Particles in 3 liters of the WFGD slurry from TPP Šoštanj were first separated into fine and coarser fractions labelled as FF and CF, with the help of a hydro cyclone. Half of the FF was returned into the separation process, where it was again separated into fine and coarser fractions labelled as FF1 and FF2. Four different size fraction samples were thus obtained in which FF contains both granulation sizes of FF1 and FF2. Granulation analyses of gypsum samples were conducted with Microtrac PSA FRA 9200 based on Fraunhofer laser diffraction. The size distribution of particles in each sample is presented in Figure 1.

Samples were then oven dried at 38˚C for a week to constant weight, in an electric drier/oven MEMMERT UFE 500. Moisture content lost during drying was 7.9% for FF, 8.7% for FF1, 3.2% for FF2 and 4.8% for sample CF.

After drying, samples were homogenized in the agate containers of a planetary mill (Fritsch planetary mill Pulverisette 7).

An aliquot of the original gypsum sample (not oven dried) was also kept for studies of the potential influence of moisture on Hg release during the temperature fractionation process.

Mercury compounds chosen for examination are presented in Table 1.

These pure compounds were “diluted” by mixing with CaSO₄∙2H₂O powder (Sigma-Aldrich Co. LLC.; purity

99% - 101% based on anhydrous substance). This substrate resembles gypsum from the WFGD in coal burning power plants. Each pure Hg compound (10 - 12 mg) was added to 5 g of substrate powder. The mixtures were carefully homogenized by mixing in zirconium containers placed in a planetary mill (Fritsch planetary mill Pulverisette 7). Homogenization was achieved in short mixing intervals of 10 minutes in order to avoid increase of temperature. The bulk homogeneity of each mixture sample (80 - 140 mg) was checked by wet digestion of 6 independent analyses of total Hg, followed by Hg determination by cold vapor atomic absorption spectrometry (CV AAS) after reduction with SnCl₂ [46] .

Only the sample with the highest Hg content (sample FF) and finer particle size was chosen. It was spiked with ten different pure Hg compounds (Table 1), each previously mixed with CaSO₄∙2H₂O (1 - 7 mg). The mixtures were placed in a planetary mill (Fritsch planetary mill Pulverisette 7) for several homogenization periods of 10 minutes.

Lignite samples were obtained at random from a conveyer belt during a normal working day to ensure representative samples. The combined mass of the samples was around 3 kg. They were dried to constant weight at 38˚C for a week in an electric drier/oven MEMMERT UFE 500. After the drying process, the combined samples

were ground with a ball mill and sieved with a Fritsch Pulverisette 0 Vibratory Micro Mill to a particle size of less than 250 µm.

An RA-915+ Hg Analyzer with a PYRO-915+ pyrolysis unit was used for the detection of Hg. The instrument was developed for direct determination of total Hg by combustion of samples and cold vapor atomic absorption detection (CV AAS) with Zeeman background correction [50] .

The apparatus used for temperature fractionation was a home-constructed device (Figure 2). It consists of a

gas tank (1) for the supply of carrier gas, a flow meter (2) for flow adjustment and quartz tubes. The first quartz tube, with an outer diameter of 20 mm and inner diameter of 16 mm, was placed in the electric tube furnace (3). The quartz boat used for samples (4) was carefully positioned in the first quartz tube in the middle of the electric tube furnace Nabertherm RT 50/250/11 where the thermocouple is located. The quartz boat is semi-circular, with an inner diameter of 5 mm and length of 35 mm. The tube furnace consists of a ceramic tube, inner diameter 50 mm, and two plugs of ceramic fiber on each side. Each plug has a drilled hole in the middle for a tight fit to a quartz tube to reduce heat loss. The second quartz tube, with same dimensions as the first, was packed with quartz wool (5) and heated to 700˚C in a small electric furnace with inner diameter of 27 mm (6) to ensure the transformation of all volatile Hg compounds to elemental Hg and to retain any particles that may be released from the sample. The Lumex Pyro 915+ pyrolysis unit, with dimensions of 350 × 350 × 120 mm (7), ensured additional decomposition of any remaining volatile compounds of Hg that might interfere with the atomic absorption measurement [50] . Elemental Hg was detected by the Lumex Ra-915+ atomic absorption detector with Zeeman correction with detector dimensions of 470 × 210 × 110 mm (8), which was connected directly to a computer (9) for data collection. An impinger containing H₂SO₄-KMnO₄ solution (10) was connected to the exhaust from the Lumex Ra-915+ to retain Hg(0) in solution by oxidation.

Around 50 mg of each gypsum sample (FF, CF, FF1 and FF2) and each spiked FF sample was weighed in a quartz boat. When pure Hg compounds mixed with CaSO₄∙2H₂O were tested only 10 - 30 mg of the sample was weighed into the sample boat and as small amount of pure Hg compound as possible, not to exceed the upper limit of the detection range (2 - 30,000 ng∙m⁻³) [50] . However, for a signal to even be recognized as a peak of released Hg, at least 20 ng of Hg should be present in the sample―due to noise caused by the detector Lumex Ra-915+. The results of mass balance for different samples showed that the intensity of the peaks does not match perfectly to recording units (ng∙m⁻³), due to a drifting measuring baseline, as a software deficiency. However, these units serve for easier comparison of the signal intensity. To resolve the drifting baseline problem, corrections were made on every thermogram by deducting the thermogram area, obtained by averaging several thermograms of blank samples. In addition to baseline correction, temperature correction was also considered, due to difference in the temperature measured at the sample and that recorded by the electric tube furnace (Figure 2, (3)). An independent measuring thermocouple was fused directly into the quartz tube, where the sample is located, giving precise temperature readings. After several experiments, the average of the temperature data was calculated and deducted from the data recorded by electric tube furnace to provide correct temperature. The sample boat (Figure 2, (4)) was carefully positioned in the middle of the quartz tube which was heated from room temperature to 700˚C with a linear heating rate of approximately 2.2˚C∙min⁻¹. Nitrogen (99.996%) was used as a carrier gas to flush Hg vapor released from the sample through the measurement train at a flow rate of 1 L∙min⁻¹. To verify that the peaks, shown on computer, really belong to Hg release, impingers were changed before every peak release and content was measured to determine retained Hg concentration.

To receive correct readings, the pyrolysis unit (Figure 2, (7)) has several different setting modes to choose the most appropriate gas flow rate and heating temperatures according to the type of a given sample. In our case, mode 1, using a gas flow rate of 0.8 - 1.2 L∙min⁻¹ and the atomizer at 680˚C - 740˚C was selected, which suits sample materials such as soils, sediments, ores, coal and minerals. Data are interpreted by the Lumex software as a signal for continuous measurement of the concentration in ng∙m⁻³ shown on the ordinate axis as a function of time in seconds displayed on the abscissa axis. Due to known heating rate and precise temperature readings, the concentration signal per time was adapted to a concentration signal per degree temperature, ˚C, to observe Hg release on increment of temperature.

For quantitative determination of Hg in gypsum samples, around 0.5 g of sample was digested in acid solution of HNO₃-HClO₄ in a ratio of 1:1 (2% v/v) and H₂SO₄ (5% v/v) in Milli-Q water at 240˚C for 20 minutes. Hg content was then determined by the CV AAS method (Automatic Hg Analyzer Model Hg-201, Sanso Seisakusho Co., LTD). For every sample at least two replicas were prepared and measured. For testing the homogeneity of samples, the same procedure was used for 6 replicas, to check the repeatability of the measured Hg concentration.

Hg released from the sample during temperature fractionation was trapped into impinger solution (4% W/V KMnO₄, 10% v/v H₂SO₄ in Milli-Q water) placed at the exit of a Lumex detector (Figure 2, (7)). The permanganate solution was sequentially pre-reduced with NH₂OH∙HCl to destroy the excess of potassium permanganate. An aliquot of the sample was reduced to Hg(0) with an acid solution of 10% SnCl₂ and determined by CV AAS (Automatic Hg Analyzer Model Hg-201, Sanso Seisakusho Co., LTD) [46] . The quantity of Hg released from the sample was used to assess the mass balance, by comparing the amount of Hg contained in solid sample aliquots and the amount trapped in the impingers.

The amount of the sample taken for temperature fractionation varied between 50 to 57 mg. Corresponding Hg content is given in Table 3, which also provides information on Hg recovered in the KMnO₄ impingers.

Every sample of a gypsum size fraction was measured at least twice. Mercury recovered in KMnO₄ impingers

ranged between 90% and 120%.

The results of the gypsum samples separated by granulation are shown as thermograms of concentration in ng∙m⁻³ of released Hg in the carrier gas versus temperature (Figure 3).

Evidently, dry gypsum sample FF (Figure 3(a)) exhibits the most distinctive peaks with a maximum temperature of 160˚C, followed by a smaller peaks at 200˚C, 290˚C and 440˚C. The thermogram of CF does not show any peaks of Hg release (Figure 3(b)) due to low Hg concentration. The intensity of Hg peaks in fine fractions corresponds to the concentrations of Hg in gypsum samples.

A slight difference in the pattern is seen between those for dry and moist FF and FF1 samples (Figure 3(a) and Figure 3(c)). The first peak, at temperature of less than 200˚C, is lower in the moist sample than in the dry one. It seems that humidity affects Hg release from moist samples; it is less likely that the structural changes in

moist samples occur during the process of sample drying. In FF2 gypsum sample the Hg release pattern for moist and dry samples is similar (Figure 3(d)).

Dried fraction (FF) of gypsum was used in all subsequent experiments with spiking. The results are shown as thermograms (Figures 4-7) in order to facilitate and enable qualitative comparison of thermograms. In every figure four thermograms are shown where the black line represents a thermogram of the FF gypsum sample, the red one represents FF sample spiked with a particular Hg compound mixed with pure CaSO₄∙2H₂O, the blue line represents a thermogram of pure Hg compound mixed with substrate CaSO₄∙2H₂O, and the green represents a thermogram of pure Hg compound. Thermograms were smoothed to reduce the noise from the measuring equip- ment for easier comparison and clarity. The contents of total Hg in each sample are listed in Table 4, which also provides the information on Hg recovered in KMnO₄ impingers. Evidently, the recovery ranged from 70 to 100%. Losses encountered during heating may be related to evaporation loses of Hg(0), instability of some Hg compounds in air, adsorption on colder parts of the instrumentation, as well as analytical errors.

As the main intention of this work was to use thermal fractionation as a tool for identifying Hg compounds present in technical gypsum, the thermogram of the fine gypsum fraction (FF) is presented as a reference in all thermograms.

The thermograms are grouped into four subgroups including Hg halides, sulfates, oxides and HgS and HgSe for easier understanding and interpretation.

Mercury halides

Based on the presence of halogen elements (Cl, Br, F) in the WFGD system (Table 2) we selected Hg₂Cl_2, HgCl₂, HgBr₂ and HgF₂ as test compounds (Figure 4).

Thermograms for pure Hg chlorides and bromide are identical (Figures 4(a)-(c)), while that for pure HgF₂ differs significantly (Figure 4(d)), with several overlapping peaks between 200 and 370˚C and one strong peak at 440˚C [51] .

The reason for similar patterns of thermograms for Hg₂Cl₂ and HgCl₂ lies in Hg(0) and HgCl₂ having the same vapor pressure [16] . At higher temperatures, Hg₂Cl₂ follows a decomposition step, described in Equation (1) where, as a result, two peaks are released (Hg(0) and HgCl₂) [34] [52] [53] , emerging at the same temperature due to the same vapor pressure [51] . At the same temperature HgCl₂ compound also decomposes, following the decomposition step shown in Equation (2). Because of the similar chemical properties of Hg-Cl compounds, HgBr₂ decomposes at the same temperature.

Owing to the HgF₂ high boiling and melting points and the reactivity of HF (in the presence of residual moisture) released during thermal decomposition (Equation (3)) [54] , the resulting thermograms show very different patterns. The first peaks can be attributed to the decomposition of the HgF₂, while the second peak, at 440˚C, corresponds to HgO decomposition as described later (Figure 4(d)) [51] .

Consistent with the general observation, the thermograms of pure Hg halides and those mixed with CaSO₄∙2H₂O substrate differ considerably. The released peaks of Hg from Hg₂Cl₂, HgCl₂ and HgBr₂ mixed with CaSO₄∙2H₂O (Figures 4(a)-(c)), appearing at temperatures of around 670˚C, are difficult to explain since no theoretical background exists. We assume that this is caused by the experimental set up in which ionic Hg could have been deposited on the cooler surfaces of the inner wall of the tubing and finally released by high temperature gas flow towards the end of the temperature cycle [51] . This may be confirmed by the good mass balance observed during the experiments (Table 4).

The analyses of thermograms of technical gypsum (FF fraction) spiked with Hg halides show the two typical initial peaks and skewed declining trend up to 400˚C, the exception being a new clear peak at 340˚C for HgF₂. The relative intensities of the first two peaks in spiked and non-spiked FF gypsum are most similar for Hg₂Cl_2, and HgCl₂, less so for HgBr₂ and much less for HgF₂. The Hg halides can be concluded to be present in the gypsum samples with a high degree of certainty, the Hg chlorides predominating. This is in agreement with the relative concentrations of halogens in the WFGD solution [7] [34] [52] [53] .

Mercury sulfates

The thermograms shown in Figure 5 indicate some distinct differences in the decomposition patterns of Hg(I) and Hg(II) sulfates, which was not observed for Hg(I) and Hg(II) chlorides.

Pure Hg(I) sulfate (Figure 5(a)) shows a larger peak at 300˚C and a second at 520˚C, while Hg (II) sulfate (Figure 5(b)) exhibits only one major peak at 540˚C [51] . According to a report [55] this is a result of a multistep decomposition of Hg(I, II) sulfates. Decomposition of Hg₂SO₄ should follow the following decomposition steps as follows:

The presence of pure CaSO₄∙2H₂O causes a change in the appearance of the second peak for Hg(I) sulfate. A new peak appeared at 420˚C, while the second peak of pure substance Hg₂SO₄ almost disappeared. In the case of Hg(II) sulfate mixed with CaSO₄∙2H₂O a new peak at around 420˚C appeared beside a strong peak with a maximum at about 560˚C. These phenomena can be explained by the change in chemistry induced by the presence of crystalline water in CaSO₄∙2H₂O.

The thermograms of Hg(I) and Hg(II) sulfates mixed with technical FF gypsum differ significantly (Figure 5(a) and Figure 5(b)). In the case of Hg(I) sulfate, besides the two indicative peaks, two additional peaks ap-

peared at 260˚C and 340˚C but no peaks above 400˚C. This observation indicates that Hg(I) sulfate is not present in the FF gypsum sample. Hg(II) sulfate, however, shows a number of non-separated peaks in the temperature range from 320 to 500˚C and no peaks above 500˚C. A small peak of the FF gypsum sample (black line) appears within this temperature range (with a maximum at 450˚C) which indicates the possible presence of a small amount of Hg(II) sulfate in the original FF sample. This conclusion is further strengthened by the fact that only Hg(II) sulfate spiked to FF gypsum sample shows the peak within this temperature range.

Mercury oxides

Thermograms of red and yellow Hg oxides are shown in Figure 6(a) and Figure 6(b).

Similar thermograms are observed for pure Hg oxides with maximum temperatures of 460˚C and 420˚C for red and yellow HgO, respectively [51] . Both HgO compounds have an orthorhombic crystalline structure but they differ in particle size [56] . Yellow HgO (Figure 6(b)) consists of smaller particles which may affect the decomposition temperature. A multistep decomposition of HgO is suggested [57] , with the first decomposition step occurring at a temperature of 160˚C - 220˚C, the second at 220˚C - 400˚C and the third decomposition step at 400˚C - 500˚C. This is consistent with the results obtained in this study.

The presence of CaSO₄∙2H₂O caused the release of Hg at lower temperatures, with an additional, not well separated peak at about 300˚C. The effect of the matrix (substrate) on Hg release from the samples, especially from spiked gypsum samples, where Fe²⁺ and Mn²⁺ could play a role of catalysts and in the presence of Cl, could additionally enhance the reactions [7] . Thermograms of spiked FF gypsum exhibit the appearance of a maximum peak at about 350˚C and 300˚C for red and yellow HgO, respectively. This temperature range corresponds to the area of the not well separated skewed region of Hg release from technical FF gypsum sample. This is indicative of the presence of a small amount of HgO in FF gypsum.

Mercury sulfide and selenide

The thermograms presented in Figure 7 show that pure compounds exhibit patterns similar to those of substrates mixed with pure compounds [51] . Major peaks for HgS (Figure 7(a)) appear in the range between 290 and 310˚C, while for HgSe (Figure 7(b)) the maximum peaks appear between 250 and 290˚C [51] . The decomposition temperatures for the two compounds appear in a much narrower range (within 60˚C) than those for other Hg compounds, where Hg appears in much larger temperature ranges. Therefore, we can conclude that HgS and HgSe interact with the constituents of the substrate to a much lesser degree than other tested Hg compounds [51] .

According to published data and without considering the fugacity effect, the decomposition temperature of HgS should be 585˚C. Thus, when all factors are taken into account, a lower decomposition temperature is expected [58] . The single peak observed, is a consequence of a decomposition process of HgS consisting of only one step, which can also be termed a reactive sublimation (Equation (7)) [58] .

It has been assumed by other authors that HgSO₃ and HgO compounds could appear as decomposition product of HgS, which is probable when considering the gypsum produced in forced oxidation FGD. However, until now there is no evidence of such a decomposition process (Equations (8)-(10)) [32] also confirmed by our experiments.

The decomposition of pure HgSe follows just one step, according to the single released peak. It is interesting to note, that Hg and Se are among all elements far more enriched in fine fractions of the gypsum compared to lignite (Table 2). The enrichment factors are 110 and 88 for Se and Hg respectively. For other elements, these factors are in the range between 0.3 and 8.8. Both, Se and Hg are volatile, but highly concentrated in fine gypsum fraction, so we assumed the presence of a very stable HgSe in the gypsum. Based on the thermograms presented in Figure 7(b) this assumption could not be confirmed.

Both, HgS and HgSe, appear in the temperature range where the technical gypsum (FF sample, black line) shows a declining skew of Hg release with unclear peaks that cannot clearly be attributed to a specific compound tested.

More detailed analysis of all thermograms reveals that, in all cases except for HgS, those for FF gypsum spiked with different Hg compounds show two initial peaks at 170 ± 10 and 205˚C ± 5˚C. Although the oxidative environment of the WFGD system would prevent the formation of HgS in the scrubber, small amounts of H₂S present in the flue gas can produce trace amounts of highly insoluble HgS.