3.1 From the Idea to the Proof of Concept

To evaluate the applicability of FC-ICP-MS for Te(IV)-Te(VI) speciation, it is crucial to identify chromatographic conditions that fulfill the key requirements of this low-resolution technique, specifically ensuring that the two target species possess completely distinct chemical properties. This characteristic enables the selective retention of one species in the chromatographic column, while the other species remains unretained.

The Pourbaix diagram (obtained using Hydra-Medusa software, Fig. 2) clearly shows that in acidic nonreducing environments (approx. pH < 3, and with an Oxidation–Reduction Potential around 0.75 V) the above discussed requirement is met: Te(IV) exists as a cationic species (TeOOH+) whereas Te(VI) occurs as a neutral species (H2TeO4).

Fig. 2

Pourbaix diagram of tellurium species constructed using Hydra-Medusa software [39] (T = 25 °C, Te total concentration = 1 × 10−9 M)

Under these conditions (e.g., in diluted HNO3), cationic Te(IV) is expected to interact with a strong cation-exchange resin, while Te(VI) should pass through the column without any interaction.

For the first preliminary experiment, all chromatographic parameters were selected arbitrarily, leveraging our expertise in the field. The chosen parameters included: HNO3 concentration (0.25 M), column length (50 mm), column width (3 mm), and sample flow rate (1.14 mL/min). The results of this experiment are presented in Fig. 3.

Fig. 3

Frontal chromatogram of a 0.25 M HNO3 solution containing 1 μg/kg of Te(VI) (red line), Te(IV) (gold line), and both Te(VI) and Te(IV) species (blue line). Experimental conditions are detailed in the figure

As can be seen, when a mixture of Te(IV) and Te(VI) is introduced into the FC system, two distinct sigmoidal curves are observed. At this stage, we recorded the frontal chromatograms by monitoring the interference-free 125Te isotope, which is the default and preferred option for Te determination using ICP-MS. The first curve corresponds to unretained neutral Te(VI), while the second, broader curve is related to cationic Te(IV), which is partially retained by the stationary phase. Notably, two well-defined plateaus can be distinguished,for Te(VI) between 60–100 s and for Te(IV) between 250–300 s, allowing for the quantification of both species by measuring the height of their respective fronts. As a final remark, the chromatogram is reported as corrected by internal standard 73Ge which occurs as a neutral species (H2GeO3) under these conditions, thus it is not retained by the cation-exchange resin. This choice was made to compensate for any fluctuation of the 125Te signal (see Fig. S2 for a comparison with 125Te and 73Ge raw data).

In addition to the effective separation of Te(IV) and Te(VI), the elution time is relatively short, around 5 min. However, there is space to further shorten this time to make it suitable for routine analysis, while still maintaining comparable separation efficiency. A Design of Experiments (DoE) approach was therefore employed to optimize the chromatographic parameters (see next paragraph). Furthermore, the exploitation of the so-called “carbon effect” and the careful selection of isotopes were also investigated to further enhance the method's sensitivity. These steps are essential to meet the analytical requirements for detecting ultratrace levels of Te in environmental samples. All these optimizations will be detailed in the following paragraphs.

3.2 Design of Experiment: Optimizing Chromatographic Parameters

Optimal conditions were sought using a full factorial 24 design, which systematically evaluated the effects of (i) column length (25 and 50 mm), (ii) column width (2 and 3 mm), (iii) sample flow rate (1.39 and 1.14 mL/min), and (iv) HNO₃ concentration (0.25 and 0.50 M) on two responses: analysis time and chromatographic resolution. The aim is to find analysis conditions that maintain sufficient resolution to effectively separate the two species while keeping minimum analysis time. The specific conditions tested in each experiment are outlined in Table 1 and Table S3.

Table 1 Chromatographic conditions and corresponding levels used in the Design of Experiments (DoE) model for data standardization and centering, selected for all experiments within the explored domain

The resulting 16 conditions (full-factorial DoE) were applied to the analysis of Te(IV)-Te(VI) mixture solutions (concentrations = 1 μg/kg). The effect of each single parameter can be immediately observed in Fig. 4.

Fig. 4

Effects of chromatographic parameters on the elution profile of Te species: a column length; b column internal diameter; c HNO3 concentration, and d sample flow rate. The chromatographic conditions for each test are reported in the corresponding panel

The chromatograms show that all the parameters influence the elution profile. Since Te(VI) does not interact with the cation-exchange resin, its elution time is very slightly influenced by sample flow rate and column geometry whereas the chemistry of the mobile phase does not play any role. Contrarily, the elution time and width of the front of cationic Te(IV) are significantly affected by all investigated parameters. Generally, increasing the sample flow rate and decreasing column size speed up the elution of Te species as expected. Moreover, higher HNO3 concentrations notably reduce analysis time at the expense of the resolution, as H⁺ ions enhance the rapid elution of Te(IV).

Multivariate regression analysis was conducted to quantitatively assess the effects and optimize the chromatographic conditions that were initially observed qualitatively. Once the resolution and the analysis time were rigorously determined (see details in Fig. S1), their responses were modeled by second-order equations including interaction terms. Figure 5 shows the obtained coefficients, while the surface responses are depicted in Fig. S3.

Fig. 5

Coefficients resulting from multivariate regression analysis on the data presented in Table S3 to model a chromatographic resolution and b analysis time. Green bars mean statistically significant parameters with a p-value of ≤ 0.05 and red bars mean not statistically significant factors

The goodness of the regression model is demonstrated by the high explained variance (94.4% for the resolution model and 99.1% for the analysis time model) and the low, randomly scattered residuals, as illustrated in Fig. S4.

As shown in Figs. 5a-b, all main effects have a statistically significant impact (p-value ≤ 0.05) on both resolution and analysis time, except for the sample flow rate, which does not significantly affect resolution. This is consistent with the data in Fig. 4, where column geometry (length and internal diameter, i.d.) and acid concentration are identified as the primary factors influencing responses. Furthermore, interaction terms involving column geometry and HNO3 concentration significantly affect both resolution and analysis time, while other interaction terms are negligible. The surface response plots in Fig. S3 provide further illustration of these findings. As expected, regions within the experimental domain that exhibit high resolution are also associated with longer analysis times. This underscores the need to find a compromise in chromatographic parameters. As summarized in Table S1, resolutions varied from 0.38 to 1.77, and analysis times ranged from 59 to 290 s. Given the low separation capacity of the FC system, a resolution greater than 1 is considered satisfactory. Among the tested conditions, the chromatographic run with all parameters set to high levels (see experiment No. 9 in Table S1) proved to be optimal, achieving a resolution of 1.05 and a very short analysis time (147.9 s). The chromatogram in Fig. S5 shows effective separation between Te(IV) and Te(VI), allowing for their speciation analysis in around 2.5 min. Anyway, we opted to slightly enhance the resolution by decreasing the HNO3 concentration from 0.5 M to 0.43 M: this straightforward adjustment is expected to enhance chromatographic separation while also unlocking the direct analysis of soil and sediment extraction solutions for geochemically reactive elements, utilizing the standardized method with 0.43 M HNO3 [38]. With this modification, we finally achieved a resolution of 1.12 and an analysis time of 152.6 s (Fig. 6).

Fig. 6

Frontal chromatogram obtained for a Te(IV)-Te(VI) mixture solution (both species at a concentration equal to 1 μg/kg) under optimized conditions (column internal diameter = 3 mm, column length = 50 mm, HNO3 concentration = 0.43 M, and sample flow rate = 1.39 mL/min). Red and gold regions denote the time intervals utilized to determine the height of the fronts

The front height of Te(IV) can be reliably estimated as the average signal within the range of 50–55 s, while Te(VI) can be quantified within the range of 150–155 s. Given that these conditions were satisfactory, further optimizations were performed using the following chromatographic parameters: column length = 50 mm, column width = 3 mm, HNO3 concentration = 0.43 M, sample flow rate = 1.39 mL/min.

As a final remark, we verified that Te species do not undergo interconversion and that their concentrations remain stable for at least 72 h in 0.43 M HNO3.

3.3 Increasing the Sensitivity: Carbon Effect and Isotope Selection

Once optimized the rapid elution of Te species, we assessed the possibility of enhancing the sensitivity of the method by the so-called “carbon effect” [36], i.e., the matrix effect induced by simple organic compounds that can significantly enhance the limit of detection (LOD) for elements with high ionization potential, including Te. Although the exact mechanism behind this signal enhancement remains unclear, it seems primarily ascribable to improved ionization efficiency through charge transfer between C+ ions and analyte atoms in the plasma, increasing signal intensity [36].

Organic modifiers were introduced post-column in the FC-ICP-MS to prevent any alteration of the chromatographic system (refer to the configuration in Fig. 1). The flow rate of the additive was adjusted to minimize the online dilution of the samples (dilution factor of approximately 1.18). As shown in Fig. 7, we tested three different alcohols, namely ethanol (EtOH), methanol (MeOH), and 2-propanol (iPrOH) at three different concentrations for the analysis of a Te(IV)-Te(VI) mixture solution (1 µg/kg). For comparison purposes, an experiment was conducted using ultrapure water as the additive to simulate the same dilution.

Fig. 7

Evaluation of the carbon effect using three different organic modifiers at three concentration levels. The sensitivity enhancement factor was calculated by normalization against ultrapure water sensitivity. Te(IV)-Te(VI) mixture solutions (1 µg/kg) were analyzed for this purpose

Regardless of the organic modifiers used, sensitivity significantly increases. The best result was achieved with 4% V/V MeOH, which increased the 125Te signal by a factor of 2.9 for both Te(IV) and Te(VI) species. No specific effects instead were found for the internal standard 73Ge.

The step forward involved the choice of the best Te isotope to be used to improve sensitivity without incurring significant interference. Tellurium presents challenges for ICP-MS determination due to the numerous isotopes suffering from isobaric interference from xenon (Xe) isotopes originating from plasma gas impurities. Additionally, the most abundant isotope, 130Te, is affected by isobaric interference from 130Ba and 130Xe. This Te isotope was not therefore considered in this study. We considered the following isotopes:

125Te: 7.14% abundant, with no known interferents, the default isotope in ICP-MS analysis [1].

126Te: 18.95% abundant, interfered with by 126Xe (0.09% abundant).

128Te: 31.69% abundant, interfered with by 128Xe (1.91% abundant).

Signal corrections were applied based on the equations reported in Table S4. Sensitivities were determined from the slopes of the calibration curves within the range 0.01–1 µg/kg. Excellent linearity was observed for all studied isotopes, with R2 > 0.9998 (Fig. S6). Enhanced analytical signals compared to 125Te were observed for 126Te (2.7-fold increase) and 128Te (4.7-fold increase). Despite the higher sensitivity offered by 128Te, the best compromise between sensitivity and signal-to-noise ratio was achieved using the 126Te isotope. This is primarily due to the more significant corrections required for Xe interference, which greatly increases signal noise. Consequently, the LODs were achieved on the 126Te channel: 1.0 ng/kg for Te(IV) and 1.3 ng/kg for Te(VI). For comparison, 128Te under the same conditions provided LODs of 2.8 ng/kg for Te(IV) and 2.7 ng/kg for Te(VI), while uncorrected 125Te produced LODs of 2.7 ng/kg for Te(IV) and 2.5 ng/kg for Te(VI). This difference can be attributed to the higher variance induced by the much more marked interference correction needed for 128Te, supporting the greater reliability of the 126Te isotope which was chosen as the optimal channel to be monitored. All LOD values were estimated by analyzing ten replicates of 5 ng/kg standard solutions of Te(IV) and Te(VI) (see Fig. S7 for 126Te signals): this concentration is approximately five times the standard deviation of the background signal, in line with European guidelines for estimating LOD [40].

The notably low LOD values warrant careful consideration, as the simultaneous presence of both species might introduce errors, particularly at ultratrace levels when one species is significantly orders of magnitude more prevalent [31, 32]. To address this, the method's predictive accuracy was tested by analyzing nine different Te(IV)-Te(VI) mixture solutions with various concentrations (10, 100, and 1000 ng/kg). The findings are summarized in Table 2.

Table 2 Quantification of Te(IV) and Te(VI) in spiked ultrapure water solutions. The FC-ICP-MS analysis was carried out under optimized conditions. Uncertainties are expressed as two times the standard deviation (n = 3)

The recoveries are close to 100% in most cases, validating that the method performs well for Te(VI) with minimal loss or enhancement. As expected, significant overestimation occurred only under extreme conditions, specifically when the Te species ratio was 1:100, where a relative recovery of 200% was observed for both Te(IV) and Te(VI). It is important to note that these high relative recoveries are associated with very low absolute errors, around 10 ng/kg (1% of the interferent species signal), which is approximately three times the limit of quantification. Furthermore, the presence of an interferent Te species in concentrations two orders of magnitude higher (1 µg/kg) does not represent a realistic scenario. As demonstrated, we can safely state that concentrations below 100 ng/kg do not present any cross-interference issue given that the absolute error of 1 ng/kg lies close to the LOD.

The repeatability was assessed through three replicates of the nine standard solutions detailed in Table 1. The relative standard deviations (RSDs) observed were 4.2% for Te(IV) and 2.8% for Te(VI), aligning well with the expected RSD values for conventional ICP-MS measurements. Inter-day precision was assessed by analyzing both single standard solutions and Te(IV)-Te(VI) mixture solutions over three consecutive days. The results showed inter-day precision below 5% for both Te species, further demonstrating the robustness of the method.

3.4 Application to Mineral Water, Sediment, and Soil Samples

Six natural water samples were analyzed to validate the applicability of the FC-ICP-MS method across various mineral compositions. The physiochemical parameters and major ion concentrations are provided in Table S2. All samples contained extremely low levels of Te species, with Te(IV) consistently below the LOD, while Te(VI) was detected at concentrations up to 7.4 ng/kg (Table 3). No clear correlations between Te speciation and the specific characteristics of the water samples were observed, suggesting that the geochemical setting and geographical origin of the water source may influence Te distribution.

Furthermore, mineral waters spiked with very low concentrations of Te(IV) and Te(VI) (5 ng/kg) were analyzed for validation. This choice was made because that no certified reference materials are available for Te speciation. As shown in Table 3, recoveries were (97 ± 5)% for Te(VI) and (115 ± 16)% for Te(IV). This demonstrates the reliability of the method for detecting ultratrace levels of Te species under challenging conditions, with the spike concentration approaching the limit of quantification.

Table 3 Te(IV) and Te(VI) quantifications in mineral water samples and mineral water samples added with 5 ng/kg of both Te species. Quantifications were performed by external calibration. Uncertainties are expressed as two times the standard deviation (n = 3)

Similarly, we applied the developed method to determine the weakly bound, geochemically reactive fraction of Te species in sediments and soils. The standardized extraction procedure ISO-17586:2016 [38] was employed. As this protocol involves an extraction using 0.43 M HNO3, it is fully compatible with the direct analysis of the filtered samples without requiring dilution. This feature significantly enhances analytical sensitivity, enabling the detection of ultratraces even in solid samples. The analyses (Table 4) show, as expected, extremely low concentrations of Te species (with concentrations in solid samples in the range 0.34–3.5 µg/kg and 0.13–0.56 µg/kg for Te(VI) and Te(IV), respectively). These results are consistent with the uncontaminated setting in which the samples were collected, i.e., a remote area presenting negligible local sources of contamination. The ability to detect such ultratrace levels highlights the method's capability to accurately measure Te even in samples that should be regarded as background references. Concerning Te species distribution, Te(VI) is largely the predominant species in sediment samples, with Te(IV) in some cases falling below the LOD. This consideration applies also to the certified reference material (total Te concentration equal to 24 µg/kg), where the weakly bound Te fraction is made of Te(VI), only.

Table 4 Te(IV) and Te(VI) quantifications in soil and sediment samples, as well as the samples spiked with 50 ng/kg of both Te species (100 ng/kg for the certified reference material GBW07302). Spiked solutions were analyzed 72 h after the spiking. Quantifications were performed by external calibration. Uncertainties are expressed as two times the standard deviation (n = 3)

Regardless of the extraction capabilities of the standardized ISO-17586:2016 procedure (and hence what “geochemical reactive species” clearly refers to), we spiked the post-extraction solutions to verify whether the co-extracted substances interfere with the Te species concentrations and distributions. As can be seen in Table 4, Te(IV) and Te(VI) remain stable in the extraction media even after 72 h, with recoveries not statistically different from 100%, thus confirming the consistency and accuracy of the obtained data. These results strongly highlight also the robustness of this protocol: the analyzed samples present significant geochemical anomalies, including elevated levels of potential interferents (i.e., Sn and W for GBW07302, as well as Ni and Cr in the samples from the Ventina Valley [41]). This demonstrates the method's reliability, even when applied to challenging matrices.

3.5 Comparison with Existing Methods

Table 5 presents a comparison of representative analytical features between the developed method and recent protocols reported in the literature for Te speciation analysis in water and soil samples. Both chromatographic and non-chromatographic methods were considered in this evaluation. No focus will be given to extraction techniques from soil or sediment samples since a standardized extraction protocol was used here (and therefore this aspect was neither optimized nor developed in the present study). Differently, any other relevant pretreatment (e.g., pre-concentration, derivatization) steps will be discussed.

Table 5 Comparison of the analytical features between existing procedures for the determination of Te species in natural water or soil samples and the proposed method

Currently, only a limited number of chromatographic methods are available for the simultaneous determination of Te species [27, 28]. However, these methods fall short compared to the proposed approach, particularly in terms of LOD. Most existing strategies are inadequate for quantifying Te in uncontaminated natural matrices, as their LODs typically exceed 1 µg/kg, restricting their practical application. While some methods offer analysis times comparable to the FC-ICP-MS (around 300 s for the fastest HPLC-based strategies in the literature [28]), they do not match the sensitivity of the proposed method.

The lowest LODs (down to sub-ng/L levels) have been achieved using hydride generation systems coupled with ICP-MS/MS [14]. Despite their sensitivity, vapor generation-based methods do not support the simultaneous determination of both Te species [14, 43]. Typically, they quantify either Te(IV) or total Te, with the latter requiring the pre-reduction of Te(VI) using agents like Ti(III) in combination with strong acids at high temperatures. However, the reduction with TiCl3 can be problematic since nitrates in water samples may interfere [26]. This additional step complicates the process, extends the analysis time, and reduces environmental sustainability.

The limited LODs have also driven the adoption of preconcentration techniques to enhance sensitivity. However, these methods often suffer from analyte loss and discrepancies in the preconcentration behavior of Te(IV) and Te(VI). This issue frequently necessitates the pre-reduction of Te(VI) to Te(IV), further complicating the procedure and preventing the simultaneous determination of both species [21, 22]. Nevertheless, the LODs achieved through these techniques are comparable or, in some cases, superior to those of the present work.

Electrochemical methods also face similar challenges, as they are selective for Te(IV) and do not detect Te(VI), which is non-electroactive [26]. Although preconcentration is not required, these methods are limited to Te(IV) and, while they can achieve LODs as low as 5 ng/kg, they are unsuitable for comprehensive speciation analysis [26].

By comparison, the proposed method offers notable LODs similar to or better than most of the state-of-the-art techniques for Te analysis, with the great advantage of allowing a direct and fast speciation analysis, having minimal or no sample workup. As a potential drawback, this method may appear to require a much larger sample volume compared to other techniques (e.g., HPLC) due to the high flow rate at which the sample solution is continuously fed to the ICP-MS. However, the actual sample volume needed is still relatively small, with just 5 mL sufficient for a single analysis: this volume should not pose a critical limitation for the analysis of environmental samples. Finally, we successfully addressed potential limitations arising from common matrix effects that could impact chromatographic separation through the validation of environmental samples. The traceability of the investigated interferents is ensured by data reported on the composition of all analyzed water samples (Table S2) and soils and sediments (standard GBW07302 and reference [37]).

In summary, the simplicity and strong analytical performance of this approach make it highly suitable for high-throughput routine Te speciation analysis, while also offering an environmentally friendly option.