Experimental approach

In order to test whether the δ18_O

p determined from an individual protocol is accurate, it is important that the source δ18_O

p used within an experiment is known. Therefore, 10 litres of the WWTP effluent and of the two river waters were stripped of Pi prior to analysis, to prevent Pi that may have been originally present within a matrix from influencing the final δ18Op measurement. Stripping of Pi was achieved through a batch mode shake using a zirconium oxide (ZrO)/acrylamide binding gel that has been used to remove P from aqueous solutions within a diffusive gradient in thin films (DGT) technique (Zhang et al., 1998; Ding et al., 2010). The matrix was shaken for 24 hrs following the addition of each sheet of ZrO binding gel, until the Pi concentration in the matrix was below 0.009 mg P.L-1 (equating to <2% of the Pi in the sample after the addition of the KH2PO4 spike). This method was chosen to minimise any potential alterations of the initial matrix; for example, chemical stripping of P using the MagIC precipitation would have stripped the matrix of Pi, however this process would have also introduced both Mg2+ and Cl- ions into the matrix. The presence of these additional ions during the spiking experiment could have significant, and differing, effects on the performance of the two methods.

A KH2PO4 spike was subsequently added to 3 x 1 L of the four matrices to attain triplicate 0.4 mg P.L-1 _{solutions for each method. KH}

2PO4 was chosen to create the Pi spike as it is readily combustible, which allows direct analysis of δ18_{O within the}

TCEA-IRMS without the need for additional chemical reactions which could alter the source δ18_{O value. The spectator cations – K}+ _{and H}+ _{– should also have minimal} interference on the protocols for Ag3PO4 precipitation.

Aliquots of the spiked matrices were taken throughout the initial stages of Method 3 (steps 3c-6c in Figure 3.12) to determine Pi losses and pH. The concentrations of SRP and TP were measured using the techniques described in Section 3.2.2, and DOC concentrations as in Section 3.4.2. The following approaches were then used:

Method 1: 0.1 mol MgOH2.6H2O was added to 3 x 1 L of each Pi-spiked MilliQ water matrix and sample matrix, followed by 25 mL 1M NaOH solution, and mixed occasionally to precipitate brucite over 10 mins. The precipitate was collected through centrifugation at 3500 RCF for 10 minutes and dissolved in a 1:1 mixture of 10M HNO3 and c.CH3COOH. The addition of a CH3CO2K buffer and CeNO3 formed a CePO4 precipitate, which was dissolved in 2 mL 1M HNO3. The solution was diluted to 0.2M HNO3 with MilliQ water before the removal of Ce+ ions using an overnight shake with Dowex 50x8 cation exchange resin in batch mode. After separation of the liquid from the resin, Ag3PO4 was precipitated using the fast precipitation method by addition of 0.5 g AgNO3, c.NH4OH, 3M HNO3 and 3M NH4NO3. The resulting Ag3PO4 was captured on a 0.2 µm polycarbonate membrane filter and transferred to a glass vial and dried at 40 ˚C overnight.

Method 3: 0.1 mol MgOH2.6H2O was added to 3 x 1L of each Pi-stripped sample matrix, followed by 25 mL 1M NaOH solution and mixed occasionally to precipitate brucite over 10 mins. The precipitate was collected through centrifugation at 3500 RCF for 10 minutes and dissolved in the minimum volume of c.CH3COOH necessary to fully dissolve the brucite pellet. The acidic solutions were loaded at 5 mL.min-1 through 1 x 7 mL cation exchange resin, 1 x 10 mL DAX-8 column, 1 x ENV+ resin and onto a 12.5 mL anion exchange resin. Pi was eluted using 0.25M KCl at 1 mL.min-1. 0.01 mol MgOH2.6H2O was added and diluted to form a 100 mL solution. 2.5 mL 1M NaOH solution was added to a beaker containing the 100 mL solution and mixed occasionally to precipitate brucite over 10 mins. Following the brucite precipitation, the method proceeded as for Method 1, as described above, from

H2O2 was added to each vial and left to decompose any remaining organic compounds (minimum 3 hrs) and then removed by evaporation at 40 ˚C overnight, and finally washed with 3 x ~1.5 mL MilliQ water, centrifuged at 3500 RCF for 20 minutes and dried at 40 ˚C overnight.

Ag3PO4 samples were analysed for δ18Op on an IsoPrime100 mass spectrometer coupled to a varioPYRO cube elemental analyser. For each sample, 700 μg Ag3PO4 was weighed into a silver capsule with 800 μg carbon black, dried at 40°C overnight and converted to CO by pyrolysis in an ash crucible at 1450 C. The resulting gases are passed through Sicapent (phosphorus pentoxide) to remove water. The CO is separated from other impurities, namely N2, using a purge-and-trap system and helium carrier gas. 18O/16O is derived from the integrated mass 28 (12C16O) and 30 (12C18O; 14_C16_O; 13_C17_{O) signals from the sample CO pulse, compared to those in an} independently introduced pulse of pure CO reference gas. These ratios are then calibrated to the Vienna-Standard Mean Ocean Water (VSMOW) scale in per mille notation (‰) using standards – NBS127 (+9.3‰), EM Ag3PO4 (+21.7‰) and Acros Ag3PO4 (+14.2‰). The precision obtained from repeat analysis of standard materials is generally better than ±0.5‰VSMOW.