Occurrence of titanium dioxide nanoparticle in Taihu Lake (China) and its removal at a full-scale drinking water treatment plant

The occurrence of titanium dioxide nanoparticle (TNP), an emerging contaminant, in Taihu Lake of China was investigated. Ti was present at a concentration of 224 ± 59 μg/L in the water samples collected from a water source in east Taihu Lake. Approximately 0.19% of the Ti-containing matter was at the nano-scale. Scanning Electron Microscope analysis verified the existence of Ti-containing components, such as TiOx and FeTiOx. Furthermore, Ti K-edge X-ray absorption near-edge structure spectroscopy was used to detect the phase composition of nano-scaled Ti-containing matter. The spectra showed the three characteristic peaks of TiO2 in the samples, suggesting the occurrence of TNP in Taihu Lake. A least-squares linear combination fitting analysis indicated that the TNP concentration in the water source was ~0.77 μg/L in water and ~0.85 μg/g-dry in sediment. The removal performance of the TNP at a full-scale conventional drinking water treatment plant indicated that ~61% of TNP was removed via coagulation/sediment, sand filtration, and disinfection/clear water reservoir. The coagulation/sediment process accounted for approximately 70% of the total removed TNP. The finished water contained ~ 0.30 μg/L TNP. This study is the first that reported the presence and transport of TNP in a full-scale drinking water treatment system.


Introduction
Engineered nanoparticles (ENPs) represent an important class of emerging contaminants due to their release into the environment (Gottschalk et al. 2015;Kiser et al. 2009;Mueller and Nowack 2008). Compared with the bulk counterpart, ENPs possess a significant difference in their physical and chemical properties (Nel et al. 2006), which stimulates the wide application of ENPs globally and results in their toxicity in the environment (Jomini et al. 2015;Yu et al. 2015).
Titanium dioxide nanoparticle (TNP) is one of the most commonly produced ENPs and is extensively used in photovoltaic cells, cosmetics, catalysts, coatings and cleaning agents, plastics, dietary supplements, etc. (Weir et al. 2012). Most of the environmental risk assessments of ENPs report that TNP is one of the highest risk ENPs in the aqueous environment based on their toxicology data and the predicted environmentally relevant concentration (Gottschalk et al. 2015;Gottschalk et al. 2009;Mueller and Nowack 2008).
Taihu Lake is located in the core area of the Yangtze River delta in China. It is one of the most important water sources of the ambient cities, including Shanghai, Suzhou, Nanjing, and Wuxi. Its drainage area is 36895 km 2 , accounting for 0.4% of the total area of China (Qiu et al. 2004). Taihu Lake supplies water to 2.9% of the population and provides more than 10% of the gross domestic production (GDP) in China (Bureau of Taihu Lake Basin 2013). However, in recent years, as the economy has developed, heavier pollution and the eutrophication of lakes have worsened, and the primary source of Taihu Lake pollutants is considered an accumulation of nutrient-rich sewage and agricultural runoff into the shallow lake (Guo 2007). A similar exposure pathway of TNP was proposed in previous studies (Gottschalk et al. 2015; Responsible Editor: Thomas D. Bucheli * Shuili Yu liliu8566@163.com 1 Gottschalk et al. 2009;Mueller and Nowack 2008), which implied the presence of TNP in the lake along with sewage and agricultural runoff. However, the status quo concentration of TNP in Taihu Lake remains unknown.
To determine the TNP concentration in the environment, the most commonly used method to date is exposure modeling (Gottschalk et al. 2015). For example, Mueller et al. modeled the quantities of nanosilver, TNP, and carbon nanotubes released in Switzerland using a life-cycle perspective and reported that their expected concentrations in water were 0.03-0.08, 0.7-16, and 0.5-0.8 ng/L, respectively (Mueller and Nowack 2008). However, the actual TNP concentration may be different from the modeled values due to its complicated transport behavior in the environment (Battin et al. 2009). Moreover, the models are always performed under the assumption that TNP is naturally non-existent in the environment, which does not agree with the facts (Grafe et al. 2014). For the direct detection of TNP in the environment, studies have adopted element quantification (Donovan et al. 2016;Kiser et al. 2009;Weir et al. 2012). However, the element quantification technologies including ICP-MS, ICP-OES, and SP-ICP-MS could not fight off the effect of the other Ti-containing compounds such as ilmenite and titanium chloride which have been proved to be present widely in the environment (Gaillard et al. 2001;Grafe et al. 2014). X-ray absorption near edge spectroscopy (XANES) is a powerful tool to investigate the local coordination environment of most of the metals in the environment, which could distinguish the various compounds, even those possessing the same metallic element (Eley et al. 1986;Gaillard et al. 2001;Manceau et al. 2002). Based on the linear fitting of the characterized pre-edge peaks of Ti K-edge XANES spectra, Lee et al. calculated the phase ratio of anatase to rutile for TNP mixtures (Lee et al. 2003). For environmental samples (wastewater and sediment), Tong et al. reported credible Ti fitting results using standard substances containing anatase, rutile, brookite, amorphous TiO 2 , and ilmenite (Tong et al. 2015b). Although their study did not focus on TiO 2 on the nano-scale, XANES was shown to be a useful technique to detect the TiO 2 phase ratio in the environment.
Drinking water treatment is one of the main strategies to prevent the ingestion of harmful contaminants, including ENPs, from an aqueous environment . Conventional drinking water treatments typically involve coagulation/sediment, sand filtration, and disinfection/clear water reservoir. The removal efficiency of ENPs using these technologies has been investigated extensively on the laboratory scale Hyung and Kim 2009;Li et al. 2013). However, the removal performance of ENPs during full-scale drinking water treatments is largely unknown.
In this study, the occurrence of TNP in a water source of east Taihu Lake was investigated via acid dissolution, size grading, inductively coupled plasma mass spectrometry (ICP-MS), scanning electron microscopy (SEM), and XANES. Furthermore, the removal performance of TNP at a full-scale drinking water treatment plant using the conventional treatments of coagulation/sediment, sand filtration, and disinfection/clear water reservoir was also studied. Our study represents the first report of the presence and the transport of TNP in such environmental samples, providing new insights for assessing the potential risks of TNP to people via the drinking water system.

Sampling protocol
The water source (WS: 31°1′20″N/120°26′40″E) is located in east Taihu Lake, serving as the water source for the Wujiang Drinking Water Plant (WDP). WDP supplies more than 200 billion liters of drinking water per year and meets the water demands of approximately 420,000 households, using the conventional drinking water treatments, i.e., coagulation/sediment, sand filtration, and disinfection/clear water reservoir (Fig. 1).
To investigate the occurrence of TNP, water samples were collected at the WS 16 times over a one-month sampling period (December 10, 2014, to January 11, 2015, at a depth of 50 cm under the water surface. Sediments of the WS (WSS) were also collected under the water sampling points (top 30 cm sediment). Locations of the sites for water/ sediment sampling relative to WDP were detailed in the supporting information (Fig. S1). To study the removal performance of TNP in the drinking water treatment plant, effluent from various stages of the treatments of WDP was randomly collected five times over another onemonth sampling period (January 11, 2015, to February 15, 2015 (Fig. 1). The hydraulic retention time in the treatments was considered during the water sampling. For each sampling point, three replicates were collected. After sample collection, the samples were sealed in glass bottles and stored below 4°C until analysis.

Sample pretreatment
To investigate the acid solubility of the Ti-containing matter, the water samples were injected with HCl and HNO 3 to achieve the pH value of 2 and then centrifuged at 24149 g (12000 rpm) for 30 min (Jouan KR25i, Thermo Fisher Scientific, USA) to obtain the supernatant. The Ti concentration in the supernatant was then determined to evaluate the acid solubility of the Ti-containing matter. To evaluate the size fractionation of the Ti-containing matter, the water samples were treated with ultrasonication and then filtered through 5, 1.2, 0.8, 0.45, 0.1 μm PTFE membranes (Millipore, UK Ltd.) using a vacuum filtration device (Xinghua Glass Instrument Plant, Zhengzhou, China), respectively. The filtrate of 0.1 μm PTFE membranes was further centrifuged at 24149 g (12000 rpm) for 30 min (Jouan KR25i, Thermo Fisher Scientific, USA) to obtain the supernatant. The centrifugal treatment was estimated that the TiO 2 particles larger than 1 nm could be separated as calculated by Stokes law (Perret et al. 1994). The Ti concentrations in the filtrates or the supernatant were then determined to evaluate the size grading of the Ti-containing matter. The recovery of the size fraction was testified to be 80.4-132%. To prepare the samples for XANES analysis, 100 L of each water sample was centrifuged at 24149 g (12000 rpm) for 30 min to obtain the concentrate (Jouan KR25i, Thermo Fisher Scientific, USA). Then, the concentrate was ultrasonically treated and filtered through a 0.1 μm PTFE membrane (Millipore, UK Ltd.) using vacuum filtration. The filtrate was centrifuged at 24149 g (12000 rpm) for 30 min. After the supernatant was discarded, the obtained sludge was dried for 12 h at 75°C, which is a condition that does not alter the phase composition according to a previous study (Tong et al. 2015b). The dried sludge was stored in a drying oven until analysis.
The sediments were subjected to the same procedures as the water samples after being dissolved in ultrapure water for acid solubility, size fractionation, and XANES analysis.

Titanium quantification
Titanium quantification was performed after chemical digestion following the method reported by the previous literature (Khosravi et al. 2012). Briefly, an appropriate amount of samples (1 mL for the WS and S1 samples, 100-150 mL for the other water samples) was dried on a plate digestion apparatus (DB-3EFS, Lichen Technology Co., LTD, China). Then, the samples were mixed with 5 mL mixtures (with a weight ratio of 7:4 of sulfate: ammonium sulfate) and fired on the plate digestion apparatus at 300°C for 30 min. Then, the mixtures were diluted to 50 mL with ultrapure water and analyzed via ICP-MS (Icap Q, Thermo Fisher Scientific, USA). The Ti recovery was 95 ± 3% when 0.3 μg of nano-scaled anatase was spiked into 10 mL of the raw water. The lowest limit of detection was calculated as 0.15 μg/L following the method reported by (Khosravi et al. 2012).

XANES analysis
Ti K-edge XANES spectra were acquired at the BL14W1 XAFS beamline at the Shanghai Synchrotron Radiation Facility. Ilmenite (FeTiO 3 ), Anatase, Rutile, Brookite, and Amorphous TiO 2 were adopted as the standards to represent the components of Ti-containing matter in the samples. Ilmenite (catalog no. 400874), Anatase (catalog no. 232033), Rutile (catalog no. 224227), and Brookite (catalog no. 791326) were purchased from Sigma Aldrich Co., USA. Amorphous TiO 2 was synthesized in the laboratory according to the procedure described by Wang et al. (Wang et al. 2003). Briefly, a 0.5:1 mixture of dodecylamine and tetrabutyl titanate was stirred and warmed until a homogeneous colorless solution was obtained. To this solution, distilled deionized water was added and the mixture was stirred for 30 min. The reaction mixture was aged at ambient temperature for 24 h, and then further aged at 100°C for 48 h in an autoclave. The as-synthesized products were filtered and washed with the mixture solution of ethanol and distilled deionized water to remove the surfactant, and dried at ambient temperature.
The dried sludge obtained from the samples was ground with lithium fluoride at a weight ratio of 1:10 due to the scarcity of the samples and then mounted on Kapton tape (DuPont) for XANES spectroscopy analysis.
The XANES spectra of all the standards were collected in the transmission mode using Oxford ionization chambers with 296-mm path lengths due to the high abundance of Ti (> 10%), while the spectra of the environmental samples were determined in the fluorescence mode using a germanium array detector. During the measurement, the synchrotron was operated at the energy of 3.5 GeV and a current between 150 and 210 mA. Background subtraction and edge normalization were performed using the Athena program (version 0.9.17). The TiO 2 concentration was calculated according to the method reported by Tong et al. (2015b). Briefly, least-squares linear combination fitting (LCF) was performed from −6.4 to 80 eV relative to the absorption edge (4966.4 eV) to quantify Ti speciation. Then, the TiO 2 concentration was calculated using the following equation.
Where 79.87 is the molecular weight of TiO 2 , 47.87 is the molecular weight of Ti, and ilmenite represents the percentage of ilmenite in the samples.

Results and discussion
Occurrence of TNP in the water source in east Taihu Lake As shown in Fig. 2, the Ti concentration in WS substantially shifted during the sampling period, achieving an average Ti concentration of 224 ± 59 μg/L during the sampling period. The Ti concentration in sediments was relatively stable (0.70 ± 0.06 mg-Ti/g-dry). Similarly, a few studies have reported the Ti concentration in natural water and sediment. Tong et al. reported the Ti concentration in the sediment of a North Shore Channel in Skokie (USA) as 1.98 ± 0.04 mg-Ti/g-dry (Tong et al. 2015b), and from < 5 to 15 μg/L Ti in the effluent of a wastewater reclamation facility (Kiser et al. 2009). Thus, the Ti concentration in the water of Taihu Lake is significantly higher than the previously reported values, whereas the Ti concentration in sediment is similar. Principle component analysis (PAC) revealed the close correlation of Ti and turbidity (Fig. S2), indicating Ti in water may present in the state of colloid or suspended solids. Accordingly, the high contents of Ti in the water of Taihu Lake are much likely due to the hydraulic disturbance of the sediment as the average depth is merely 1.9 m (Yang et al. 2008).
Due to the insoluble property of TiO 2 in the strong acids at room pressure and temperature (Khosravi et al. 2012), the acid-dissolved fraction was determined to estimate the presence of TiO 2 in the samples. It was found that the major fraction of Ti-containing matter in both WS and WSS could not be dissolved either in HCl or in HNO 3 implying that the main component of the Ti-containing matter was present as acid-insoluble matter such as TiO 2 , FeTiO 3 , Ti(OH) 3 , and Ti 2 O 3 as shown in (Fig. S3). Interestingly, HNO 3 could not decrease the insoluble Ti percentage as much as HCl implying the coexistence of the Ti-containing matter which could be oxidized to a species that is insoluble in acid, such as TiO. After being decomposed by an ultrasonic treatment, the size fractionation of Ti-containing matter in water and sediment of the water source was performed (Fig. 3). The main Ticontaining matter in WS and WSS was present in the > 5 μm fraction (approximately 77.3% for WS and 86.2% for WSS). The Ti concentrations between 1-100 nm in WS and WSS were 0.49 ± 0.62 μg/L and 0.57 ± 0.17 μg/g-dry, respectively. This result confirmed the presence of nano-scaled Ticontaining matter in WS and WSS. SEM analysis confirmed the presence of TiO X and FeTiO X in WS as shown in (Fig.  S4). Most of the Ti-containing particles are present in the μm size, whereas some Ti-containing particles are aggregates of nano-scaled particles.
As shown in Fig. 4, the spectra of WS and WSS show the three characteristic peaks in the pre-edge (labeled by the imaginary lines) and the characteristic shapes after edge (Fig. 4a). The three peaks are due to the Ti 4p-Ti 3d hybridization and Ti 4p-O 2p hybridization in TiO 2 , suggesting the presence of TiO 2 in the samples (Lee et al. 2003). Therefore, XANES further confirmed the presence of TNP in the water and sediment of east Taihu Lake.
According to the general estimation of the Ti-containing matter of east Taihu Lake in the above experiments, as well as the previous fitting results of Ti-containing matter in the environment (Tong et al. 2015b), we chose Ilmenite (FeTiO 3 ), Anatase, Rutile, Brookite, and Amorphous TiO 2 as the standards to represent the components of Ti-containing matter in the samples. The XANES spectra of the Ti reference compounds, on which the linear combination fitting (LCF) was based, are shown in the supporting information (Fig. S5). LCF was performed from −6.4 to 80 eV relative to the absorption edge (4966.4 eV) to quantify Ti speciation, and the data and results of the curve fitting are presented in Fig. 4b and c. The Combined with the Ti concentration on the nano-scale, the TNP concentration in the WS and WSS was calculated using Eq. 1. The TNP concentration in WS was~0.77 μg/L, with a crystal composition of 30.0% amorphous, 33.5% anatase, and 28.5% rutile. The TNP concentration in WSS was~0.85 μg/gdry, with a crystal composition of 15% amorphous, 33% anatase, and 41% rutile.
It was found that no brookite was detected in the water samples and sediment. The natural scarcity of Brookite and its poorer production due to the lack of proper usage compared with the other phases may be the primary reasons for this result (Allen et al. 2008). It can also be seen that the amorphous component accounted for a large fraction of the detected TNP in the water samples and sediment. Generally, amorphous TiO 2 has not been widely used in product applications due to its poor photocatalytic activity. In view of the fact that the amorphous TiO 2 is present in natural minerals (Force 1991), the detected amorphous fraction may come from the natural minerals. Similarly, approximately 24% of amorphous TiO 2 was present in the sediments of a river (Tong et al. 2015b). In addition, rutile and anatase were the main components in TiO 2 . Rutile is widely present in nature as a TiO 2 -containing mineral (Yu et al. 2013). Moreover, rutile has been massively produced for several decades. Similarly, anatase has a high photocatalytic activity and is used widely in products. Therefore, it is not surprising that rutile and anatase are detected in the samples. Previous studies reported similar environmentally relevant concentrations of TNP based on the exposure modeling (Boxall et al. 2007;Gottschalk et al. 2009;Mueller and Nowack 2008).

Removal of TNP at the drinking water treatment plant
As shown in Fig. 5, a significant fraction of both Ti and nanoscaled Ti was removed via coagulation/sediment (99.1% for Fig. 3 Size fractionation of Ti-containing matter (mean ± SD) in the water samples (left y-axis) and sediments (right y-axis) collected from a water source in east Taihu Lake. WS and WSS represent the representative water samples and sediments of the water source in east Taihu Lake, respectively (n=3) Fig. 4 Normalized Ti K-edge XANES spectra (a), fitting results of the normalized Ti K-edge XANES pre-edge spectra (b), and the Ti speciation composition of nano-scaled Ti-containing matter (c) in the representative water sample and sediment collected from a water source in east Taihu Lake Ti and 58.8% for nano-scaled Ti). The coagulation/sediment accounted for 61.5% in the total removed nano-scaled Ti. The removed fraction via sand filtration was 20.9%, whereas the disinfection/clear water reservoir removed 17.5% of nanoscaled Ti-containing matter. Additionally, the removal performance deteriorated as the size of the Ti-containing matter decreased. Similarly, Ti larger than 0.7 μm was well removed by a wastewater treatment plant in the USA, while the < 0.7 μm fraction was poorly removed and was present in effluents in concentrations ranging from < 5 to 15 μg/L (Kiser et al. 2009).
As shown in Fig. 6, all the spectra of the effluent at various stages at the drinking water plant contained the three characteristic TiO 2 peaks in the pre-edge of TiO 2 (Fig. 6a), which indicated the presence of TNP at the various stages at the drinking water plant. LCF was performed from −6.4 to 80 eV relative to the absorption edge (4966.4 eV) to quantify Ti speciation, and the data and results of the curve fitting are presented in Fig. 6b and c. The best LCF yielded 0.00042, 0.00051, and 0.0031 of the R-factor for the samples collected after coagulation/sediment, sand filtration, and the effluent of WDP, respectively (the fitting result of the WS was used as the influent Ti species composition). Combined with the Ti concentration on the nano-scale, the TNP concentration in the influent was calculated as~0.77 μg/L, whereas the TNP concentration in the effluent was~0.30 μg/L, with a crystal composition of 21% amorphous, 47% anatase, and 20% rutile as shown in Fig. 6d. Similarly, O'Brien et al. predicted that the mean TNP concentrations for the local drinking water treatment schemes ranged from 44.1 to 1450 ng/L (O'Brien and Cummins 2010).
The overall removal of the TNP was~61%. The coagulation/sediment, sand filtration, and disinfection/storage tank accounted for approximately 70%, 15%, and 15% of the total removed TNP, respectively. The effective removal of TNP during coagulation/sediment was also found in the jar test (Abbott Chalew et al. 2013), and the electrostatic interaction was considered to be the main removal mechanism. Sand filtration did not provide efficient TNP removal. Similarly, Rottman et al. reported that the batch isotherms revealed poor adsorption between quartz sand and TNP, and the quartz sand provided nearly zero retention of a 50 mg TiO 2 per one liter stream in a column experiment (Rottman et al. 2013). It is should be noted that the other factors such as water treatment processes, water quality, and TiO 2 concentration could also influence the removal of TNP from water. Donovan et al. reported that over 95% of the nano-TiO 2 were removed by the drinking water treatment using lime softening, alum coagulation, filtration, and disinfection; among this, the lime softening accounted for 89%, while the other treatments for 11% (Donovan et al. 2016). The removal percentages of nano-TiO 2 suspended in the deionized water and the raw waters shifted from 52.6-97.3% to 9.3-35.1% when the nano-TiO 2 concentration decreased from 10 mg/L to 0.1 mg/L (Chang et al. 2017).
In addition, different removal performances for the various TiO 2 crystals were observed during the drinking water treatments. During coagulation/sediment, the removal performance decreased in the order of amorphous, rutile, and anatase. Previous studies have reported the different characteristics of some types of TiO 2 crystals in the aqueous environment. For example, Liu et al. showed that crystallinity and morphology are not influential factors in determining the stability of TNP suspensions; however, the differences in their chemical compositions, notably, the varying concentrations of impurities (i.e., silicon and phosphorus) in the pristine materials, determined the surface charge, therefore determining the sedimentation and aggregation of TNP in the aqueous phase (Liu et al. 2011). Thus, the different removal performances during coagulation may be caused by the varying surface charge of the TiO 2 crystals.
In terms of exposure pathways to humans, TNP in drinking water is most likely to gain exposure via the gastrointestinal and skin routes (Hagens et al. 2007;Lomer et al. 2002). Previous studies producing lethal and sub-lethal effects in animal toxicity tests revealed the toxicity thresholds of 5 g/kg for rat (Wang et al. 2007), 10 g/m 3 for Daphnia magna (Lovern and Klaper 2006), 1 g/m 3 for Rainbow trout (Federici et al. 2007), and 40 g/m 3 for Algae (Hund-Rinke and Simon 2006).
According to China's water specifications, the drinking water consumption amounts per person per day is 75-220 L including washing and bathing (Ministry of Construction of the People's Republic of China 2002). The annual accessible TNP through the gastrointestinal and skin routes per person is 8.20-24.09 mg assuming the TNP concentration in the water was 0.30 μg/L. The exposure quantity is several orders lower than the abovementioned toxicity thresholds. Moreover, this intake concentration is significantly smaller than the concentration to which humans are exposed orally through daily food (73-730 mg/kg body weight per year of nano-TiO 2 /E171 in the USA) (Weir et al. 2012). Thus, exposure to drinking water is extremely unlikely to result in a nano-specific toxicological response (O'Brien and Cummins 2010).
However, the presence of TNP in the drinking water system may result in secondary pollutants. For example, when TNP is present in the biological treatment, e.g., biological activated carbon filter, it may affect the pollutant removal performance because of its eco-toxicity (Li et al. 2014). For example, Liu et al. reported that all of 10 μg/L, 100 μg/L, 1 mg/L, 10 mg/L, and 100 mg/L TiO 2 NPs showed a significant negative effect on the nitrification potential of the microbes in biological activated carbon filters for drinking water after 36 h of exposure (2016). Additionally, TNP may exert an influence on the disinfection effectiveness due to its affinity for bacteria (Wei 2011). Furthermore, TNP could alter the potential biological uptake of heavy metal ions, such as arsenic and lead (Miao et al. 2015;Sun et al. 2009). The coexistence of different types of nanoparticles was also reported to alter the original toxicity of individual nanoparticles (Tong et al. 2015a). Thus, a comprehensive assessment of potential TNP toxicity in the drinking water system still requires careful integration of complex physicochemical interactions between TNP and other components in water.

Conclusions
In this study, a comprehensive and systematic survey was performed to evaluate the occurrence of TNP in the water source of east Taihu Lake (China) and its removal performance at a conventional full-scale drinking water treatment plant. Based on the combination of size grading, ICP-MS, and XANES, the TNP in the water source was calculated as~0.77 μg/L in water and~0.85 μg/g-dry in sediment. The removal performance of TNP at a full-scale drinking water treatment plant using the conventional treatments of coagulation/sediment, sand filtration, and disinfection/clear water reservoir was also studied. The overall removal of TNP was~61%. However,~0.30 μg/L TNP was still detected in the treated waters, although at trace levels, suggesting the incomplete removal of TNP through the conventional process. Additionally, further studies are required to evaluate the secondary pollutants caused by the presence of TNP in a drinking water system. Fig. 6 Normalized Ti K-edge XANES spectra (a), fitting results of the normalized Ti K-edge XANES pre-edge spectra (b), the Ti speciation composition of nano-scaled Ti-containing matter (c), and crystal composition of the TNP (d) in the effluents after various stages in the Wujiang Drinking Water Plant