Analysis of organophosphorus flame retardants in sewage water treatment plant influent and effluent samples

(1)

Institute for Biodiversity and Ecosystem Dynamics (IBED)

Analysis of organophosphorus flame

retardants in sewage water treatment

plant influent and effluent samples

Bachelor Thesis Chemistry

June 30, 2015

Author:

Mimi den Uijl

Student ID:

10410465 (UvA) / 2544890 (VU)

Supervisor:

dr. J.R. Parsons

(2)

1.

Popular Abstract

Flame retardants play a key role in the prevention and delay of fire. Especially with the development and application of polymers everywhere around us, flame retardants have become of great interest in the polymer industry. There are four types of flame retardants: : inorganic flame retardants, organohalogenated flame retardants, organophosphorus flame retardants and nitrogen-based flame retardants. Brominated flame retardants were most popular, because they were relatively cheap and had decomposition temperatures in the combustion range of the polymers. But brominated flame retardants seem to have toxic properties so other types of flame retardants are applied. Among those relatively new flame retardants are organophosphorous flame retardants (OPFRs). In this research three OPFRs were studied: triphenylphosphate (TPP), resornicol bis(diphenyl phosphate) (RDP) and bisphenol A bis(diphenyl phosphate) (BDP). A method was optimized to analyse concentration of these three OPFRs in water samples of a sewage treatment plant. This method was developed to see if there are any OPFRs in water and if so, if there are any concentrations differences between influent and effluent samples. The procedure consistsed of an extraction step with Solid Phase Extraction (SPE) and an analysis with LC-MS/MS of the eluates from the SPE. When the optimized method was applied, traces of TPP have been found in influent and effluent samples. Influent gave concentrations of 0.50 µg/L and the concentration in effluent decreased to 0.20/0.30 µg/L. This decrease could indicate the removal of TPP, but the experiment has to be repeated to be reliable. The presence of RDP and BDP was low or not present at all, even though the method worked for all three OPFRs in the optimization process. A possible reason for the low concentrations is the lack of a good liquid/liquid extraction method, because RDP and BDP are expected to adsorb to the sludge in the samples. The optimized method should be adjusted to be more suitable for the detection of RDP and BDP.

2.

Abstract

Flame retardants play a key role in the prevention and delay of fire. Especially with the development and application of polymers, flame retardants have become of great interest in the polymer industry. There are four types of flame retardants: : inorganic flame retardants,

organohalogenated flame retardants, organophosphorus flame retardants and nitrogen-based flame retardants. Brominated flame retardants dominated the market, because they were relatively cheap and had decomposition temperatures in the combustion range of the polymers. But brominated flame retardants seem to have toxic properties so other types of flame retardants are applied. Among those relatively new flame retardants are organophosphorous flame retardants (OPFRs). In this research three OPFRs are studied: triphenylphosphate (TPP), resornicol bis(diphenyl phosphate) (RDP) and bisphenol A bis(diphenyl phosphate) (BDP). A method was optimized to analyse concentration of these three OPFRs in water samples of a sewage treatment plant. This method was developed to see if there are any OPFRs in water and if so, if there are any concentrations differences between influent and effluent samples. The samples will be collected in polypropylene bottles and will undergo Solid Phase Extraction (SPE). Adsorption will be minimized with an extraction of the polypropylene bottles. Eluates were analysed with LC-MS/MS. Traces of TPP were found: influent gave concentrations of 0.50 µg/L and the concentration in effluent decreased to 0.20/0.30 µg/L. The presence of RDP and BDP was low or not present at all, even though the method worked for all three OPFRs in the optimization process. A possible reason for the low concentrations is the lack of a good liquid/liquid extraction method, because RDP and BDP are expected to adsorb to the sludge in the samples. The optimized method should be adjusted to be more suitable for the detection of RDP and BDP.

(3)

3. Introduction

3.1 Flame Retardants

Fire is one of the greatest dangers of mankind, because it spreads quick and is uncontrollable. The idea to prevent fire from happening or to delay the spread of fire, dates back to the old Egyptians and the Romans.1_{They used alum and vinegar to prevent the wood from catching fire.}2_{Some 2000} years later with the development and application of polymer products, which are relatively

combustible, there is a great need for flame retardants.3_{Fire retardants prevent or delay fire by} interfering in the fire cycle.4_{There are three necessities for fire: heat, fuel and air and flame retardants} interfere in the cycle by stopping or by weakening the supply of one of those three components. There are four classes of FRs: inorganic flame retardants, organohalogenated flame retardants,

organophosphorus flame retardants and nitrogen-based flame retardants.5

Inorganic flame retardants, the oldest group of FRs, decompose at high temperatures and produce non-flammable gasses. In this way, the flame retardant shields the surface of the polymer against oxygen and prevents fire. The group of organohalogenated flame retardants, especially brominated flame retardants (BFRs) is the largest market group of FRs: the demand for BFRs in 1999 was estimated over 200000 metric tons and this number has only gone up.6,7_{BFRs interfere with the} fire cycle in the gas phase to prevent fire. Organophosphorus flame retardants (OPFRs) interfere with reactions in the condensed phase.5_{Nitrogen-based flame retardants are the smallest group of FRs and} little is known about their preventing mechanism.5_{There are two methods of applying the FR: a} reactive and an additive method. Reactive flame retardants are chemically bond to the product and the additive flame retardants are blended with the polymers. Because additive FRs aren’t bonded to the product, they are more likely to end up in the environment.3

3.2 Brominated Flame Retardants

The reason for the great market share of BFRs is that they are relatively cheap and have a high performance efficiency, since they are effective in the range of decomposition temperatures of plastic between 150°C and 400°C.5,6_{Because of recent discoveries of toxicity of those FRs, the European} union restricted the use of a great number of BFRs.6,8_{BFRs appeared to cause liver disfunctions}9_, endocrine disruptions10,11_{and neurotoxicity}6,9_{. Little is known about the toxicity of the other flame} retardants. The toxicity is a big problem because recent studies have discovered traces of flame retardants in house dust12,13_{, wastewater}14-16_{, sewage sludge}7,17_{, sediment}7_{, human adipose tissue}7_and even in breast milk7,18,19_{. The accumulation of the FRs in breast milk poses a risk for the unborn child,} which is more prone to possible dangers.

3.3 Other options: Organophosphorus flame retardants

Some BFRs were listed under the Stockholm Convention on Persistent Organic Pollutants and others have been restricted.20_{For this reason the market of FRs shifts from BFRs to different} flame retardants like organophosphorus flame retardants (OPFRs).21_{As mentioned before, little is} known about its toxicity, its biodegradability and its behaviour in aquatic systems.

In this research three additive22_{OPFRs will be studied: triphenylphosphate (TPP), resornicol} bis(diphenyl phosphate) (RDP) and bisphenol A bis(diphenyl phosphate) (BDP). All three OPFRs have one or two phosphate centres surrounded by multiple phenyl-groups and are shown in Figure 1. All three OPFRs act through flame poisoning in the gas phase, whereby TPP and RDP perform slightly better than BDP.23_{Nonetheless, BDP shows a clear condensed phase action and TPP and RDP}

(6)

show none. The combination between gas and condensed phase action makes BDP superior in fire load and decreasing flame speed.24_{Compared to TPP, less is known about the properties of BDP and} RDP.

3.4. Properties

of TPP, BDP

and RDP

When searching for the properties of TPP, BDP and RDP, it leads to varied results. Especially for RDP and BDP, not much is known about their (toxic) properties. The toxic properties that are mentioned in the next section are not completely reliable: a lot of numbers are estimates on other OPFRs. All three substances don’t seem to have toxic effects on humans, only in extreme doses: RDP has an inhalation toxicity of 12.4 mg/L and TPP has an inhalation toxicity between 0.36 – 290 mg/L. The LD50 of TPP when ingested was measured on guinea pigs, rats, mice and rabbits and gave values between 1320 – 20000 mg/kg body weight. The aquatic toxicity and chronic toxicity of BDP is minimal. On the other hand, TPP is acutely toxic to aquatic organisms.22_{This could be because of the} stability under neutral and acidic conditions in water.25_{TPP is strongly adsorbed in soil and sediment} and degrades quick in the atmosphere via indirect photolysis (t1/2 = 12h). The DT50 value in soil ranges from 21 – 37 days. The distribution of TPP in the environment is as follows: soil (43.9%), sediment (41.0%), water (14.3%) and air (0.7%).25_{Research found that TPP enters the environment via} hydraulic fluid leakages, via leaching and volatilization from plastics and from manufacturing processes.22_{The solubility and log Kow of TPP, BDP and RDP is given in Table 1. As can be seen,} RDP has a low solubility which results in the high log Kow of the substance. The log Kow is the

logarithm of the ratio of the compound’s concentration in n-octanol to its concentration in water when in equilibrium. The solubility of TPP is five times higher than the solubility of BDP and this could explain why TPP is acutely toxic to aquatic organisms. Because of the low solubility of RDP, this flame retardants is superior to the other two. On the other hand, little can be said because an efficient method to analyse concentrations of BDP and RDP is missing.

Table 1. Solubility and Log Kow of TPP, BDP and RDP. 22

OPFR Solubility (mg/L) Log Kow

TPP 1.9 4.59

BDP 0.4151 4.5

RDP 1.11 x 10-4 _7.41

Figure 1. Structures of 1) triphenylphosphate (TPP), 2) resornicol bis(diphenyl phosphate) (RDP), 3) bisphenol A bis(diphenyl phosphate) (BDP)

(7)

3.5. Aim of this study

Now OPFRs are being used in multiple polymer product instead of BFRs, multiple research groups found traces of some OPFRs in surface waters26-29_{, wastewater}16,30_{and air}27,28_{. Because they} found an accumulation of some OPFRs in the environment, several analytical methods have been tested.22_{While other studies focused on a lot of OPFRs including TPP, Rodil et al. started to focus on} the extraction of RDP and BDP in water samples, because an efficient method was missing.16

Unfortunately, Rodils research group failed to detect BDP and RDP in water samples, while they did develop a plausible method for other FRs. They ascribed the non-detection to low concentrations in the water samples, because these OPFRs are relatively new. Helmus31_{developed an extraction method} based on the work of Rodil and this research will further build on the discoveries of Helmus. The purpose of this study is to develop an analytical method to detect low concentrations of TPP, RDP and BDP in influent and effluent of a sewage water treatment plant based on Helmus and Rodil. When this method is optimized in the first part of this study, samples will be taken from the sewage treatment plant for analysis with this method. The influent and effluent samples will be compared in

concentration of TPP, RDP and BDP. If there are any differences, this means that the sewage treatment is able to filter out any OPFRs.

The research question of this study will be: Are there any traces of TPP, BDP and RDP in wastewater? If there are traces of those OPFRs, do they exceed toxic quantities?

If there are traces of OPFRs, are any OPFRs cleansed out of the wastewater in the water treatment plant?

If there are any OPFRs in the water samples, it is expected that the concentration of TPP is higher than the concentration of RDP and BDP, because it has a higher solubility. RDP has the lowest solubility, so it is likely that the concentrations of RDP will be the lowest. A negative difference between influent and effluent is expected, because the OPFRs tend to adsorb and the purification process will remove the OPFRs.

4. Materials & Methods

4.1. Chemicals

In this study a mix solution (200 mg/L) of TPP (99%, Sigma Aldrich, 115-86-6), BDP (≥80%, ICL, 5945-33-5) and RDP (≥80%, ICL, 125997-21-9) was made and used to make dilutions for several purposes. The internal standard was deuterated TPP (dTPP, 98%, Sigma Aldrich) and several dilutions of the stock solution (500 mg/L) were made. The solvent in all stock solutions was

acetonitrile. Further chemicals that were used are methanol (100%, Biosolve[1]_{, ULC/MS grade),} acetone (100%, Biosolve, ULC/MS grade), acetic acid (100%, Merck, Darmstadt, Germany), formic acid (99%, Biosolve, ULC/MS grade), hexane (100%, Rathburn Chemicals Ltd, Walkerburn, Scotland) and ultrapure water (ULC/MS grade). For the analysis of concentration in LC-MS/MS 9 levels were used with level 0 to 8 with concentration 0, 1, 10, 25, 50, 100, 500, 1000, 2000 μg/L respectively with an internal standard concentration of 500 μg/L dTPP.

[1] = Biosolve, Valkenswaard, the Netherlands

4.2. Samples

All samples were collected in polypropylene bottles of 200 mL. They were washed with acetone (3 times) and methanol (ULC/MS grade, 3 times). Within 2 hours after sampling , formic acid

(8)

(2%) was added and the samples were stored at 4 °C. The water samples were 24 h composite samples of waster water samples from the RWZI in Amsterdam, The Netherlands.

4.3. Two methods: controls and internal standard.

4.3.1. Controls

Because the internal standard was dTPP, it wasn’t sure if this internal standard would

compensate for RDP and BDP, because they have slightly different properties. For this reason another compensating method was used. Controls were made of the water sample before executing SPE. One control consisted of 50% water sample and 50% methanol, and the other control consisted of 25% water sample and 75% methanol. The controls were also analysed with LC-MS/MS and the

concentrations of the eluates were compensated with the recoveries found in the controls. The control method was only used in the optimization process and not in the sewage water sample analysis. 4.3.2. Internal Standard

Even though the control method did give high recoveries, the control method wasn’t applicable to unknown volumes: it will always be necessary to know the concentrating factor to calculate concentrations with the control method. The internal standard method however calculates the concentrations in the elution and compares this with the concentration that it needed to be. Than the loss-factor is calculated and processed in calculating the concentration of the OPFR. TPP is very similar with dTPP, which means that dTPP is a very good IS for TPP. However because RDP and BDP have other adorption rates and log Kow, dTPP could overcompensate or undercompensate for these OPFRs.

4.4. Procedure

4.4.1. Overall Program

In this experiment concentrations of the three flame retardants (TPP, RDP and BDP) were measured in influent and effluent samples of water treatment plants. Those samples underwent solid phase extraction (SPE) to extract the OPFRs out of the water sample and the eluate of the SPE was analysed with LC-MS/MS. From the LC-MS/MS analysis the concentration was calculated.

Concentrations in influent and effluent samples of the sewage treatment plant were compared to see if there were any differences. Differences would indicate that the sewage treatment plant can remove the OPFRs present in wastewater. Prior to the analysis of water samples from the water treatment plant, several optimization experiments took place.

4.4.2. Experiment 1: Choosing the best SPE cartridge

The first step in this process was choosing the best SPE cartridge. Two cartridges were tested: Waters Oasis HLB-150 and Waters Oasis MCX-150. Both columns were acidified, because acidified columns gave the best results in Rodil et al.16_{and Helmus}31_{. The mechanism of the SPE is explained} further in section 4.5.1. Both cartridges were tested in threefold with a loading volume of 7 mL (100 μg/L) added with a volume pipette and an elution volume of 6 mL. The used volumes and

concentrations were based on previous research of Helmus.31_{To investigate if any analyte gets lost} during other steps in the SPE, the loading volume, the washing volume and the two elution volumes were collected and analysed with LC-MS/MS.

4.4.3. Experiment 2: Testing larger volumes

To test for breakthrough with bigger volumes, different volumes were tested in this experiment. The different volumes were 25 mL, 50 mL, 75 mL, 125 mL and 200 mL. Because the

(9)

final samples will be taken in polypropylene bottles, the solutions from 50 mL were made in polypropylene bottles and connected to the SPE with a connecting tube. The volumes from 50 mL were done in twofold. The smaller volume of 25 mL was done in fourfold: 2 erlenmeyers were connected to the SPE with a connecting tube and two SPE cartridge were loaded with a volume pipette. This was done to see if there were any differences between the two methods, because the OPFRs tend adsorb to the glassware and pipettes. To each volume a quantity of 1.25 µg was added. The concentrations of each volume are shown in Table 2.

Table 2. Loading Volumes with Concentration.

# Volume (mL) Concentration (μg/L) Tube/Pipett e Erlenmeyer/Polypropylene flasks 1 25 50 Pipette Erlenmeyer 2 25 50 Pipette Erlenmeyer 3 25 50 Tube Erlenmeyer 4 25 50 Tube Erlenmeyer

5 50 25 Tube Polypropylene flask

7 75 16.6 Tube Polypropylene flask

4.4.4. Experiment 3 and 4: Minimalizing adsorption with different hexane:acetone combinations Because OPFRs tend to adsorb and have a low solubility, a method had to be developed to minimize the adsorption. Eight polypropylene flask were filled with OPFR solution (100 mL , 50 μg) and spiked with dTPP (50µg). Samples were taken before and after shaking the solution in the flask for 30 minutes. After sampling, the flasks were emptied and different ratios of hexane and acetone were added with a total volume of 15 mL. The four ratios were 3:1, 0:1, 1:0 and 1:1 hexane:acetone. The extracting volume was transferred into tubes, in which the solution was evaporated under nitrogen flow. This was repeated 3 times. To analyse the extracted OPFRs in the tube, 10 mL of methanol and water (50/50) was added. The concentrations were calculated with the IS method.

The same experiment was repeated with a lower starting concentration (100 mL. 10 μg) in the polypropylene flask. Also the spike dTPP was lower (10 µg). The smaller amount of OPFR was chosen to end with a plausible concentration in the eluate volume, which wouldn’t exceed the calibration standard and to prevent having concentrations higher than the solubility values. Also the polypropylene flask weren’t emptied, but the OPFRs in the solution were collected in an HLB cartridge by solid phase extraction. The methanol (13 mL) which was used for the elution step was transferred into the polypropylene flask before the extracting steps to collect residual OPFR mix. The loading step and the elution step both went through a connecting tube before getting to the SPE. For the extracting steps, only hexane (15 mL) was used to extract, because hexane evaporated more easily than any combination with acetone or only hexane and the results of experiment 3 gave only small differences in extracted amount OPFR. The evaporation of the extraction solvent was a time consuming activity and had to be minimized. This whole procedure is shown in Figure 2. The concentrations were calculated with the IS method.

(10)

Figure 2. Graphic representation of experiment 4. Step 1: Loading the sample over the SPE tube. Step 2: Addition of methanol (15mL) to the polypropylene flask and 30 minutes shaking. Step 3: Eluting the SPE column with the methanol from the flasks. Step 4: Addition of 15 mL hexane to the empty polypropylene flasks and 30 minute shaking. Step 5: Transferring the hexane to a smaller vial (15 mL). Step 6: Evaporating the hexane (step 4, 5 and 6 are repeated 3 times). Step 7: Addition of methanol (15 mL) to the vial for analysis.

4.4.5. Experiment 5: Using water/methanol combinations for adsorption minimization

Because OPFRs tend to solve better in methanol than water, an extra adsorption experiment was executed in which the starting OPFR solutions were mixed with methanol before extracting with solid phase extraction. The ratios had to be chosen thoughtfully: a small amount of methanol won’t alter the results, but a great amount of methanol would prevent the OPFRs to adsorb to the HLB cartridge. Because all three OPFRs have different solubilities, it could be that one water/methanol ratio works for one OPFR but not for the other. The three chosen ratios were 50/50, 60/40 and 75/25 water to methanol. A total volume of 100 mL was made in polypropylene flasks with a OPFR concentration of 12.5 μg/L (100 mL, 1.25 µg). All samples were spiked with dTPP (1.25 µg). All flasks were shaken for 15 minutes. The flask were connected to the SPE with a tube. The concentration were analysed with the IS method.

4.4.6. Experiment 6 & 7: First and final trial with influent and effluent samples

Influent (n = 10) and effluent (n = 10) samples were taken at the RWZI (Amsterdam West, the Netherlands) on May 26th_{2015. Formic acid (4 mL) was added to all water samples. Also all samples} were spiked with internal standard dTPP (0.4 μg) and half of the samples were spiked with OPFR mix (0.2 μg). In this experiment 4 influent and 4 effluent samples were used. The reason half of the samples were spiked with OPFR mix is to test the recovery and breakthrough. To know the exact added amount of sample, the weight before and after SPE was measured. The methanol used for the elution step was added to the polypropylene flask as in the previous experiment. This was transferred to a smaller tube (15 mL) and centrifuged to prevent the solids in the influent and effluent from clogging the cartridge. The methanol was transferred to the SPE for the elution step with the same used connecting tube as mentioned before. The elution volume of 13 mL was evaporated to 2 mL to concentrate the samples.

The final trial with influent and effluent had some differences with the first trial. Again 4 influent and 4 effluent samples were used, but because there was some sludge in the influent and a little in the effluent, attempts were made to prevent the cartridge from clogging. For this reason a small volume with sludge was kept in the polypropylene flasks when the sample was loaded. To the volume of water in the polypropylene flask twice the volume of hexane was added. In the effluent bottles this was 10 mL and the influent bottles this was 15 mL. The bottles were shaken and put in an ultrasonic bath for 15 minutes. The hexane was collected after shaking and the extraction was repeated

(11)

twice. When all hexane was collected in a separate tube, it was evaporated and filled with methanol (2 mL) for analysis. The complete optimized procedure is shown in Figure 3.

Figure 3. Graphic representation of the second trial with influent and effluent. Step 1: Loading the sample over the SPE column, keeping a small volume in the polypropylene flask with sludge. Step 2: Addition of hexane to the water volume and 15 min shaking & 15 min sonicvibrations. Step 3: Transferring the volume to a smaller vial. Step 4: Transferring the hexane layer to another vial. Step 5: Evaporating the hexane (This part (2, 3, 4 and 5 are repeated 3 times). Step 6: Addition of the elution methanol (13 mL) to the original polypropylene flask. Step 7: Eluting the SPE column. Step 8: Transferring the eluate to the evaporated hexane vial. Step 9: Evaporating the methanol to 2 mL for analysis.

4.4.7. The matrix effect

The matrix effect is an enhancement or a suppression of the concentration in a sample. It is caused by other chemicals in the matrix. When analysing the self-made samples in experiment 1 to 5, the samples consist only of ultrapure water, OPFR and internal standard. When analysing the water samples from the sewage water treatment plant, the matrix effect could occur, because other chemicals could interfere with the signals. To tackle and to map the problem, a known amount of OPFR mix is added to the eluate of the non-spiked samples. Because this quantity is known, the theoretical difference in concentration is also known. The concentration of both the non-spiked sample and the afterwards spiked sample are measured with LC-MS/MS and their difference is calculated. The calculated difference is divided by the theoretical difference and this results in a percentage, which will be shown in the results.

4.5. Solid Phase Extraction settings

4.5.1 Used SPE Cartridges

In this experiment two columns of Waters were used: Oasis HLB (hydrophilic-lipophilic-balanced) and Oasis MCX (mixed-mode cation exchange). Their properties are shown in Table 2. Both columns were acidified to pH = 2 with formic acid, because acidified columns performed best in previous studies with OPFRs like BDP, RDP and TPP16,31_{. The acidified MCX gave great recoveries in} the experiments with Helmus31_{and other researchers at the VU found the acidified HLB to be the best} performing cartridge. This is contradictive with the expectations for two reasons. Even though the MCX and the HLB cartridge are similar, the MCX cartridge is also based on strong-cation-exchange

(12)

(SCX) (Table 3), which means that it is a great columns to extract charged products. On the contrary, the three OPFRs in this research will be neutral so they should interact less to an SCX column. Second both cartridges will be acidified with formic acid. While acidifying a column is also done for charged products, the analytes in this research will be neutral.

Table 3. Columns of the Solid Phase Extraction with their properties.

Manufacturer Column Sorbent (mg) Material Interactions

Waters Oasis HLB 150 HLB vdW, polar

Waters Oasis MCX 150 MCX vdW, polar, SCX

4.5.2 Steps of Solid Phase Extraction (SPE)

The column was first conditioned with methanol (5 mL). Then the equilibration of the column took place with H2O + 2% formic acid (5 mL). After those preparation steps, the sample was loaded. The volume of this step depended on the experiment. Afterwards the column was washed with the same formic acid solution as in the equilibration step (2 mL). The next step in the SPE process was drying the column. This took some 30 minutes. The eluates were collected with methanol (13 mL). Those elutuates were analysed by LC-MS/MS. The vacuum in all steps was between 10 and 15 mbar and the drop speed was always kept under 1 drop a second.

Table 4. Steps of the Solid Phase Extraction (SPE).

Step Solvent & Amount

1. Conditioning Methanol (5 mL) 2. Equilibration H2O + 2% FA (5 mL) 3. Loading Sample 4. Washing H2O + 2% FA (2 mL) 5. Drying -6. Eluting Methanol (12,5 mL)

4.6. LC-MS/MS settings

To analyse the eluates of the SPE, a C18 column was used. The gradient was used to separate the three different flame retardants and existed of bottles A and B: A = 50% water, 50% methanol, 0.2% acetic acid and ammonium acetate (0.39 g/L) and B: 100% methanol, 0.2% acetic acid and ammonium acetate (0.39 g/L). At t=0 the flow consisted of 25% A and 75% B and this changed to 100% B within t=0 and t=11. Until t=18 the flow was constant and finally at t=20 the flow was the same as the starting composition. In Table 3 the retention times are given for this method, which are based on the studies of Helmus and Van Duin32_.

Table 5. Retention times FRs for the LC-MS/MS.32

Compound Retention Time (min)

TPP 7.97

RDP 12.57

BDP 10.11

All compounds were ionized and were measured in positive mode. The SRM (Selected Reaction Monitoring) transition from the positively charged product to the product ion pair was measured in de MS/MS. The chosen transitions are shown in Table 4, which are based on previous studies.31,32_{The fragmentation is shown for in Figure 4 (TPP), Figure 5 (RDP), Figure 6 (BDP) and} Figure 7 (dTPP). The results gave different concentrations for the two transitions, so when the concentration of the OPFR was measured, the average was calculated.

(13)

The concentrations were calculated by Analyst Software on the basis of a calibration line, which was made of 9 levels (0 to 8, shown in the section 4.1. Chemicals). The LOQ was the second level (10 µg/L). The maximum was level 7 for RDP and BDP (1000 µg/L) and level 6 for BDP (500 µg/L).

Table 6. Transitions of the used compounds.

Compound Transition TPP transition 1 327.000 - 215.000 TPP transition 2 327.000 - 153.000 BDP transition 1 693.000 - 367.000 BDP transition 2 693.000 - 327.000 RDP transition 1 575.000 - 419.000 RDP transition 2 575.000 - 481.000 dTPP transition1 342.000 - 243.000 dTPP transition 2 342.000 - 164.000

Figure 4. Fragmentation of TPP, transition m/z 327 - m/z 215 and 327 m/z – 153 m/z.

(14)

Figure 6. Fragmentation of BDP, transition m/z 693 – m/z 367 and m/z 693 – m/z 327.

(15)

5. Results

5.1. Results of Experiment 1: choosing the best SPE cartridge

In Figure 8 the results of the cartridge experiment are shown. In this experiment two cartridges of Waters Oasis (MCX and HLB) were tested in threefold. The loading volume, the washing volume and the two elution volumes (2x 6 mL) were collected and tested. The first three columns show the recovery of the three OPFRs in the MCX cartridge, and the last three columns show the three OPFRs in the HLB cartridge. The recoveries were calculated with the controls, because the internal standard overcompensated for BDP and RDP. What can be seen in Figure 8 is that the HLB cartridge

performed better than the MCX cartridge. Second, the HLB cartridge needed a second elution to recover more OPFRs. This means that the elution in the optimal method will be at least twice the elution volume (13 mL). TPP MCX BDP MCX RDP MCX TPP HLB BDP HLB RDP HLB 0 10 20 30 40 50 60 70 80 90 100

Recovery of the load per OPFR per cartridge

Elution2 Elution1 Wash Load OPFR + cartridge R e co ve ry ( % )

(16)

5.2. Results of Experiment 2: different volumes

In Figure 9 the results of the second experiment are shown. Different volumes with different concentrations were made in a polypropylene flask (Table 2) and loaded over the SPE column with a connecting tube. The recoveries are calculated with controls as in the first experiment. It is remarkable that the recoveries in this experiment are lower than the recoveries in the previous experiment, because the method with the controls is similar. A possible reason is that in this experiment the sample has come in contact with a lot more materials which would’ve led to more adsorption. Every volume was tested in duplicated and the average was taken of the two for the results in the graph.

What can be seen in Figure 9 is that the loading volume of 125 mL seem to be the better volume compared to the other volumes (it has to be noticed that the scale of the y-axis runs from 0 to 40 instead of 100 so the differences are small). The mean of the three OPFRs is 20% in 200 mL and 22% in 125 mL. The concentrations in the water sample that will be analysed with this method will be lower than the concentrations made here. This means that the concentrations in the eluates will be lower too. When choosing a smaller volume, the eventual concentration of the eluate could be too low to detect. Second, when the samples are taken, the polypropylene flasks will be fully filled, which means that decreasing the volume includes transferring steps in which adsorption can take place. For this reason a larger volume of 200 mL is chosen. The recoveries for the volume of 25 mL and the different methods is shown in the appendix (Figure 16).

50 mL 75 mL 125 mL 200 mL 0 5 10 15 20 25 30 35 40

Recovery OPFRs in different loading volumes and methods

TPP BDP RDP Loading Volume R e co ve ry ( % )

(17)

5.3. Results of Experiment 3: adsorption minimization

In the next experiment different acetone hexane ratios were tested to minimize adsorption. A solution of OPFRs of 500 μg/L (100 mL) was made in 8 polypropylene flasks and samples were taken before and after shaking the bottles. The results of the shaking experiment are shown in Figure 17 in the appendix. It appeared that the concentration of BDP was higher after shaking and that TPP decreased enormously, while RDP stayed the same. The results of BDP could be due to homogenization and TPP seemed to be adsorbed to the flask.

After shaking the flasks were emptied and a volume of 15 mL was added of the desired ratio of acetone : hexane (1:3, 1:0, 0:1 and 1:1). The extracting volume was transferred into a smaller tube and evaporated. This was repeated three times. Eventually, methanol was added to the empty smaller tube and the sample was analysed. The different ratios were executed in duplicate and the mean of the duplicates and transition concentrations is shown in Figure 10. The results of RDP and BDP were similar in all four combinations, but the recovery of TPP was a little higher in 1 : 3 acetone : hexane and 1 : 0 acetone : hexane. In this situation again, it should be noticed that the y-axis runs from 0 to 40 and the differences in recovery aren’t much. When executing the evaporation of the four mixtures, it has come to the attention that the mixtures with only or partial hexane evaporated much easier than the ones with acetone. Because all combinations have comparable results and because hexane is much easier evaporated, hexane was the best extraction solvent and this was used in the second absorption extraction experiment. 1 3 1 0 0 1 1 1 0 5 10 15 20 25 30 35 40

Extraction: percentages of original concentration

TPP BDP RDP

Ratio Acetone : Hexane

R e co ve ry ( % )

(18)

5.4. Results of Experiment 4: adsorption minimization

Experiment 4 continues the adsorption minimization experiments. The process of the experiment is shown in Figure 4. In this experiment a lower concentration (100 μg/L) and the same volume (100 mL) were loaded over the SPE. This concentration is 5 times lower than the previous experiment (section 4.4.4.). The extraction was performed three times with only hexane (15 mL). The experiment was executed in fourfold and the mean of each OPFR is shown in Figure 11. In the appendix all four experiments are shown in Figure18.

What can be seen from the results is that the recovery of TPP, BDP and RDP lies between 40% and 50% of what is expected. The extra extraction steps afterwards do show an extra recovery between 5% (for BDP) and 17% (for TPP). For TPP it increases from 40% to almost 60%, which means an enlargement of almost 50%. For BDP it gives an enlargement of almost 20% and RDP increases with 40%. These findings confirm that using hexane as an extraction solvent can enlarge the recovery percentages, especially for TPP and RDP.

TPP BDP RDP 0 10 20 30 40 50 60 70 80 90 100

Percentage of the original concentratio in SPE elution and hexane extraction: averages

Hexane Extraction SPE Elution OPFR R e co ve ry ( % )

Figure 11. Percentage of original concentration in SPE elution and hexane extraction: averages.

5.5. Results of Experiment 5: using different water:acetone ratios

In the appendix in Figure 19 the results of the different water/acetone ratios are shown. Those results are shown in the appendix, because the method wasn’t applied to the optimal method, because it gave bad results for TPP. In this experiment the water samples were mixed with methanol in different ratios to prevent adsorption from happening. The ratios water to methanol were 50/50, 60/40 and 75/25.

TPP gave low recoveries in every experiment: all recoveries were below 30%. An explanation could be that TPP is easily dissolved when water is mixed with methanol. On the contrary, RDP gave better results and the recoveries of BDP were remarkably high. The difference between TPP and the other two flame retardants can be ascribed to their differences in solubility and log Kow. Because the ratio 75/25 gave impossible results for RDP (higher than 100%), Figure 20 shows only the ratios 60/40 and 50/50. Those results could be the result of dTPP flushing through the column and a minimal adsorption of RDP. This could mean that the eventual concentration of dTPP in the eluate was lower than the concentration of RDP, which would lead to an overcompensation. From Figure 19 and 20, it

(19)

can be said that for analysis of all three flame retardants mixing the water samples with methanol gives undesired results. This is because the TPP (and dTPP) recoveries are extremely low.

5.6. Results of Experiment 6: first batch

When the method was optimized, real influent and effluent samples were taken from the RWZI (Amsterdam, May 26th_{2015) and analysed with the optimized method. Some extra issues came} to light: the extra sludge in the samples, especially in the influent samples, was a risk for clogging the SPE columns. The decision was made to try to load the largest volume possible without getting the sludge in the tube. This resulted in one clogged SPE column. The spiked samples are shown in Figure 12, given in percentages of the expected concentration for which the samples were spiked.

The non-spiked samples and blanks are shown in Figure 13. These samples were only treated with formic acid (2%) and dTPP (0.4 μg). The errors of the columns in Figure 13 show divergent results. The remarkable peaks in the blanks show that either something has gone wrong or that the polypropylene bottles contain traces of flame retardant. This will be discussed further in the discussion. TPP BDP RDP -20 -10 0 10 20 30 40 50 60 70 80 90 100 110

Percentage of expected concentration in spiked influent and effluent samples

SPE Infl 4 SPE Infl 5 SPE Effl 3 SPE Effl 6 OPFR P e rc e n ta ge o f ex p e ct e d c o n ce n tr ati o n

(20)

TPP BDP RDP 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Concentrations of original non-spiked blanks, influent and effluent samples

Blank 1 Blank 2 Influent 7 Effluent 9 Effluent 10 OPFR C o n ce n tr ati o n ( μ g /L )

Figure 13. Concentration OPFRs in original non-spiked blanks, influent and effluent samples in μg/L.

5.7. Results of Experiment 7: second batch

In the second batch, the method was adjusted to the water samples with the sludge to prevent clogging of the tube and SPE cartridge. A small volume of water sample was kept in the

polypropylene flask and a liquid/liquid extraction was performed with hexane. The complete method is shown in Figure 3. In Figure 14 the concentrations of the blanks and the non-spiked influent and effluent samples are shown. The blank samples again show remarkable results: they show peaks for every OPFR. It can also be seen that TPP is present in influent and effluent, while the concentrations of RDP are lower and BDP are almost zero. In Figure 15 the spiked (0.2 μg) influent and effluent samples are shown. The added OPFR solution for the spike should have resulted in a concentration at about 1 µg/L, but they seem to be doubled. It could be that a mistake was made was preparing the spike solution: diluted 5 times instead of 10 times. The results will be further discussed in the discussion.

(21)

TPP BDP RDP -0.15 -0.05 0.05 0.15 0.25 0.35 0.45 0.55 0.65

Concentrations of original non-spiked blanks, influent and effluent samples

Blank 1 Blank 2 Influent 2 Influent 10 Effluent 5 Effluent 2 C o n ce n tr ati o n ( μ g /L )

Figure 14. Concentration of non-spiked blanks, influent and effluent samples in μg/L.

TPP BDP RDP 0 0.5 1 1.5 2 2.5 3

Concentration of OPFRs in spiked samples

Influent 6 Influent 9 Effluent 8 Effluent 1 C o n ce n tr ati o n ( μ g /L )

Figure 15. Concentration of OPFRS in spiked influent and effluent samples in μg/L.

Table 7. The Matrix effect in percentages for TPP,BDP and RDP in three water samples and their average.

Sample TPP BDP RDP

Influent 2 _87.7 _52.3 _139.2

Influent 10 _118.1 _83.2 _172.0

Influent average _102.9 _67.8 _155.6

Effluent2 _90.8 _88.2 _191.4

Table 7 shows the matrix effects, calculated with the concentration of non-spiked samples and the concentrations of these samples when spiked afterwards. This experiment shows how the matrix influences (suppresses or enhances) the concentrations of the three OPFRs.

(22)

6. Discussion

In this experiment, the extraction of three different flame retardants has been optimized. This was done by different sub-experiments. When optimized, this method was used on water samples from the RWZI on 26th_{of May, 2015. Now all results of the optimization will be discussed in a separate} header and concluded with an optimized method to measure OPFRs in water and with its future improvements. After that the results of the first and second batch will be discussed and compared with the values in the literature.

6.1. Method optimization

6.1.1. Discussion of the results of the optimization process

The first part of the method optimization included testing two different cartridges of Waters Oasis: MCX-150 and HLB-150. As can be seen in Figure 8, the HLB column gave better results. This was expected because the mechanism of the MCX cartridge is based on strong cationic exchange and the analysed OPFRs have no charge at all. The concentrations of the analysis were low in both cartridges, but when compensated with the controls, all HLB columns gave recoveries of some 80% for all three flame retardants. What is also remarkable is that the second elution had some 10% extra analyte of the original quantity. This means that in further experiments HLB columns will be used with an elution volume of at least 12 mL.

In the second part of the optimization different volumes were tested. A larger volume is desired, because concentrations in water samples will be extremely low. When comparing the results of the larger volumes in Figure 9, it can be seen that a loading volume of 125 mL gave the best results and should be used in further experiments. It has to be noticed that the scale on the y-axis runs from zero to 40 instead of 100. This means that the difference between 125 and 200 mL is only small: the mean of the three OPFRs is 20% in 200 mL and 22% in 125 mL. It has to be noticed that the

recoveries aren’t as high as the previous experiment and this could be an effect of a higher adsorption, because the solution had been in contact with more materials than the first experiment (like the tube and the polypropylene bottle). Because the expected concentrations in the final water samples will be extremely low, a volume of 200 mL is preferable and will be used in further experiments.

In this experiment a smaller volume of 25 mL in an Erlenmeyer was tested in two different methods: the pipette method and the tube method. The tube method is preferable, because this method isn’t as labour-intensive as the pipette method. From the results in Figure 16, it can be concluded, when working with a smaller volume like 25 mL, it is better to use the pipette method. This can easily be explained when the 25 mL is compared to the volume of the Erlenmeyer flask. The diameter of these flasks was about 8 cm, so 25 mL would only reach a height of some 1,5 cm. When the tube method is used, the last load in the flask was hard to catch. When using a smaller volume like 25 mL, the last load in the flask was a larger part of the whole volume than in larger volumes like 50, 75, 125 and 200. This is why, when analysing the smaller volumes like 25 mL, the pipette method is a better technique. Another difference between the 25 mL volume and the higher volumes is that adsorption rates for the OPFR on the glassware and on the polypropylene flasks could be different. Because a higher volume is necessary for detectable results, the results of the smaller volume are no longer of interest.

In the adsorption experiments several results were obtained. The results of the pre and post shaking experiment are shown in Figure 17 in the appendix. From these results it is confirmed that all the samples should be in contact with as few materials as possible, because the adsorption rates are high, especially for TPP and RDP. Further this experiment focused on choosing the best extracting solvent. Four different ratios of acetone and hexane were used and the results are shown in Figure 10. The results of RDP and BDP were similar in all four combinations, but the recovery of TPP was a

(23)

little higher in 1 : 3 acetone to hexane and 1 : 0 acetone to hexane. In this situation it should again be noticed that the y-axis runs from 0 to 40 and the differences in percentage recovery aren’t much. When executing the evaporation of the four mixtures, it has come to the attention that the mixtures with only or partial hexane evaporated much easier than the ones with acetone. Because all combinations have comparable results and because hexane is much easier evaporated, only hexane was the best extraction solvent and this was used in the second absorption extraction experiment. The complete procedure of this experiment is shown in Figure 2. Because the experiment was executed in fourfold, Figure 11 shows the mean of the values in Figure 18 (in the appendix). What can be seen from the results is that the recovery of TPP, BDP and RDP lies between 40% and 50% of what is expected. The extraction steps afterwards do show an extra recovery between 5% and 15%. This confirms that using hexane as an extraction solvent can enlarge the recovery percentages, especially for TPP and RDP.

The experiment in which the water samples were combined with methanol gave mixed results. TPP gave low recoveries in every ratio: all recoveries were below 30%. An explanation could be that TPP is easily dissolved when water is mixed with methanol and for this reason flushes through the SPE cartridge in the loading step. On the contrary, RDP gave better results and the recoveries of BDP were remarkably high. The difference between TPP and the other two flame retardants can be ascribed to their differences in solubility and log Kow. The ratio 75/25 gave impossible results for RDP (higher than 100%). Those results could be the result of dTPP flushing through the column and a minimal adsorption of RDP. This could mean that the eventual concentration of dTPP in the eluate was lower than the concentration of RDP, which would lead to an overcompensation. From Figure 19 and 20, it can be said that for analysis of all three flame retardants mixing the water samples with methanol gives undesired results. This is because the TPP recoveries are extremely low. The water methanol mixtures will not be of interest for the optimized method.

6.1.2. Choosing the optimal method

When revising the results of the optimization steps, the optimal method can be assembled. First, an HLB column will be used with an volume of 200 mL. The high volume was chosen because a high volume contains more possible flame retardant than a smaller volume, even though smaller volumes like 125 mL gave 5% to 10% more recovery. When a volume of 125 mL is used and the final concentration of the elution is too low to measure, no peaks will be detected.

When choosing the best extracting solvent, the time-consuming activity of evaporating the solvent had a major role in making the decision between the solvents or mixture. Mixtures with acetone took almost twice as long as the extractions with only hexane. Because the differences between the combinations were minimal, hexane was chosen as optimal extraction solvent. The extraction step gave an extra recovery between 5% and 15%, which made it a crucial step.

The final decision had to be made if methanol should be added to the water samples and if so, in what ratio. The reason to try this was because it would prevent adsorption before the sample had undergone solid phase extraction. The decision was made to not add any methanol to the samples: first because TPP would flush through the SPE column and detection would be impossible and second because a maximal volume is preferred. Adding methanol to the flask would mean a smaller volume of water sample and because a smaller water volume would lead to low concentrations in the eluates. 6.1.3. Possible improvements for future analysis

In the previous part a sufficient method has been developed, however there are still

components that could be improved. The recoveries of a great deal of the experiments was low, i.e. under 50%.This could be due to taking dTPP as internal standard, which works good for TPP, but not for BDP and RDP, because they have different properties. The research of Rodil et al. used

(24)

be better to develop a method which is only for RDP and BDP and method which is only for TPP or similar flame retardants like TPP. The experiment in which methanol was added to the water sample for example gave better results for RPP and BDP than before so this could be a good starting point for a analysing method for only RDP and BDP.

For future experiments it should be found out whether the polypropylene bottles contain any kind of flame retardant. This should be very likely, but it will interfere with the results if they contain traces of TPP, RDP and BDP. If they do contain one of the three OPFRs, they should be excluded from the experiment and other flasks should be used.

Another part that needs improvement is the cleaning method. The tubes were flushed 3 times with hexane (20 mL) and three times with methanol (20 mL) before they were used in other

experiments. If the polypropylene bottles don’t contain any OPFR, the contamination could have been from a poor cleaning method or the tubes contained the OPFR. This needs to be found out in further research, because it could interfere with the results. In the case that the cleaning method isn’t

sufficient, another method needs to be found. When the tubes appear to contain the analysed OPFRs, other transferring methods should be used: either the pipette method or another tube, which doesn’t contain OPFRs.

6.2. Sample analysis of RWZI, discussion of the results

6.2.1. Batch 1

When the optimized method from paragraph 6.1.2. was first applied to real samples, some extra issues came to light. The extra sludge in the samples, especially in influent samples, was a risk for clogging the SPE columns. This resulted in a conflict between choosing for the largest volume possible and preventing the column from clogging. The decision was made to try to load the largest volume possible without getting the sludge in the tube. This resulted in one clogged SPE columns, which was no longer usable. In Figure 12 the concentrations of OPFRs in the LC vials are shown, while Figure 13 shows the concentration in the original sample. All concentrations are corrected for the added amount of loading volume. A possible reason for the absence of peaks (Figure 12 & 13) in the influent sample could indicate a breakthrough of the SPE column, because the water samples contain a lot of other chemicals. Another reason for the absence of peaks is adsorption of OPFRs to the sludge in the samples. The peaks in the blank samples pose another problem. This should be impossible, since the blanks were only spiked with dTPP and not with OPFR mix. A possible explanation of the peaks is that the polypropylene bottles contain OPFRs. This experiment was executed to investigate whether the experiment with water samples from the sewage water treatment plant could undergo the same optimized method as proposed in the previous section. It can be concluded that some adjustments had been made to make the method more suitable to real samples. The most important factor in the adjustments is a liquid/liquid adsorption for the water with sludge. This will prevent the SPE column and connecting tube from clogging. The adjustments will be applied in the second batch and the results of batch 2 will be used to answer the research questions.

6.2.2. Batch 2

When the analysis of the second batch took place, some parts of the method were different from the first batch. The sludge underwent a triple liquid/liquid extraction with hexane, and more precautions were taken to prevent clogging of the SPE tubes. This liquid/liquid extraction was necessary, because the last 5 mL of the water sample were kept in the polypropylene bottle. This means that the added hexane could be transferred and evaporated, because there was still water in the extraction. What is remarkable is that the blanks again show signs of the OPFRs. This confirms the hypothesis that the polypropylene bottles contain traces of OPFR and that the contamination is not a

(25)

result of contamination of poorly cleaned materials. What is also remarkable is that the internal standard corrects the concentrations with a factor 2. All spiked concentrations should be around 1 μg/L, but they seem to be doubled or higher. It could be that a mistake was made when preparing the spiking solution. It could have been diluted 5 times instead of 10. This would explain the doubled concentrations.

If this hypothesis that the spikes have been doubled is accepted, Figure 14 shows the

concentrations of the three OPFRs. There is a clear difference between influent and effluent samples for TPP. The concentration in influent is about 0.50 µg/L and the concentration decreases to 0.20/0.30 µg/L. This decrease could indicate that TPP could be removed in a sewage water treatment plant. In the literature Rodil et al.tried to document the concentration of TPP in water samples from treatment plant with solid phase extraction, but their concentrations in the analysis were all below LOQ.15,16 Quintana et al. found a concentration of 0.063 (±0.008) µg/L TPP in influent and a concentration 0.091 (±0.005) µg/L.33_{This is a remarkable result, because the hypothesis is a decrease in}

concentration and not an increase. While the concentration of TPP in effluent in this experiment is 30% lower, the concentration of TPP in influent does come close to the previously mentioned

concentration of Quintana et al.. However, if the amount TPP in the blanks is a result of traces of TPP in the polypropylene flasks, this amount has to be subtracted from the concentration in influent and effluent. This would lead to a lower concentration further away from the concentration Quintana found. Differences in concentration could also be ascribed to differences in the water samples. Quintana et al. took their samples in Sartorius, Germany while the samples in this experiment were taken at Amsterdam, the Netherlands. Rodríquez et al., who analysed water samples in Spain, also found a decrease in TPP concentration between influent and effluent samples.34_{However there was not} enough data to draw conclusions and the concentration were below 1 ng/L, which is a factor 500 lower than the concentrations found in this study.

With these results for TPP, plausible concentrations have been found which do correspond with the literature. For further research a lot more samples have to be taken and analysed to get more reliable results. It also has to be found out if the polypropylene flask contain OPFRs or if the peaks are a result of poor cleaning of the materials.

BDP and RDP seem to be absent in the water samples. The blanks have a higher

concentrations for these two OPFRs than the influent and effluent samples. This is very contradictive, because the blanks could have had peaks for the OPFRs for 2 reasons. The first was a poor cleaning method of the tubes, but this would have resulted in the same peaks for both the blanks and the influent and effluent. The second possible reason was traces of OPFR in the polypropylene bottles, but all samples were collected in these bottles. The only difference between the blanks and the samples was the little amount of sludge in the samples. Suppose the polypropylene bottles do contain traces of TPP, RDP and BDP, than the OPFRs in the sample bottles could have adsorbed to the sludge, while the OPFRs from the bottle in the blanks could not. This would explain a higher peak for the blanks and almost no peak or not a peak at all in the influent and effluent samples. The small difference between BDP and RDP in Figure 14 is also shown in Figure 15, the spiked influent and effluent samples. In that graph the BDP seemed to have adsorbed more than RDP. The literature does not give any

information about concentrations of RDP and BDP. This is either because these OPFRs aren’t used as much as other OPFRs and concentrations are expected to be low and because researchers haven’t succeeded to develop a good analysing method for these two OPFRs. They either had levels below LOQ or no peaks at all. The low concentrations could either be due to truly low concentrations in the environment, but also because they tend to adsorb to any solid material.

The found concentrations of RDP contradict the solubility value of RDP in Table 1. This value is 1.11 x 10-4_{mg/L, which is similar to a concentration of 0.11 µg/L. This could mean that a higher} concentration of 0.11 µg/L isn’t measurable. The highest peak for non-spiked samples is about 0.09

(26)

µg/L (blank 2). The higher concentrations in Figure 15 were higher than this solubility, because the OPFRs were added to the sample. The added OPFRs couldn’t have been solved in the sample, so they most probably adsorbed to the polypropylene flask and were recovered by the extraction step. This low solubility is either a problem, because the exact amount of RDP in the environment is hard to measure. On the other hand, low solubility makes it harder for RDP to leach from polymers in wastewater.

The matrix effect was calculated by subtracting the concentration of the non-spiked sample of the concentration of the sample which was spiked afterwards. This was divided by 200, which should be the extra concentration. The values in Table 7 are shown for each OPFR in each sample, but also the averages of the influent are taken in the third column. The values confirm no matrix effect for TPP, an enhancement for RDP and a suppression for BDP. In the study of Rodil16_{the matrix effect of TPP} was calculated and they found a matrix effect of 160%. This value is a lot higher than the value from this study. The experiment of Rodil had a lot of similarities with this experiment, only in this

experiment the SPE column was equilibrated with a formic acid solution and the water sample was acidified as well. Although RDP and BDP had an enhancement and suppression respectively, this experiment did gave a result for those two OPFRs. The study of Rodil failed to detect any RDP or BDP with their method.

When finalizing this project, it was found that the final experiment couldn’t give the expected answers. The optimized method had to undergo changes to be suitable for real samples of the water treatment plant with some sludge. It can be seen that TPP is found in influent and effluent samples, but the concentrations of RDP and BDP are low or not even present. It’ll be interesting to optimize a method to analyse sludge samples of the treatment plant, because the OPFRs could adsorb to the sludge. It can be concluded that it is hard to optimize a method for three OPFRs, which are all different from each other: especially TPP from BDP and RDP. It might be better to develop a method only for RDP and BDP, a method for analysing TPP had already been made.

7. Conclusion

In conclusion, traces of TPP have been found in influent and effluent samples. Influent gave concentrations of 0.50 µg/L and the concentration in effluent decreased to 0.20/0.30 µg/L. This decrease could indicate the removal of TPP, but the experiment has to be repeated to be reliable. The presence of RDP and BDP was low or not present at all, even though the method worked for all three OPFRs in the optimization process. A possible reason for the low concentrations is the lack of a good liquid/liquid extraction method, because RDP and BDP are expected to adsorb to the sludge in the samples. This should be adjusted in future research and sludge samples should be analysed as well.

8. Acknowledgements

I would like to thank several important people who helped with this experiment. First, I would like to acknowledge my supervisor dr. John Parsons for allowing me to continue this research and allowing to do my bachelor thesis at IBED at the University of Amsterdam. Second, I would like to thank my daily supervisor Rick Helmus for helping me set up the experiment and guide me through the progress of the experiment. Third, I would like to thank the sewage water treatment plant RWZI at Amsterdam-West for providing me of the 24 h composite water samples of influent and effluent and the tour they gave me through their new office and laboratories.

(27)

9. References

1. Hindersinn, R. R. Historical aspects of polymer fire retardance. Fire and Polymers: Hazards

Identification and Prevention 1989, 87-96.

2. Klitzke, R. A. Roman building ordinances relating to fire protection. The American Journal of

Legal History 1959, 173-187.

3. Alaee, M.; Arias, P.; Sjödin, A.; Bergman, Å An overview of commercially used brominated flame retardants, their applications, their use patterns in different countries/regions and possible modes of release. Environ. Int. 2003, 29, 683-689.

4. Emmons, H. Heat transfer in fire. Journal of Heat Transfer 1973, 95, 145-151. 5. Troitzsch, J. H. Overview of flame retardants. Chemistry today 1998, 16.

6. Birnbaum, L. S.; Staskal, D. F. Brominated flame retardants: cause for concern? Environ. Health

Perspect. 2004, 112, 9-17.

7. de Wit, C. A. An overview of brominated flame retardants in the environment. Chemosphere 2002,

46, 583-624.

8. Kemmlein, S.; Herzke, D.; Law, R. J. Brominated flame retardants in the European chemicals policy of REACH—regulation and determination in materials. Journal of Chromatography A

2009, 1216, 320-333.

9. Darnerud, P. O. Toxic effects of brominated flame retardants in man and in wildlife. Environ. Int.

2003, 29, 841-853.

10. Hamers, T.; Kamstra, J. H.; Sonneveld, E.; Murk, A. J.; Kester, M. H.; Andersson, P. L.; Legler, J.; Brouwer, A. In vitro profiling of the endocrine-disrupting potency of brominated flame retardants. Toxicol. Sci. 2006, 92, 157-173.

11. Legler, J.; Brouwer, A. Are brominated flame retardants endocrine disruptors? Environ. Int. 2003,

29, 879-885.

12. Stapleton, H. M.; Klosterhaus, S.; Eagle, S.; Fuh, J.; Meeker, J. D.; Blum, A.; Webster, T. F. Detection of organophosphate flame retardants in furniture foam and US house dust. Environ.

Sci. Technol. 2009, 43, 7490-7495.

13. Van den Eede, N.; Dirtu, A. C.; Neels, H.; Covaci, A. Analytical developments and preliminary assessment of human exposure to organophosphate flame retardants from indoor dust. Environ.

Int. 2011, 37, 454-461.

14. Meyer, J.; Bester, K. Organophosphate flame retardants and plasticisers in wastewater treatment plants. Journal of Environmental Monitoring 2004, 6, 599-605.

15. Rodil, R.; Quintana, J. B.; López-Mahía, P.; Muniategui-Lorenzo, S.; Prada-Rodríguez, D. Multi-residue analytical method for the determination of emerging pollutants in water by solid-phase extraction and liquid chromatography–tandem mass spectrometry. Journal of Chromatography A

(28)

16. Rodil, R.; Quintana, J. B.; Reemtsma, T. Liquid chromatography-tandem mass spectrometry determination of nonionic organophosphorus flame retardants and plasticizers in wastewater samples. Anal. Chem. 2005, 77, 3083-3089.

17. Öberg, K.; Warman, K.; Öberg, T. Distribution and levels of brominated flame retardants in sewage sludge. Chemosphere 2002, 48, 805-809.

18. Adgent, M. A.; Hoffman, K.; Goldman, B. D.; Sjödin, A.; Daniels, J. L. Brominated Flame Retardants in Breast Milk and Behavioural and Cognitive Development at 36 Months. Paediatr.

Perinat. Epidemiol. 2014, 28, 48-57.

19. Kim, J.; Isobe, T.; Muto, M.; Tue, N. M.; Katsura, K.; Malarvannan, G.; Sudaryanto, A.; Chang, K.; Prudente, M.; Viet, P. H. Organophosphorus flame retardants (PFRs) in human breast milk from several Asian countries. Chemosphere 2014, 116, 91-97.

20. Ali, N.; Dirtu, A. C.; Van den Eede, N.; Goosey, E.; Harrad, S.; Neels, H.; Coakley, J.; Douwes, J.; Covaci, A. Occurrence of alternative flame retardants in indoor dust from New Zealand: indoor sources and human exposure assessment. Chemosphere 2012, 88, 1276-1282.

21. Beard, A.; Klimes, M.; Wietschorke, U. In In Non-halogenated phosphorus, inorganic and

nitrogen flame re-tardants for electronics: update on market situation, drivers and trends;

Electronics Goes Green 2012 (EGG), 2012; IEEE: 2012; , pp 1-5.

22. van der Veen, I.; de Boer, J. Phosphorus flame retardants: properties, production, environmental occurrence, toxicity and analysis. Chemosphere 2012, 88, 1119-1153.

23. Pawlowski, K. H.; Schartel, B. Flame retardancy mechanisms of triphenyl phosphate, resorcinol bis (diphenyl phosphate) and bisphenol A bis (diphenyl phosphate) in

polycarbonate/acrylonitrile–butadiene–styrene blends. Polym. Int. 2007, 56, 1404-1414. 24. Pawlowski, K. H.; Schartel, B. Flame retardancy mechanisms of triphenyl phosphate, resorcinol

bis (diphenyl phosphate) and bisphenol A bis (diphenyl phosphate) in

polycarbonate/acrylonitrile–butadiene–styrene blends. Polym. Int. 2007, 56, 1404-1414.

25. SIDS, U. O. Triphenyl phosphate (CAS No. 115-86-6). United Nations Environment Programme

Publications (ed): United Nations Environment Programme, Organisation for Economic Co-operation and Development (OECD) Screening Information DataSets (SIDS) 2002.

26. Kim, J.; Isobe, T.; Tanoue, R.; Chang, K.; Tanabe, S. Comprehensive Determination of

Pharmaceuticals, Personal Care Products, Benzotriazole UV Stabilizers and Organophosphorus Flame Retardants in Environmental Water Samples Using SPE Coupled with UHPLC-MS/MS.

Current Analytical Chemistry 2015, 11, 138-149.

27. Quintana, J. B.; Rodil, R.; Reemtsma, T.; García-López, M.; Rodríguez, I. Organophosphorus flame retardants and plasticizers in water and air II. Analytical methodology. TrAC Trends in

Analytical Chemistry 2008, 27, 904-915.

28. Reemtsma, T.; Quintana, J. B.; Rodil, R.; Garcı, M.; Rodrı, I. Organophosphorus flame retardants and plasticizers in water and air I. Occurrence and fate. TrAC Trends in Analytical Chemistry

2008, 27, 727-737.

29. Andresen, J.; Grundmann, A.; Bester, K. Organophosphorus flame retardants and plasticisers in surface waters. Sci. Total Environ. 2004, 332, 155-166.

(29)

30. Marklund, A.; Andersson, B.; Haglund, P. Organophosphorus flame retardants and plasticizers in Swedish sewage treatment plants. Environ. Sci. Technol. 2005, 39, 7423-7429.

31. Helmus, R. Analysis and Biodegradation of Organophosphorus Flame Retardants in Aquatic Environments, University of Amsterdam, Institute for Biodiversity and Ecosystem Dynamics, Amsterdam, 2012.

32. Duin, L. v. Analysis og organophosporus flame retardants in sewage treatment plant influent and effluent; an analytical method for detection of TPP, RDP and BDP in aqueous samples, Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam, Amsterdam, 2014.

33. Quintana, J. B.; Reemtsma, T. Potential of membrane-assisted solvent extraction for the determination of phosphoric acid triesters in wastewater samples by liquid chromatography– tandem mass spectrometry. Journal of Chromatography A 2006, 1124, 22-28.

34. Rodríguez, I.; Calvo, F.; Quintana, J.; Rubí, E.; Rodil, R.; Cela, R. Suitability of solid-phase microextraction for the determination of organophosphate flame retardants and plasticizers in water samples. Journal of Chromatography A 2006, 1108, 158-165.

10. Appendix

25 mL Pipette 25 mL Tube 50 mL 75 mL 125 mL 200 mL 0 5 10 15 20 25 30 35 40

Recovery OPFRs in different loading volumes and methods

TPP BDP RDP

Volume and method

R e co ve ry ( % )

Analysis of organophosphorus flame retardants in sewage water treatment plant influent and effluent samples

Institute for Biodiversity and Ecosystem Dynamics (IBED)