Phosphorus in water: A review on the speciation analysis and species specific removal strategies

Abstract Elemental phosphorus (P) is key to all life forms on earth. Efficient P management and control in natural environments especially in water bodies is of paramount importance to the balance and stability of the ecosystem on both local and global scales. In the past decades, there have been numerous efforts devoted to the P analysis as well as on its efficient removal in water. However, natural occurrence of P species is in diverse forms with different properties, some even yet to be known, posing challenges to current analytic methods and removal technologies. In this review, we make an attempt to clarify the current advances on the analysis of different P species in water as well as the corresponding removal strategies. We hope to provide a new perspective for P management purpose, i.e., linking the P speciation analysis with the removal strategies, and offer a complementary guidance for researchers that are normally specialized in either field. Moreover, we envision future directions in both fields, and address the need for the development of P species-orientated removal strategies with high efficiency and selectivity based on advanced analytic technologies. Graphical abstract


Introduction
Phosphorus (P) is one of the crucial elements in biochemistry. The P-O bond represents the key building block for the formation of the deoxyribonucleic acid, which is essential for almost all living species on earth (Conley et al., 2009). However, the over-enrichment of P in water bodies promotes the expansion of harmful algal blooms, which is referred to as eutrophication that devastates the aquatic ecosystem and poses threat to drinking water supplies (Diaz & Rosenberg, 2008). Consequently, stringent regulations and recommendation have been proposed by governmental agencies to restrict the total P (TP) concentration in waters, for eutrophication control and sustainability of aquatic ecosystem. For examples, U.S. Environmental Protection Agency (U. S. EPA) requires the TP concentration < 20 lg/L in rivers, streams, lakes and reservoirs in the Nutrient Criteria Technical Guidance Manual, which provides technical guidance to further establish water quality criteria and standards (Nutrient Criteria Technical Guidance Manual, 2000). The China has set the maximum allowable TP concentration at 10-200 lg/L in lakes and reservoirs (China EPA, 2002). For the purpose of the P management, a systematic understanding of the precise P speciation and strategies for the removal of various P species are required.
In aqueous dissolved form, the P speciation is complicated with both inorganic and organic P species. Inorganic P (IP) species is mainly present in the form of orthophosphate (phosphate with only P-O bonds), which is the most bioavailable form of P and could be readily detected. Other IP species including pyrophosphate, other condensed phosphates, phosphite, hypophosphite, and phosphine, could also exist in water with much lower concentrations than orthophosphate and are thus of less environmental concern (Pasek, 2008). Organic P species (organic phosphorus, OP) exist in more abundant forms including phosphonates (with C-P bonds), organophosphate esters (OPEs) (with C-O-P bonds), etc., but normally in lower concentration (ng/L-lg/L) than IP (Chokwe et al., 2020;Lorenzo et al., 2019). Moreover, some P species are in non-dissolved forms that could not pass through a 0.2 or 0.45 lm filter, for example, particulate form that includes clay and silt-associated IP and OP, and Pcontaining biological substances. Consequently, it is of analytic challenge to develop reliable techniques to determine the accurate P speciation in each fraction. Currently, there have been several comprehensive reviews summarizing the state-of-the-art technologies for the detection of various P species (Alam et al., 2021;Chen et al., 2019). Here, we do not aim to provide a repetitive summarization of all analytic techniques for P detection, but rather to give a brief guidance for analysis of the dissolved P only for water treatment purpose, and thus, the P species in non-dissolved forms are not considered. The basic structures, environmental effects, matrices and concentrations of P species in different aqueous scenarios are summarized in Table S1 (Supplementary material).
In the past decades, a series of strategies have been developed for the removal of various P species from water, including chemical precipitation Peng et al., 2018), enhanced biological phosphorous removal (EBPR) (Roy, 2017;Yang et al., 2019), and adsorption (Bacelo et al., 2020;Wu et al., 2020). However, the available reviews that summarize the existing P removal strategies mainly focused on orthophosphate rather than other P species. Many OP species are present not only in natural waters but in wastewaters from various industries, such as household cleaning products, paper and textile production, membrane filtration, and cooling water systems, raising great environmental concerns. Compared to orthophosphate, the removal of OP is generally more challenging (Venkiteshwaran et al., 2018). Consequently, it is necessary to provide a complementary summary that covers the removal strategies for other P species, with the emphasis on OP.
As for the analytic efforts for P species in water, one of the ultimate goals is to achieve more efficient P management using various removal technologies toward different P species. A brief summarization of the existing analytic techniques could provide a better understanding on the P speciation in water and thereafter facilitate to develop speciation-based P removal strategies. Through this review, we aim to provide a different perspective on the combination of the P analysis and removal for P management purpose, and to stimulate more P species-orientated removal techniques with high selectivity and efficiency in the future.

Sample pretreatment
A series of sample preparation and pretreatment follows general protocols for the aqueous P analysis. For the detection of soluble reactive P(SRP), all water samples need to pass through a 0.2 or 0.45 lm filter in order to remove the particulate moiety. Afterward, different pretreatment processes are necessary for the measurement of different P species. For examples, the determination of total P (TP) requires the transformation of all P species, including P-O-P, C-O-P, and C-P bonds, to the detectable inorganic form through digestion processes such as thermal oxidation and UV photo-oxidation (Maher et al., 2002;Worsfold et al., 2005Worsfold et al., , 2016. For the quantification of trace OP, the water samples need to be enriched and purified by solid phase extraction (SPE) cartridge for further chromatography or mass spectrometry (MS) analysis. In many circumstances, chemical derivatization such as methylation is preferred to improve the chromatographic retention and/or enhance the mass spectrometry response (Chen et al., 2012;Schmidt et al., 2014). For example, we recently reported the trimethylsilyldiazomethane (TMSCHN 2 ) pre-methylation strategy for the MS detection of trace phosphonates with 2-3 orders of magnitude improvement of sensitivity by charge neutralization . Additionally, new nanomaterials have been developed for the extraction of OP analytes during analytical pretreatment process to improve the analytical speed, convenience, sensitivity and selectivity. For instances, ionic liquid-based sol-gel fiber was prepared and combined with headspace solid-phase microextraction (SPME) for extracting OPEs (Gao et al., 2013), metal-organic frameworks (MOFs) based core-shell magnetic microspheres were used for sorption of OPEs with magnetic dispersive solid-phase microextraction (MDSPME) (Pang et al., 2018), and magnetic molecular imprinted polymer (MIP) andzirconium (IV) functionalized magnetic nanocomposites were applied in MDSPME to separate and preconcentrate organophosphorus pesticides (Bazmandegan-Shamili et al., 2016;Jiang et al., 2016). Below we summarize typical chemical analytic techniques for the P detection in water treatment, including ammonium molybdate spectrophotometric method, flow injection analysis (FIA), inductively coupled plasma (ICP), ion chromatography (IC), gas/liquid chromatography (GC/LC), and other spectroscopic methods. The illustrations of these methods are presented in Figure 1. The analytical methods for P species, including the advantage and limitation, target P species, aqueous matrix, method description, and LOD are summarized in Table S2 (Supplementary material).

Ammonium molybdate spectrophotometric method
The ammonium molybdate spectrophotometric method was developed in 1962 (Jenkins, 1973;Murphy & Riley, 1962), and is considered to be the simplest, most reliable, and standard method for the chemical analysis of dissolved IP (orthophosphate and TP) (China EPA, 1989). The underlying mechanism is the reaction between phosphate and acidified molybdate to generate 12molybdophosphosphoric acid (12-MPA), which is further reduced to phosphomolybdenum blue (PMB) using reducing agents such as ascorbic acid or stannous chloride (SnCl 2 ) (McKelvie et al., 1995). The absorbance of PMB at the wavelength of 700 nm or 880 nm is linearly proportional to its concentration, allowing accurate quantification of the P concentration with the limit of detection (LOD) of 10 lg/L. Nevertheless, there are some inherent limitations of this method, including the indistinguishable detection of phosphate and acid hydrolyzable P, overestimation of free phosphate due to the presence of other P species with molybdate activity (Maruo et al., 2016;Worsfold et al., 2016), and possible interference of coexisting ions (e.g., silicate, arsenate, nitrite, nitrate, sulfide, chromium, and copper) that could form complex with PMB (Nagul et al., 2015).
With the increasing demand of the detection of P at trace concentration, various efforts have been made to further improve the sensitivity of the ammonium molybdate spectrophotometric method. For examples, strategies including pre-enrichment by extraction, pre-concentration using magnesium induced coprecipitation (MAGIC), enhanced molar absorptivity via the interaction of phosphomolybdate and dye, and increased optical path length using long capillary detector, have been demonstrated to be effective to reduce the LOD down to ng/L level (Anagnostou & Sherrell, 2008;D'Angelo et al., 2001;Itaya & Ui, 1966;Ormaza-Gonz alez & Statham, 1991).

Flow injection analysis (FIA)
FIA is a continuous flow analysis technique developed in 1975 ( _ Ruzicka & Hansen, 1975). For the detection of phosphate based on the same principle of the ammonium molybdate spectrophotometric method, FIA measures the signal of absorption intensity using a flow-through detector (Frenzel & Mir o, 2019). Compared to the conventional colorimetric method, FIA possesses the advantages of high analysis speed, high accuracy and precision, simple operation, small dosage of sample and reagent, and in-situ automatic monitoring. However, the detection using FIA might be interfered by metal ions such as Cu, Pb, Cd and high salinity in sea water (Alam et al., 2021). FIA has been widely used for the detection of P in natural water bodies such as river water, lake water, and seawater, with LOD at the lg/L level (Chen et al., 2021;Deng et al., 2020;Kortazar et al., 2016). For many years, there have been a lot of studies focusing on the optimization of the operational parameters, injection mode (Johnson & Petty, 1982), and flow cell and detector (Ellis et al., 2003;Zagatto et al., 1990), to further improve the performance for P analysis of the FIA technique. FIA has also been coupled with other techniques including inductively coupled plasma atomic emission spectrometry (ICP-AES), flame atomic absorption spectrometry (FAAS), and luminol chemiluminescence (CL) detection, to improve the sensitivity for multi-element measurement (Hartenstein et al., 1985), to enhance the anti-interference ability (Koscielniak'& Kozak, 2002), and to achieve rapid determination of phosphate at nanomolar concentrations (Yaqoob et al., 2004).

Inductively coupled plasma (ICP)
As a representative hard ionization technique, ICP targets to destruct molecules to their elemental components, serving as the preferred ionization method for elemental analysis. ICP could also be used to detect the P element, however, with strong interference by other elemental fragments (e.g., 15 N and 16 O) in plasma (Cooper et al., 2005). Such interference could be mitigated by coupling a high-resolution mass analyzer, achieving ultrahigh sensitivity to elemental P with LOD of lg/L level (Alam et al., 2021). Nowadays, for the detection of trace P in water, ICP is normally coupled with atomic emission spectroscopic (ICP-AES) or mass spectroscopic (ICP-MS) (Miller-Ihli & Baker, 2001). Moreover, both ICP-AES and ICP-MS have been used in combination with other analysis or separation techniques including capillary electrophoresis (CE), ion chromatography (IC), and high performance liquid chromatography (HPLC), for the detection of P in water with much enhanced efficiency, sensitivity, and selectivity (Brabandere et al., 2008;Fujii et al., 2009;Guo et al., 2005;Schmidt et al., 2014).

Ion chromatography (IC)
IC technique has been used for the analysis of P species in water mostly for phosphate and TP, with fewer examples for other IP (reduced and condensed P) and OP (e.g., phytates, alkyl phosphate, and phosphonates) (Ruiz-Calero & Galceran, 2005). IC possesses the advantages of high sensitivity and accuracy with LOD at lg/L level. IC includes multiple chromatographic systems for ion determination, such as ion-exchange chromatography (IEC), ion-exclusion chromatography, chelation ion chromatography, and the novel electrostatic ion chromatography. Among them, IEC based on suppressed conductivity detection is preferred for the detection of phosphate (Lu et al., 2002;Ruiz-Calero & Galceran, 2005), while techniques based on non-suppressed conductivity detection are rarely used due to the poor sensitivity of the system. Current research efforts concerning IC include the development of new stationary phase and inhibition system, and the improvement of analysis of anions in reversed-phase chromatographic columns, for more effective separation and lower LOD for P detection. Moreover, IC has also been coupled with other detection techniques such as ICP-MS, UV, evaporative laser light scattering detector (ELSD), and refractive index (Buchberger, 2000).

Gas/liquid chromatography (GC/LC)
GC/LC has been extensively used for the analysis of diverse OP, including organophosphonic acid, organophosphate ester, organophosphorus pesticide, and flame retardant, and etc. (Gao et al., 2013;Pantelaki & Voutsa, 2019;Wang et al., 2009). There are several key advantages of the GC/LC technique, including high accuracy and sensitivity, strong separation ability, fast analysis speed, and broad application range. When coupled with MS, highly accurate qualitative and quantitative detection of various organophosphorus species could be achieved simultaneously, with LOD as low as ng/L level (Gao et al., 2013;Wang et al., 2009). The accuracy of the quantitative analysis by GC/LC-MS could be further improved by using isotope labeled internal standard or isotope dilution methods. Nevertheless, the expensive large instruments and highly delicate pretreatment are required for GC/LC analysis.
GC could be equipped with selective detectors, such as electron capture detector (ECD), flame photometric detector (FPD), and nitrogen and phosphorus detector (NPD), for the analysis of trace OP (Gao et al., 2013;Pantelaki & Voutsa, 2019;Wang et al., 2009). The coupling of GC with MS detector could provide more structural information of OP with greatly improved accuracy of the qualitative analysis. Of note is that it is difficult to use GC to directly analyze OP compounds that are thermally unstable, non-vaporizable, and highly polar, and additional chemical derivatization is always required.
LC is particularly suitable for the analysis of OP compounds that are thermally unstable, strongly polar, and/or difficult to volatilize. Modern HPLC instrument is normally equipped with highly sensitive detectors, including UV absorption, MS, fluorescence detector (FLD), diode array detector (DAD), and electrochemical detector (Fu et al., 2009;Kowalski & Mazur, 2014;Wang, Sun et al., 2019), allowing full analysis with a sample of only microliter. The LC-MS technique is especially sensitive and selective. For example, LC-Orbitrap-MS has recently been reported for suspect analysis and non-targeted screening to identify unknown OP species (Ye et al., 2021). For the analysis of OP using LC-MS, inverse mode is normally adopted in LC using the combination of C18 or C8 stationary phase or reversed phase and hydrophilic interaction column, while electrospray ionization (ESI) source is mostly adopted in MS (Sud et al., 2021). It is clear that LC-MS is now widely used as the major technology for the analysis of OP in aqueous phase.

Other spectroscopic techniques
Other spectroscopic techniques, such as 31 P nuclear magnetic resonance spectroscopy( 31 P-NMR), near infrared spectroscopy (NIRS), Raman spectroscopy, and X-ray based spectroscopy (XRFS, XAS), have been used for the qualitative characterization of P-bonds and their spatial distribution (Alam et al., 2021). Among them, 31 P-NMR is the most powerful nondestructive technique aiming for simultaneous identification and quantification of all naturally occurring P-containing species in a single analysis, for both solid and aqueous samples without complex purification and pretreatment procedures (Cade-Menun, 2005). Nevertheless, there are several limitations of 31 P-NMR, including relatively high LOD (mg/L level), interference of the 31 P chemical shift by matrix substances, interference of paramagnetic ions (Fe, Mn) that could result in the failure of the detection of 31 P (Chen et al., 1999;Christ et al., 2020;Godinot et al., 2016). In addition to the analysis of P speciation in water, 31 P-NMR has also been demonstrated to be powerful to reveal the structure, transformation, and reaction kinetics of P species (Bai et al., 2015;Hafuka et al., 2021;Read et al., 2014;Zhang, Li et al., 2019).

Removal strategies for different phosphorus species
The P speciation, fraction, and concentration differ significantly in diverse water matrices. Hence, the case-specific design of phosphorous treatment strategy is recommended in real applications. Generally, the traditional separation strategies, including chemical precipitation, flocculation and adsorption target inorganic P, although some organic P may also be removed with relatively low efficiency. However, the currently available processes usually work ineffective when the initial P concentration is low (< 100 lg/L) and TP is comprised of a significant organic phosphorus component (Mayer et al., 2013). For this situation, a two-step strategy offers a feasible solution for the in-depth removal of phosphorus, that is, the transformation of organic P and/or other inorganic P such as hypophosphite/phosphite to phosphate, followed by a precipitation and flocculation process. Various oxidation techniques, including UV-based oxidation (Gray et al., 2020;Sun et al., 2019), Fenton/Fenton-like reaction (Rott et al., 2017a;Sun et al., 2022;Zhu et al., 2021), and electro-chemical oxidation (Lei et al., 2020), have been demonstrated for the transformation of organic P to phosphate, while hydrolysis and reduction technologies could be used to transform organophosphate esters (Fang et al., 2018;Li et al., 2021). In addition, biological treatment is one of the most promising methods for the removal of phosphorous in terms of commercial availability. For example, conventional activated sludge and the EBPR technique could remove P using a combination of biological accumulation and physical/chemical separation. Although organic P such as phosphonates could not be mineralized in conventional biological treatment, it can be removed by adsorption in the sewage sludge. The different treatment technologies for the removal of various P species are illustrated in Figure 2 and summarized in Table S3 (Supplementary material). We describe the details in the following sections.

Removal of orthophosphate
Apparently, the removal of orthophosphate has received the most attention due to its abundant presence as the major form of P in water. There has also been a number of comprehensive reviews on the removal of orthophosphate using various technologies (Bunce et al., 2018;Wu et al., 2020;Yang et al., 2019). Here, we only present a brief summarization of these strategies for the removal of orthophosphate.

Chemical precipitation
Chemical precipitation is a traditional method for the removal of orthophosphate with high concentration from water via the addition of divalent or trivalent metal salts, e.g., iron or aluminum (Bunce et al., 2018;Hasan et al., 2021). The addition of the salt generates precipitates as solid residuals that could be easily removed by natural settling or filtration. Due to the restriction of the solubility product of the metal phosphate, chemical precipitation could normally reduce the effluent phosphate concentration to 0.5-1 mg/L or above. Further reduction requires much higher dosage of metal salts, which is therefore not attractive from the economic perspective. Chemical precipitation has also been utilized to recover phosphate from the aqueous phase. For this purpose, calcium and magnesium ions are commonly employed as precipitators to react with phosphate to form Ca 5 (OH)(PO 4 ) 3 (hydroxyapatite ¼ HAP) and MgNH 4 PO 4 Á6H 2 O (struvite ¼ MAP), respectively .
Electrochemical precipitation has also been demonstrated to be effective for the removal of orthophosphate (Cusick & Logan, 2012;Hug & Udert, 2013). Lei et al. (2017Lei et al. ( , 2018 established an electrochemical method for the removal and recovery of orthophosphate via the precipitation of calcium phosphate in the vicinity of and on the cathode due to the elevated pH environment (Lei et al., 2017(Lei et al., , 2018. The shortcomings of electrochemical precipitation include the low selectivity due to the co-precipitation of CaCO 3 as well as high energy consumption.

Biological methods
In the past two decades, there have been extensive studies on the P removal via biological methods, particularly the EBPR using activated sludge system. During EBPR process under alternative aerobic and anaerobic conditions, phosphorous-accumulating organisms (PAO) are formed, which could absorb orthophosphate from wastewater with amount higher than that required for normal growth. The EBPR process could reduce the effluent P concentration to be 0.1-0.2 mg/L under optimum conditions (Acevedo et al., 2012;Bunce et al., 2018;Nguyen et al., 2013). However, due to the restriction of multiple factors such as organic loadings, toxic substances and operation parameters, the effluent P concentration that EBPR could achieve is much higher in the real scenarios (Brown & Shilton, 2014;Seviour et al., 2003).

Adsorption
Adsorption process has been widely used for the removal of orthophosphate, especially for deep purification. In general, there are five mechanisms involved in phosphate adsorption: (a) ion exchange (outer-sphere surface complexation), (b) ligand exchange (inner-sphere surface complexation), (c) hydrogen bonding, (d) surface precipitation, and (e) diffusion into the interior structure of the sorbent (Wu et al., 2020), which are illustrated in Figure 3. The predominant mechanism in a specific adsorption process depends on the physical and chemical characteristics of the adsorbent and the environmental/operational conditions. A variety of adsorbents has been reported for this purpose, including metal oxides/hydroxides, mesoporous materials, metalorganic hybrid materials, carbonaceous materials, minerals, and various types of modified waste. Among these materials, metal oxides/hydroxides, e.g., Al, Fe, Mn, La, and Zr, have attracted most interest because they could form specific inner-sphere complexation with orthophosphate (Liu, Chi et al., 2018). Of note is that the direct use of these metal oxides/hydroxides nanoparticles is challenged by difficult operation and potential leaking in practical application. Consequently, a lot of efforts has been made to address these issues by loading the nanoparticles onto various substrate materials, including magnetic materials Zhang, Dan et al., 2020) and porous materials (Pan et al., 2009(Pan et al., , 2014Zhang et al., 2016). This strategy creates large space for the investigation and manipulation of the interaction between nanoparticles and the host to facilitate the P removal by nanocomposites.
We have been particularly interested in investigating the adsorption and transformation of orthophosphate inside the nanocomposites. For examples, the ubiquitous presence of Ca 2þ is believed to have negative effects on the P adsorption by granular ferric oxide. However, our recent work discovered that Ca 2þ could enhance the P removal by a polymer supported ferric nanocomposites, via a Ca-P coprecipitation mechanism (Zhang, She et al., 2019). By comparing the La(OH) 3 nanoparticles in bulk and confined inside a polymer matrix for P removal, we later revealed that the polymer nanoconfinement facilitates the transition from La-P inner-sphere complexation to the formation of LaPO 4 ÁxH 2 O, and finally to LaPO 4 (Zhang, Wang et al., 2021). Very recently, we reported a new La-based P binding agent using Mg/Fe layered double hydroxide clay as host. The resultant material exhibits high P uptake of $55 mg/g and, more remarkably, superior stability against a high concentration of bicarbonate and humic substances. Such superior performance results from the unique structure of the host material for La loading, which not only contributes to direct P adsorption but also acts as a nanoshelter to provide an ideal microenvironment for specific La-P interaction against the coexisting substances (Zhang, Kong et al., 2021).

Removal of hypophosphite and phosphite
Compared to metal orthophosphate, both metal hypophosphite and phosphite precipitates normally possess higher solubility product constants, restricting their removal through chemical precipitation (Zhang, Zhao et al., 2020). Consequently, a two-step strategy has been developed for the removal of hypophosphite and phosphite in wastewater, that is, the oxidation of hypophosphite and phosphite to phosphate followed by a precipitation and flocculation process. Note that the oxidation of hypophosphite to phosphite is easy, while the oxidation of phosphite to phosphate is difficult even in the presence of oxidants like H 2 O 2 (Liu et al., 2013). In recent years, there have been several studies attempting to remove hypophosphite and phosphite using various oxidation technologies, including O 3 /H 2 O 2 combined with sequential Fe(II) catalytic process (Zhao et al., 2017), photo-electrochemical oxidation by UV/Fe 2þ /peroxydisulfate with electrochemical process (Tian & Zhang, 2019), photo-oxidation in the presence of ferric and oxalate ions (Qiu et al., 2015), Fenton and electro-Fenton oxidation of an effluent-containing hypophosphite and phosphite (Pozzo & Petrucci, 2013). In order to simplify this two-step approach and reduce the consumption of chemicals, Liang et al. (2019) reported a one-step strategy for the treatment of phosphite-containing wastewater using a single electrochemical reactor that integrates electrochemically induced oxidation and electrocoagulation process without the use of external chemicals such as oxidants, iron salts, or alkali (Liang et al., 2019). Recently, Zhang, Zhao et al. (2020) developed a photo-electrocatalytic cell system consisting of a TiO 2 /Ni À Sb À SnO 2 bifunctional photoanode and an Fe 2þ dosed carbon fiber cathode with for the oxidation of hypophosphite to P and simultaneous precipitation of FePO 4 that is easy for P recovery (Zhang, Zhao et al., 2020).

Removal of pyrophosphate
Adsorption has been considered to be very effective for the removal of pyrophosphate from water. In particular, several layered double hydroxide (LDH) materials, including CaFe-LDHs and MgFe-LDHs, have demonstrated satisfactory performance as adsorbents for the removal of pyrophosphate (Xu, Hong et al., 2019). However, pyrophosphate with high concentration might cause strong complexation effects with these materials, resulting in the dissolution of Fe 3þ , Ca 2þ , and Mg 2þ , as well as the decrease of the pyrophosphate removal efficiency. Recently, Fu et al. (2021) reported the use of plating waste as precursor to prepare Zn 2 Cr-LDHs for the efficient removal of pyrophosphate in electroplating wastewater (Fu et al., 2021).

Removal of phosphonate with C-P bonds
Phosphonates are present in municipal wastewaters and various industrial wastewaters. Rott, Steinmetz et al. (2018) have critically reviewed the existing studies until the year 2016 on the potential environmental relevance of phosphonates, their biotic and abiotic degradability, and their removal in wastewater treatment plants (WWTPs) (Rott, Steinmetz et al., 2018). It is essential to remove phosphonates from water because they are resistant to most microorganisms and could be transformed to more toxic P species, i.e., aminomethylphosphonic acid, via photolytic degradation and biological processes (Rott, Steinmetz et al., 2018).

Precipitation and flocculation
In WWTPs, phosphonates tend to form complexation with the precipitant and disrupt the precipitation of phosphate. Furthermore, the complexation also causes an overdosage of the precipitant, resulting in acidification and a constant lack of bioavailable phosphate in the system that interferes the subsequent nitrification treatment (Rott, Steinmetz et al., 2018). Rott et al. (2017b) examined the use of various precipitants, e.g., Fe 3þ , Al 3þ , Ca(OH) 2 , and NaOH, in the flocculation/precipitation for phosphonates removal from industrial wastewaters and particularly, took into account a wide range of different wastewater matrices and different phosphonates (Rott et al., 2017b). Recently, Ren et al. (2020) reported a modified chemical precipitation process using FeCl 3 as precipitant, followed by the addition of a starch-based flocculant for the removal of both inorganic and organic P from simulated turbid wastewaters. Compared with traditional chemical precipitation using FeCl 3 only, this modified method showed much higher removal of TP and turbidity, and improved floc properties (Ren et al., 2020).

Biodegradation and adsorption onto activated sludge
It is generally accepted that phosphonates could not be mineralized in conventional biological WWTPs, neither aerobically nor anaerobically, due to the rather slow biodegradation kinetics and small number of phosphonate-utilizing microorganisms (Rott, Steinmetz et al., 2018). Consequently, the phosphonates remain in adsorbed form in the sewage sludge and could only be completely oxidized and removed from the environment.

Adsorption
Adsorption is effective to remove phosphonates. Similar to phosphate, phosphonates could also form inner-sphere complexation with metal (mainly as Fe) (hydr)oxides (Mart ınez & Farrell, 2017;Yan & Jing, 2018). Therefore, various iron-based adsorbents have been used to remove phosphonates (Boels et al., 2010(Boels et al., , 2012Reinhardt et al., 2020;Rott, Nouri et al., 2018). For examples, Boels et al. (2012) used granular ferric hydroxide to adsorb nitrilotris(methylenephosphonic acid) (NTMP), which is a representative phosphonate used as scale inhibitor from membrane concentrate (Boels et al., 2012). Rott, Nouri et al. (2018) reported the use of magnetic ZnFeZroxyhydroxide adsorbent for the efficient removal of several phosphonates from real industrial membrane concentrate (Rott, Nouri et al., 2018). Although these adsorbents have exhibited attractive performance for the removal of phosphonates, their selectivity toward phosphonates is challenged especially in the presence of phosphate, because phosphate could interact with the adsorbent via stronger inner-sphere complexation than phosphonates (Nowack & Stone, 2006;Yan & Jing, 2018). Very recently, we reported a rational design of nanocomposite adsorbent by confining hydrated ferric oxide nanoparticles in hyper-cross-linked resin to realize highly selective adsorption of NTMP in the presence of various anions, organic matters at much greater levels, and particularly the organic analogue nitrilotriacetic acid and phosphate. The high selectivity toward phosphonate was achieved because of the cooperative contributions of the host as well as the embedded nanoparticles (Yan et al., 2022) (Figure 4a).

Oxidation
Many efforts have been devoted to mineralize phosphonates using various oxidation techniques, including ozonation, Fenton reaction, UV-based oxidation, electro-chemical oxidation, etc. Ozone, a powerful oxidant with oxidation potential of 2.07 V, has a strong ability to transform recalcitrant compounds to smaller and more biodegradable molecules, through either directly reacting with target pollutants or generating oxidative HO radicals dependent upon the solution pH (Wang & Chen, 2020). Xu, Wang et al. (2019) reported the use of ozonation at high ozone dosages for the oxidation of 2-phosphonobutane-1,2,4-tricarboxylic acid (PBTCA), but achieved only partial removal. More PBTCA could be further removed by the inorganic P precipitates formed in the subsequent coagulation process . Traditional Fenton reaction using H 2 O 2 as oxidant and Fe 3þ /Fe 2þ ions as catalyst could also be used for the oxidation of phosphonates. The presence of ferrous ions triggers the Fenton reaction to decompose H 2 O 2 to generate HO radicals, which could oxidize phosphonates at a low level (only < 20% transformation to phosphate) in pure water even when the H 2 O 2 was highly excessive (Rott et al., 2017a).
UV-based photo-oxidation processes, e.g., UV/H 2 O 2 , UV/persulfate, UV/chlorine, and UV/ Fenton, have been studied for the degradation of phosphonates Rott et al., 2017a;Wang, Chen et al., 2019). Since the major oxidative species are HO and/or SO 4 À radicals in these processes, the degradation of phosphonates normally suffers low selectivity due to the competitive consumption of radicals by coexisting substances especially natural organic matter (NOM) and Cl À . For example, Wang, Chen et al. (2019) reported the efficient degradation of NTMP by the UV-photolysis of persulfate in synthetic water, however, the degradation efficiency of NTMP was reduced by 40% when treating the real reverse osmosis concentrate due to the presence of Cl À and HCO 3 À (Wang, Chen et al., 2019). A series of studies has demonstrated that the UV or sunlight irradiation could convert phosphonates to phosphate in the presence of various metal ions, including Fe(III), Cr(III), Zn(II), Cu(II), and Ca(II) (Fischer, 1993;Matthus et al., 1989) (Figure 4b). In the case of Fe(III) with the highest reactivity, Matthus et al. (1989) suggested that the degradation mechanism is a photoinduced ligand-to-metal charge transfer, where phosphonate forms complex with Fe(III) ions (Matthus et al., 1989). Lesueur et al. (2005) showed that phosphonates undergo UV light conversion, which is enhanced in the presence of iron (Lesueur et al., 2005). Recently, we reported the use of a combined process, i.e., Fe(III) displacement/UV irradiation/co-precipitation, for the efficient removal of Ca(II)-phosphonate complexes from water . During this process, Fe(III) replaces the Ca(II) in Ca(II)-phosphonate complex to form Fe(III)-phosphonate complex of high photo-reactivity. The subsequent UV irradiation causes the degradation of Fe(III)-phosphonate to form phosphate and other intermediates, followed by the final removal via co-precipitation. The combined process is superior to other processes including Fenton/coprecipitation.
Transition metals could mediate the oxidation of phosphonate through oxidant-induced ligand-to-metal charge transfer process. For instance, Nowack and Stone (2000) showed that a fraction of NTMP could be oxidized in the presence of Mn(II) by using molecular oxygen as the oxidant (Nowack & Stone, 2000). The proposed mechanism is the oxidization of Mn(II)-phosphonate complex by molecular oxygen to form Mn(III)-phosphonate. A recent study in our group showed that oxidants including H 2 O 2 and permonosulfate (PMS) could oxidize phosphonates in the presence of Co(II) . For 1-hydroxyethane 1,1-diphosphonic acid (HEDP) as a target phosphonate, the Co(II)/PMS process was effective for its removal in a broad solution pH ranging from 5.0 to 10.0. Based on a series of experiments, we proposed a Co(II)-PMS complex to be the primary reactive species, while hydroxyl radicals (HO), sulfate radicals (SO 4 -) , singlet oxygen ( 1 O 2 ) and Co(III) play as the secondary reactive species for the degradation of HEDP   (Figure 4c). This unique mechanism rendered higher selectivity and efficiency for HEDP removal by the Co(II)/PMS process than those in free radicals-mediated advanced oxidation processes in real water samples. In a more recent work, we demonstrated an efficient and selective oxidation of HEDP by the Cu(II)/H 2 O 2 process at alkaline pH (Sun et al., 2022). In the presence of trace Cu(II), 90.8% of HEDP was converted to phosphate by H 2 O 2 in 30 min, whereas negligible conversion was observed in UV/H 2 O 2 or Fenton systems during the same reaction time. The intramolecular electron transfer process of the complexation of Cu(II) with HEDP was considered to be the key step responsible for the high oxidation selectivity.
It is also feasible to use electrochemical oxidation for the degradation of phosphonate. For example, Lei et al. reported the electrochemical oxidation for the efficient removal and recovery of phosphonates via the formation of calcium phosphate. At the anode, the phosphonates could be oxidized and converted to inorganic orthophosphate, while water reduction occurs at the cathode generating high local pH. Afterwards, the formed phosphate reacts with calcium ions to generate calcium phosphate precipitates on the cathode, allowing further P recovery (Lei et al., 2020).

Removal of organophosphate esters with C-O-P bonds
Organophosphorus flame retardants (OPFRs), including organophosphate esters (OPEs) with C-O-P bonds, are rapidly growing as important alternatives to brominated flame retardants (Moser et al., 2015). OPEs have been widely used in building materials, chemicals, upholstered furniture, electronics, and other industries (Liang et al., 2019;Wei et al., 2015). The annual use of OPEs is estimated to reach tens of thousands of tons, and its presence in urban sewage, river water, groundwater, and drinking water range from several thousands to tens of thousands ng/L . Most of the detected OPEs are stable (Wang et al., 2018a). Conventional water treatment technologies including physical membrane separation and biochemical methods are normally not effective for the removal of OPEs. Therefore, the development of new effective strategy for OPEs control is desired.

Adsorption
Various adsorbents, e.g., resins and activated carbons, have been investigated for the adsorption of OPEs (Wang et al., 2018a(Wang et al., , 2018b. During the adsorption, several interactions might be involved, including hydrophobic interactions, electrostatic interaction, van der Waals interaction, p-p interaction, and hydrogen bonding. However, different from other P species (phosphate and phosphonates), it is difficult to find suitable adsorbents that could provide specific interactions with OPEs. Nevertheless, adsorption remains as a potential choice for pretreatment before further degradation of these OPEs due to their low concentration in wastewaters.

Hydrolysis
OPEs might undergo hydrolysis due to the characteristic C-O-P bonds. There are many studies examining the hydrolysis of OPEs agents and pesticides by enzymes (Bigley et al., 2016;Le et al., 2015).  reported the hydrolysis of organophosphate esters flame retardants (OPFRs) via a mixture of lactonases designed by semi-rational evolution engineering . Although hydrolysis catalyzed by enzymes could be fast, the reaction is limited by the reactivity of the enzymes, which strongly depends on the operational conditions including temperature and pH. Moreover, the delicate and complicated preparation procedures is generally cost ineffective. Interestingly, Fang et al. (2018) reported that OPEs flame retardants undergo adsorption on the surface of metal oxides, followed by subsequent hydrolysis process (Fang et al., 2018) (Figure 4d).

Oxidation
There are many studies reporting the removal of OPEs from water using various oxidation processes, mainly through the generation of oxidative radicals (Liu, Ye et al., 2018;Yu et al., 2019). For examples, Yuan et al. (2015) demonstrated that UV/H 2 O 2 is more powerful than ozonation for the removal of OPFRs in a municipal secondary effluent (Yuan et al., 2015). Ye et al. (2017) reported effective degradation of tris(2-chloroethyl) phosphate (TCEP), a representative OPFR, using a UV/TiO 2 photocatalytic system (Ye et al., 2017). Xu et al. (2017) showed that the UV/ peroxymonosulfate system could achieve a high removal efficiency of TCEP (Xu et al., 2017).

Reduction
A few studies have demonstrated that OPEs could be reduced under anoxic conditions. For examples, Saint-Hilaire and Jans (2013) reported the reductive degradation of three Cl-OPEs using reduced sulfur species under alkaline condition. Recently, Li et al. (2020) showed the efficient removal of TCEP by FeS with cetyltrimethylammonium bromide (CTAB) as reducing agent , and the removal of three Cl-OPEs using zero valent iron (nZVI) and S-nZVI as reducing agents .

Conclusions and perspectives
In this review we have briefly introduced the currently available techniques for the analysis of different P species. Clearly, the combination of two or more techniques is an effective strategy to overcome the limitation of an individual technique and to realize the complementary advantages. Moreover, the recent advance in novel nanomaterials has driven the optimization of the preconcentration process in P analysis, effectively improving the analysis speed, convenience, sensitivity, and selectivity, especially for trace OP species. Additionally, the development of non-targeted screening analysis methods in recent years has allowed the identification of more unknown OP species. Even though, there are still several challenges for the speciation analysis of P. Generally, new analytical techniques with simple pretreatment and suitability for in situ rapid P monitoring are strongly desired to improve the analysis efficiency and reduce the operational cost, and new in situ techniques could find applications in freshwater eutrophication monitoring, sewage treatment plant discharge monitoring, and marine monitoring. In addition, the highly selective and sensitive analysis for trace P species with the interference in complex water matrix is still challenging, especially in environments where the concentration of P species is significantly lower than the interfering ions. A plausible solution is to develop new preconcentration materials and analysis methods that are highly selective toward the target P species. In the future, these preconcentration materials and methods could be extended to more scenarios, including the in-depth cognition of complex sewage water systems, as well as the cognition of the degradation, migration, and transformation of phosphorus species. With respect to the emerging new OP species revealed by non-targeted screening analysis, it is important to combine their structural information with the subsequent toxicological analysis, to elucidate their ecotoxicological effect and to provide additional cognition of all P speciation for prioritizing P management and control.
We have also presented a brief summarization of available removal strategies for different P species. Apparently, the removal of inorganic phosphate especially orthophosphate has received the most intensive research with the corresponding strategies well established. In contrast, the removal strategies for OP are much less reported and lack systematical methodology. OP compounds possess complicated structures, posing challenges in terms of selectivity to the removal approaches via physicochemical interactions, such as adsorption. Chemical oxidation approaches also suffer from low selectivity and efficiency for the removal of OP compounds due to their low concentration in complicated water matrix. Consequently, it is of practical significance to develop methods with high efficiency and selectivity for the removal of OP compounds in future, e.g., selective adsorption and selective oxidation. Currently, there are examples concerning the selective removal of trace OP in aqueous systems. However, it is still far to fulfill the selectivity for practical application given the vastly different and rather complicated water matrices in real scenarios. Hence, the design of selective phosphorous treatment processes should be case-specific in real applications. Moreover, we can also take advantage of the nature of the water per se, such as pH, complexation capability of the target pollutant, and coexisting active metals as catalysts, etc. during the development of phosphorous treatment strategies.
The purpose of this review by providing a combinative overview of the analysis of different P species in water and the corresponding removal strategies is to highlight an important direction for P management in the future, that is, the development of P species-orientated removal technologies with higher selectivity and efficiency. The efficiency of a treatment process depends highly on the features of target P species, including adsorption property, redox property, etc. Therefore, the identification of the common features of the target P species, for examples, functional groups, hydrophilicity/hydrophobicity, redox potential, and size, is necessary to ensure a suitable match with the removal technology for high reactivity and selectivity. The ultimate goal of developing phosphorous removal strategies in lab is to implement them in real water treatment, requiring the consideration of the vastly different and rather complicated water matrices in practical scenarios. The accomplishment of this goal is based on the advances of the analytic techniques which could provide precise information of the P species, and could be achieved through several means including the nexus of multiple technologies and more delicate designs of functional materials.