Assessment of the DGT technique in digestate to fraction twelve trace elements.

This study proposes an evaluation of the diffusive gradients in thin films technique (DGT) for studying trace elements in digested sewage sludge samples. Twelve elements were monitored by Chelex (Al, Cd, Co, Cr (III), Cu, Fe, Mn, Ni, Pb) and zirconia-DGT (As, Mo, Se) samplers exposed from 4 h to 9 days. Twenty-four hours' deployment time was suitable for most of the studied elements. However, short deployment led to insufficient element accumulation or non-establishment of steady state while long deployment (from 18 to 144 h depending on the element) led to saturation of the binding gels and/or competing effects with other major elements. In addition, this study showed that the matrix of the digested sewage sludge lowers the accumulation of some trace elements in the DGT samplers, leading to labile concentrations underestimation of roughly 10-30% (depending on the element). Moreover, compared to the conventional total dissolved elements measurement, DGT technique allowed to quantify 7 out of 12 labile elements whereas only 3 out of 12 dissolved elements were quantified. These results highlight the potential of DGT technique to assess labile trace elements in digestate samples, provided a careful adaptation of the deployment time as well as an evaluation of the matrix effect is performed.


Potential interference from digestate matrix on trace elements accumulation 116
To evaluate the potential interference from the digestate matrix on the diffusion and accumulation 117 of trace elements in the binding gels, the Chelex and Zr-DGT samplers were exposed in triplicate 118 to the digestate sample for 24 hours to load their diffusive gels with the digestate matrix. The pre-119 exposed diffusive gels were then recovered to build new DGT samplers with new Chelex and Zr 120 binding gels (henceforth named "soiled" DGT samplers). Additionally, triplicate DGT samplers 121 were built with new diffusive and binding gels as control in the experiment.  Tables S1 and S2.  131 To check the contamination of the binding gel brought by the "soiled" diffusive gel, three blank 132 DGT samplers were built with "soiled" diffusive gels and new Chelex and Zr binding gels. The 133 blanks were stored at room temperature (20±1°C) in a moistened plastic bag and disassembled 134 after 4 hours alongside the other samplers. 135 For statistical analysis of the results, a F-test was performed using Microsoft Excel 2013 to 136 determine the variances of the two sets of samples, then the two-tailed t-test was applied at 95% 137 confidence interval. 138

Size fractionation of labile elements 139
Fractionation of labile elements based on their size was investigated through the simultaneous 140 deployment of DGT samplers equipped with restricted or standard diffusive gels. The Chelex and 141 Zr-DGT samplers were deployed for 24 hours in 20 L of digested sludge sample continuously 142 stirred at 30 rpm. The deployment time was chosen according to the results obtained from the 143 experiment described in 0. 144 1.4 Analytical procedures 145

DGT-labile concentration 146
After retrieval, DGT samplers were rinsed with ultrapure water and disassembled to recover the 147 binding gels. The accumulated mass (m) of trace elements in each DGT sampler was determined 148 after elution of the binding gel. The Chelex binding gels were eluted in 2 mL of 1 M HNO 3 for 149 24 hours and the Zr binding gels in 2 mL of 5·10 -3 M NaOH and 0.5 M H 2 O 2 for 24 hours. Then 150 the concentration of trace elements in the eluents (C e ) were quantified by the inductively coupled 151 plasma mass spectrometry (ICP-MS) or microwave plasma atomic emission spectroscopy (MP-152 AES) (see section 0). The accumulated mass is determined according to equation (1) [15]: 153 , Eq. (1)  154 where V e is the volume of the eluents (2 mL) and f e is the elution factor (values are reported in 155 Table S3). 156 The concentration of labile trace elements, C DGT , in the sample is then derived using equation (2)  157 based on Fick's first law [16]: 158 , Eq. (2)  159 where Δ MDL is the thickness of the material diffusion layer (i.e. diffusive gel plus membrane), t is 160 the time of DGT samplers' exposure in the sludge, D is the coefficient of diffusion of the 161 considered element in the diffusion layer and A is the geometric area of the DGT holder window 162 (3.14 cm 2 ). The values of D were corrected for the average temperature (T) recorded every 10 163 min by a Tinytag data logger during each deployment using Stokes-Einstein relation [13]  where η is the viscosity of the water taken from the NIST chemistry WebBook [17]. The values 167 of D at 25°C used in our study for a standard diffusive gel are summarized in Table S4 in 168 supporting information. The D values for the restricted gel are equal to 70% of the D for a 169 standard gel, based on the work of Scally et al. [14] and Shiva et al. [18] as summarized in Table  170 S5. 171

Physicochemical analysis 172
The pH was measured with a Mettler Toledo pH electrode. The total solids (TS), volatile solids 173 (VS), total suspended solids (TSS) and volatile suspended solids (VSS) were measured according 174 to the French standard AFNOR NF T90-105 method. The supernatant recovered during the TSS 175 and VSS procedure was used to estimate dissolved trace elements (see section 0). 176

Sample treatment and trace elements analysis 177
At the beginning and at the end of each experiment, an aliquot of digested sewage sludge was 178 sampled to measure the total and dissolved elements' content. About 5 g of raw sample (total 179 content) or 2 mL of supernatant (dissolved content), recovered after centrifugation at 3.000 g for 180 20 min, were digested with 6 mL of 69% HNO 3  Moreover, quality controls at 5 and 10 µg/L were added to check the performance of the analysis. 187 The recovery was equal or above 86% for each element among all analyses performed by ICP-188 MS or MP-AES. 189

Method's limits of detection 190
The method's limits of detection were determined for each procedure (i.e. digestion or DGT 191 handling) to account for sample contamination. For the acid digestion procedure, ultrapure water 192 blanks were treated alongside samples with the procedure described in 0. Blank DGT devices 193 were prepared in duplicate and treated alongside exposed devices during the "short" and "long 194 term" experiments (see section 0). The method's limit of detection (MLD) and quantification 195 (MLQ) were calculated according to IUPAC as the average plus three or ten times the standard 196 deviation of the blanks for MLD and MLQ, respectively. 197  Table S6. For each parameter, the difference in 201 percentage is low (ranging from 4% to 7%). 202

Results and discussions
The total and dissolved element concentrations of the samples is reported in Table S7. Dissolved 203 element concentrations were below the MLQ except for As, Fe and Mn. A small discrepancy 204 between the samples is observed for the dissolved Fe (9% difference) whereas a high discrepancy 205 for the dissolved As (75% difference) and Mn (31% difference). Regarding the total element 206 concentrations, only Se is not quantified in the samples. A small discrepancy is observed for Fe 207 and Mn (≤ 9% difference) between the samples, whereas a discrepancy higher than 10% is 208 observed for the other elements. 209

Steady state establishment 211
During the "short term" experiment, Cd, Cu, Mo and Pb were below the MLD whereas Al, Cr 212 (III) and Se were below the MLQ of DGT deployment. Therefore, these elements are not 213 discussed further in this section. According to DGT theory, steady state is rapidly established in 214 the sampler (≤1h, [19]) and the accumulated mass should behave linearly over time. 215 The mass of elements accumulated over time on the Chelex and Zr-DGTs is reported in Figure 2. 216 We observe a linear accumulation trend from 0 to 24h for Co, Mn and Ni. Therefore, the system 217 (DGT-digestate) is rapidly in steady state and Eq. (2) holds for these elements regardless of the 218 deployment time (until 24h at least). We also observe a linear accumulation trend for As and Fe 219 from 0 to 18h and from 4 to 24h, respectively. For As, it indicates that the steady state is rapidly 220 reached and that Eq. (2) holds up to 18h deployment. Deviation from linearity after 18h is likely 221 caused by competing effect. Indeed, Zr-binding gels are known to bind both As and P [20] that 222 are chemical analogous (in the form of arsenate AsO 4 3and phosphate PO 4

3-
). Consequently, P 223 could have replaced As on the binding gel. This hypothesis is supported by data shown in Figure  224 S1 where P displays the same linear behavior as As, but its accumulated mass on the Zr-binding 225 gel was about 40-fold higher than As up to 24 h deployment time. 226 Fe presents a unique behavior since we observed linearity only after 4h, indicating delayed 227 establishment of steady state in the sampler. Such behavior can be explained by the properties of 228 Fe complexes (partially labile complexes) or by interactions between Fe and the diffusive gel 229 [21]. Such properties indicate that Eq. (2) does not hold at 4h deployment and its use will result 230 in an underestimation of C DGT . Indeed, we calculated C DGT from the regression line and compared 231 to the value estimated with Eq. (2) using 4 and 24h deployment and we found that C DGT is highly 232 underestimated at 4h (i.e. 70% less) than 24h deployment (i.e. 16% less). 233 We observed the establishment of steady state in the samplers for all the quantified elements, 234 therefore the principle of DGT are validated for short deployments (≤24h) in the studied digestate 235 matrix. However, the non-significant accumulation of Al, Cd, Cu, Cr (III), Mo, Pb and Se during 236 this "short term" experiment suggests that these elements may be countered by deploying the 237 DGT samplers longer. 238  During the "short term" experiment, we observed an accumulation of As in the samplers over 268 time, whereas not anymore during the "long term" experiment. Such behavior is consistent with 269 the competing effect of P already highlighted and discussed in section 0. 270 Regarding Se, we cannot state that its accumulation trend is linear after 24h deployment time 271 (R 2 <0.6). Consequently, this element cannot be correctly estimated using Eq.

Impact of digestate matrix on accumulated labile elements 280
To check the interference of the digestate matrix on the trace elements accumulation by DGT 281 samplers, some diffusive gels were pre-exposed for 24h to the digestate before deployment in a 282 well-defined spiked solution as described in 0. Since As (III) and Mo (VI) were below the MLQ 283 of the DGT blanks, these elements are not further discussed in this section. 284 The mass of the elements accumulated by the control and "soiled" DGT samplers are presented in 285  Except for Se (IV) and Cd (II), we observed that the accumulated mass of the elements measured 289 by the control DGT samplers is significantly higher (p<0.05) than the one measured by the 290 "soiled" DGT samplers. In particular, the "soiled" DGT devices accumulates 11%, 18%, 24%, 291 28% less Co (II), Ni (II), Pb (II), Cu (II), respectively, compared to the control DGT devices. 292 Such low accumulation could be even more pronounced in the digested sludge since its pH is 293 higher than the one measured in the spiked solution of this study (4<pH<6, Table S1). A high pH 294 is favorable for element binding to organic matter [23], at least for cations. In fact, organic matter 295 is known to diffuse within diffusive gels [16,[24][25][26]. We hypothesize that organic matter 296 accumulated on the diffusive gel during pre-exposure and promoted element sorption onto the 297 gel, resulting in a delay of element diffusion as already observed by Davison et al. [27] for Cu 298 with river or soil organic matter. 299 Here, we showed that DGTs pre-exposure to the matrix of the digestate lowers the accumulation 300 of most of the studied trace elements, leading to underestimation of the labile element 301 concentrations in the medium. 302

Sensitivity of DGT method 304
The limit of detection and quantification of the method for DGT (MLD DGT and MLQ DGT ) are 305 given in Table 1. Compared to the instrumental limit of quantification (which only counts for the 306 analytical sensitivity of the ICP-MS or MP-AES), the MLQ DGT is at least two times higher (data 307 not shown), meaning that some contamination of the samplers occurred during the samplers 308 handling. 309 Additionally, we compared the MLQ DGT to MLQ for dissolved element (MLQ dissolved , Table 1).It 310 arises that DGT greatly increased the sensitivity for element monitoring in the digested sludge 311 than the conventional method (i.e. dissolved elements measurement). In particular, the MLQ DGT 312 for Al, Cd, Co, Cr (III), Pb and Se is more than 1000 lower than the MLQ dissolved . For the other 313 elements the ratio decreases in the following order Fe>Ni>Cu>Mn>As>>Mo. This high 314 sensitivity is inherent to the sampling method since DGTs concentrate analytes whereas dissolved 315 elements measurement requires acid digestion of the sample and subsequently its dilution. 316 However, we must stress that both methods do not target the same chemical fraction since the 317 labile fraction targeted by DGT represents only a part of the dissolved elements. 318 Besides, from a monitoring point of view, DGT appears a very interesting method since it 319 allowed to quantify several of the labile elements during the experiments (Table S8) whereas it  320 was not possible for most dissolved elements (Table S7). Therefore, we consider DGT as a 321 sensitive method to monitor trace elements in digested sludge. 322 Table 1 Surprisingly, a significant higher concentration of labile As and Ni was estimated with restricted 339 gels (p<0.02) than standard gels. Such results are not consistent since restricted gels have smaller 340 pore size (i.e. <1 nm) than standard gels (i.e. >5 nm) and it should not allow diffusion of a higher 341 amount of labile elements. Such discrepancy could derive by the use of a non-adapted D value for 342 the restricted gels. In fact, the values reported in Table S5

Practical implementation for other digestate samples 350
In the studied digestate, the "short" and "long term" experiments revealed the following optimal 351 deployment times for each element (

354
A 24h deployment appears a good compromise to allow quantification of most elements. 355 However, these results cannot be generalized to any digestate sample given the variable 356 composition of digestate in terms of trace elements and organic compounds which may interfere 357 with elements' accumulation in DGTs. Therefore, preliminary tests to optimize the deployment 358 time are strongly recommended. In general, we advise to avoid long deployment time because 359 22 saturation of the binding gel can occur due to the presence of other major compounds. Very short 360 deployment time (i.e. <4h) should also be avoided, since the mass of trace elements may not 361 accumulate in the device or the steady state is not established. 362 The studied digestate matrix altered accumulation of labile elements in DGT devices by 10-30% 363 for Co (II), Ni (II), Pb (II), Cu (II). Such alteration was due to diffusion of organic matter in the 364 sampler from the digestate matrix. This behavior is probably expected in most digestate samples 365 given their high organic matter content [28,29]. Further studies are needed to determine the 366 diffusion rate of trace elements in the presence of digestate matrix. From such work one should 367 be able to correct for matrix effect with the aim to accurately determine labile trace elements 368 concentrations. Unless this, it is safe to limit interpretation of labile concentration established 369 with DGTs to general trends (e.g. evolution over time, order of magnitude) in order to limit 370 misinterpretation of the absolute DGT labile trace elements concentrations. 371 Finally, size fractionation by coupling the restricted and standard gels was investigated in this 372 study. Our results show the presence of large labile complexes for Fe (>1 nm) and small labile 373 complexes for Al, Co, Cr (III) and Mn (<1 nm). However, these results must be confirmed and 374 cannot be generalized at this stage. 375

Interpretation of DGT fractionation 376
One of the main objective when performing trace element fractionation is to predict their bio-377 accessibility. The DGT technique demonstrated to perform well mostly in natural waters and 378 soils [7]. Currently, data regarding the relationship between DGT-labile element concentrations 379 and their bio-accessibility in digestate are very sparse. To our knowledge, only the study of 380 Bourven et al. [8] addressed this topic. They showed, in the context of whey anaerobic digestion, 381 that DGT-labile Cd content is linked to the initial alteration of biogas production and enzymatic 382 activities (i.e. β-galactosidase and TTC-dehydrogenase). However, such correlation was absent 383 after 21 days of anaerobic digestion. DGT based fractionation of Cd appears, therefore, 384 encouraging to predict its bio-accessibility, but not straightforward. Similar works could be 385 performed for several trace elements and in various digestates. Therefore, new studies are 386 required to fully establish the extent to which DGT fractionation can be used to predict elements 387 bio-accessibility in digestates. 388

407
The supporting information is available at the following link (to be mentioned). 408