. LBO varies greatly between all the oxo-anions analyzed, with the largest LBO values, and strongest bonds observed for HCO₃⁻ (Table 5). Generally, C-O, Cr-O, S-O and N-O bonds among the strongest, P-O bonds are moderately strong, and As-O and Se-O are weakest amongst the oxo-anions. These results enable development of electrostatically selective adsorbents as oxo-anions with higher LBO values are less polarizable.^11,133,181

Polarizability differences can be exploited in the development of selective adsorbents, particularly for targeting oxo-anions with relatively low charge density and high polarizability such as arsenate and selenite. For example, Yamani et al. used a Cu(II)-chitosan adsorbent to electrostatically select for (measured as $\propto$) arsenate and selenite in the presence of phosphate.¹¹ As Cu(II) binds to the amine groups of chitosan, the Cu(II)-chitosan complex gains cationic character that electrostatically attracts oxo-anions.^{11,24,29,79,133,181} Chitosan is a large polymeric molecule with many electrons, so would expected to be highly polarizable.¹⁹⁶ Therefore, arsenate and selenite, both of which are more polarizable than phosphate,^{11,57,133,181} will be more attracted to Cu(II)-chitosan bidentate complex due to their similar diffusion of charge.^11,133,181 Pincus et al. further developed this adsorbent, optimizing the loading of Cu(II) and n-TiO₂ to selectively adsorb (measured as %R) As(III) and As(V) over phosphate.^133,181 These adsorbents were notable as they are able to select for arsenate over its strongest adsorptive competitor, phosphate, by taking advantage of some of their few dissimilar traits, charge density and polarizability (Table 2).^{11,57,63,133,181}

Other studies have focused on developing adsorbents that select for oxo-anions of higher charge density. Many of these technologies use amine,¹⁹⁸ amide,¹⁹⁸ urea,⁹⁸ and pyridinium⁹⁸ groups among others to promote electrostatic interactions and hydrogen bonding between oxo-anions of higher charge density and the adsorbent. Hydrogen bonding is more directional than Coulombic and van der Waals forces, meaning it can potentially be better exploited for selective adsorption.^17,64 Generally, smaller anions interact more strongly as hydrogen bond acceptors due to their relatively higher charge density.¹⁷ For example, Hu et al. found that for strong base anion exchangers, development of selectivity is dependent on both the polymer composition and the functional groups on the polymer.¹¹⁰ Strong base anion exchangers with smaller spacing between functional groups electrostatically prefer divalent ions with higher charge density compared to anion exchangers with wider spaced functional groups.¹¹⁰ Their study took place at pH >8 where chromate is divalent, and were successfully able to select for (measured by $\propto$) chromate over sulfate and nitrate using a hydrophobic strong base anion exchange resin with small spacing between amine functional groups.

An additional approach is developing 3D hydrogen bonding networks through ring structures in hydrogen bonding capsules, MOFs, and MIPs. For example, Custelcean et al. developed hydrogen-bonding capsules lined with urea and pyridinium binding groups and found that using six urea groups to develop twelve hydrogen bonds was most complementary to tetrahedral sulfate and selenate.⁹⁸ Fang et al. developed a MIP with imidazole functional groups that formed four hydrogen bonds with As(III) had the following selectivity (measured as K_d and $\beta$): As(III)>P>S>N.

These electrostatically selective adsorbents have shown great promise in their ability to separate between the strongest competitive adsorptive pairs, such as arsenate and phosphate. However, one important note regarding this mechanism is that its success is highly pH dependent. Therefore, pH must be carefully monitored and controlled in order to optimize performance.

A further subset of electrostatic interactions is hydrophobic effects. As dictated by the Hofmeister Bias, oxo-anions with a lower degree of hydration are more strongly attracted to hydrophobic surfaces, and conversely, more hydrated oxo-anions are more attracted to hydrophilic surfaces.^{17,98,103,199} Therefore, by controlling the hydrophobicity/hydrophilicity of the adsorbent material, more or less hydrated oxo-anions can be rejected or selected.

The degree of hydration of an oxo-anion is largely controlled by the ratio of charge/volume.^17,103 For ions with the same charge, the larger the radius, the lower the charge density and hydration of the ion, and therefore the more hydrophobic the ion.¹¹⁰ In reviewing the literature, there is not yet a comprehensive comparison of degree of hydration between various oxo-anions. Based on our calculations of molecular volume and assuming protonation in environmental conditions, we can rank the oxo-anions in order of decreasing hydration energy: SO₄^2->SeO₄^2->NO₃⁻>HCO₃⁻>HSeO₃⁻>H₂PO₄⁻>HCrO₄⁻>H₂AsO₄⁻>H₃AsO₃>H₄SiO₄.^65,200,201 Generally, monovalent oxo-anions have a lower hydration energy, and therefore are more strongly attracted to hydrophobic adsorbents, than divalent oxo-anions.^17,64,199

Several technologies have utilized differences in hydrophobicity to develop selectivity. For example, it was found that hydrophobic activated carbon preferred (measured as %I) HCrO₄⁻ over H₂PO₄⁻, NO₃⁻ and SO₄^2-.¹⁰³ Luo et al. used surfactant modified clays and found that generally, monovalent ions were preferred for adsorption within the hydrophobic, surfactant-coated interlayer space over divalent ions due to their lower hydration energies.¹⁹⁹ They observed the following selectivity sequence (measured as %I): NO₃⁻ >HCrO₄⁻ > SeO₄^2-~SO₄^2- > HSeO₃⁻>H₂AsO₄⁻.

Another hydrophobic material type frequently utilized in selective adsorbents is polymers. Resin hydrophobicity can be varied by changing the composition of the polymer and the chain length of functional groups on the polymer.¹¹⁰ Polymers have been widely used in a variety of adsorbents including MOFs⁶⁴ and anion exchange resins^{29,57,110,202} to enable selective adsorption of oxo-anions (Table 6). Custelcean and Moyer found that by including higher charged metals such as Fe(III) and Zr(VI) in their MOF framework, they selected for smaller and more hydrophobic oxo-anions.⁶⁴ Further research is needed to fully explore the potential of MOFs to hydrophobically select for target oxo-anions.

Many anion exchange resins are made up of cross-linked polymers with functional groups such as amine groups that serve as binding sites for oxo-anions.²⁰² Conventional anion exchange resins, due to their hydrophobic nature, follow the Hofmeister bias and prefer less hydrated, more hydrophobic anions.^57,202 When targeting HCrO₄⁻, a more hydrophobic ion than most of its strongest competitors, a more hydrophobic adsorbent is more effective in developing selectivity. Hydrophobicity of the adsorbent can be influenced by polymer composition, and the chain length of the functional groups.¹¹⁰ For example, in a bicarbonate-form strong base anion exchanger, by selecting a more hydrophobic resin they were able to successfully select for (measured as $\propto$) HCrO₄⁻ over weakly competing SO₄^2- and NO₃⁻, but did not evaluate stronger competitive oxo-anions such as phosphate.¹¹⁰

Several studies used of weak-base anion exchange resins and found that they typically exhibit non-Hofmeister selectivity.^57,202 For example, Awual et al. modified cross-linked polystyrene through the addition of primary amino groups, resulting in a more hydrophilic weak-base anion exchange resin that exhibited the following selectivity (measured as %R): H₂PO₄⁻>H₂AsO₄⁻>SO₄^2->NO₃⁻.⁵⁷ While this weak-base anion exchange resin was successful at reversing the Hofmeister bias, it was not able to select for As(V) over phosphate. Some studies also examined the potential of MOFs and hydrogen bonding capsules to exhibit non-Hofmeister selectivity.^17,64,98 They found that through combinations of rigid hydrogen bonding networks and metal coordination, anti-Hofmeister selectivity can be developed. For example, Custelcean et al. observed the following selectivity (measured via differences in Gibbs free energy of crystallization) with a Mg(II)-hydrogen bond capsule: SO₄^2->SeO₄^2->HCO₃⁻.⁹⁸

8. Use of Quantum Simulation to Understand and Predict Oxo-anion Sorption

Over the years, density functional theory (DFT) and wave-function methods (Hartree-Fock, MP2, couple cluster) have found important application in understanding oxo-anion adsorption on the metal oxide surfaces.^203–209 Specifically, DFT methods provide two tools to complement and propel experimental approaches in development of selective adsorbents: 1) the ability to isolate individual cause/effects by focusing on processes at the site of interaction of the studied system; and 2) the use of rapid screening approaches to identify novel sorbent materials, predicting the most promising candidates that could then be validated experimentally.

Due to limitations dictated by system size (<500–1000 atoms) and simulation time scale (either thermodynamic states or a few femto-seconds), the DFT calculation of competitive adsorption energies and Gibbs free energies²¹⁰ must be completed on a specific material surface facet, with a specific pH, ionic strength, etc.^168,211 Competing effects, such as presence of multiple facets, defect sites, and sample analysis treatments are generally avoided allowing the direct delineation of the effect of changing variables. This is commonly done by calculating oxo-anion-surface complexation energies;^212–214 with DFT calculated Ultra Violet, Infrared and EXAFS spectra of adsorbed oxo-anions aiding in experimental interpretation and cross-method validation.^214,215 In addition, DFT calculations enabled direct elucidation of the nature of the chemical and physical bonds between adsorbate and surface^{168,209,210,216} becomes useful in identifying material characteristics required for higher adsorption selectivity.²¹⁷ The predictive and design power of DFT methods have been underutilized in selective sorbent design. In general, DFT based design of novel materials follows two routes: 1) brute force materials screening or 2) descriptor identification and materials selection.

Effective screening requires rapid calculation of competitive oxo-anion adsorption on hundreds to thousands of candidate material compositions and morphologies which is complicated and time-consuming process. Only limited number of studies going beyond simple comparisons of two or three materials have been reported to date.^{210,217–219} Several recent studies of Cr(VI) adsorption on hematite have suggested outer-sphere, monodentate, bidentate mononuclear, and bidentate binuclear inner-sphere complexes as being favorable configurations.^{209,215,220,221} However, the specific hematite facet was not considered, even though it was later found to be critical factor in determining the most energetically stable (i.e. lowest energy) configuration.^{168,215,220,221} Therefore, screening studies require physically representative model systems, where the material composition, surface characteristics, and facet; pH dependent effects; as well as surface and oxo-anion solvation effects, are necessary.^{168,213,216,221,222} Additional effects such as oxo-anion/oxo-anion interactions,^212,223 and pH / presence of other ions in the solution should also be included in the screening models.

The development of adsorption descriptors which use simple formulas or material characteristics to predict adsorptivity mostly rely on adsorption energy and selectivity derived from the DFT calculations and identification of materials characteristics that correlate with desired adsorptive performance. For example, Corum and Mason found that strong adsorption of As oxo-anions on Keggin Al-nanoclusters was correlated with large local shape and electronic potential gradients.²¹¹ With a suitable and clearly defined descriptor, materials could be screened without detailed DFT calculation for interaction of oxo-anions and each sorbent material, and only candidates with most promising properties would be subjected to full DFT calculations (i.e. adsorption energies, Gibbs free energies, spectra calculations, etc.). This approach has become standard in other fields (e.g., use of Volcano plots in catalysis²²⁴). One of the difficulties in effective application becomes establishing an adequate database of selective or non-selective sorbents that would be used for defining the most appropriate descriptor(s).

DFT has a great potential to aid in the discovery of selective oxo-anion adsorbents. The key areas that limit its implementation are the size of model system needed to adequately describe the complex sorbent/sorbate environment, the development of efficient computational materials screening tools, and sufficient high-quality experimental validation data. Use of novel methods, based on linear scaling methods,²²⁵ artificial intelligence/machine learning methods,²²⁶ or combined quantum mechanics/molecular model systems²²⁷ can help with development of complex model systems and identifying features that categorize behavior while decreasing computational time.

9. Future Directions and Opportunities

Reviewing the literature, it is clear that variation in surface complexation, HSAB interactions, sterics, and electrostatics, can be exploited for the development of selective adsorbents. Rarely do these mechanisms occur singularly, often acting synergistically to strengthen the specificity of a selective adsorbent (Table 6). The strength of these mechanisms varies, with those that involve inner-sphere complexation and steric interactions exhibiting stronger selective adsorptive performance compared to weaker outer-sphere or electrostatic interactions in antagonistic situations. Further research is needed to elucidate these synergistic and antagonistic relationships between selectivity mechanisms. The most widely used mechanisms to develop selectivity were electrostatics and sterics, with HSAB and crystal facet engineering requiring more study. However, very few of the relevant studies contained discussions of adsorption mechanisms, and even fewer critically examined and proved the adsorption and selectivity mechanisms. In addition, due to a lack of a standardized definition for quantification of selectivity, it is difficult to quantify the reliability of removing target contaminants to beneath regulatory limits in competitive systems. Kinetics also warrant more consideration since differences in diffusion rates between oxo-anions can have a large influence on selectivity in competitive systems.¹⁴² This is particularly important since steric selectivity mechanisms tend to more strongly control selectivity, leading to a need to more carefully consider the influence tortuosity and pore size on diffusion rates in these competitive systems.

Great opportunities exist to develop selective adsorbents for arsenite, selenite, and selenate as very few studies have targeted these oxo-anions (Table 6). The largest number of selective sorbents have been developed for chromate and arsenate. However, critically, very few of these studies have successfully developed selectivity for these target contaminants over their strongest competitors (e.g., phosphate and silicate) or if silicate is evaluated, its surface polymerization is rarely considered. Studies of singular target/contaminant pairs are important to truly understand fundamental competitive relationships, more complex multicomponent waters which more accurately reflect environmental conditions should also be tested. Use of SCMs derived from this data could enable easier prediction of systems conditions in which selectivity will be achieved. Additionally, the potential trade-off between highly efficient selective removal and decreased regeneration potential also needs to be evaluated.²²⁸

The majority of adsorbent materials reviewed that successfully developed selectivity were organometallics including MOFs, PLEs, AEs, and MIPs; however, these were also the mostly studied materials (Table 6), so there is room for innovation to other sorbent systems. Further systematic empirical studies, iterated with computational techniques, are needed to more accurately and efficiently screen materials and develop selective adsorbents. For example, DFT modeling could be used determine which crystal facets of NMOs show high binding energy for targets and low binding energy for competitive oxo-anions. Sustainability considerations should also be considered during this screening process, including questions of materials criticality, energy consumption for production and regeneration, and environmental and human health impacts from any potential waste streams across the life cycle of these systems. Additionally, any opportunities to develop multifunctionality such as simultaneous reduction/oxidation of a contaminant into a more easily adsorbed form and sensing/quantification of contaminants upon adsorption are worthy of exploration and development.