Determination of mercury thermospecies in South African coals in the enhancement of mercury removal by pre-combustion technologies

Samples of South African bituminous coals were analysed for total mercury (Hg) and Hg thermospecies concentrations using an RA-915 + Zeeman Mercury Analyser. Total mercury concentrations in samples of coals (n = 57) ranged between 10 ng g−1 and 493 ng g−1 with a mean value of 150 ± 53 ng g−1. Thermospecies of Hg were determined by monitoring Hg response as a function of sample temperature, increasing at 0.8 °C/s from ambient to 720 °C. This approach provides important information on thermal release of Hg species, as indicated by their appearance over specific temperature intervals. This permits identification of the presence of Hg thermospecies in coal and their quantification in each time (temperature) interval. It was found that 76% of tested bituminous coal samples release Hg species within low temperature intervals (20–180 °C and180–360 °C). The information generated in this study will aid in the selection of suitable coals for pre-combustion treatment that can lead to significant reduction of atmospheric Hg emission during coal combustion at power stations. This analytical approach can also be used for the creation of a system of coal classification based on the temperature of release of various Hg thermospecies.

Coal combustion during electricity generation is the main source of Hg emission to the South African atmosphere 15 . Exact data on the amount of Hg emitted was acquired only after several controversial publications. Thus, according to Pacyna et al. 16 , the amount of emitted Hg in South Africa in 2000 was 50 t. Recalculated data reported by Dabrowski et al. 17 showed a much smaller number (9.8 t) but data reported by Masekoameng et al. 18 was three-fold higher (29.47 t). Garnham and Langerman (2016) 19 subsequently showed that the amount of Hg from coal combustion in South Africa could range between 16 and 20 t/annum. Such differences may arise simply from uncertainty in the results of analytical determinations of the Hg content in coals.
In order to reduce Hg emissions from coal combustion, it was recommended that only coal having lower Hg content should be used 20 . Reduction during coal combustion is possible by preliminary cleaning, either by washing or by thermal treatment prior to combustion [21][22][23][24][25] . Coal washing removes the Hg associated with incombustible mineral materials such as pyrite (FeS 2 ) and thus reduces ash content to improve coal heating value. Some coals contain large amounts of Hg associated with their organic fraction and this can be a setback as it cannot be removed by coal washing. Therefore, controlled pyrolysis of coal provides a more promising alternative for mercury removal from all types of coals prior to combustion. However, the rate of mercury removal is affected by the treatment temperature, heating time, sweep gas flow and especially by the forms of mercury present. The efficiency of such thermal cleaning will be greatest in coal containing highly volatile Hg species. Mercury compounds that decompose at high temperature require much higher energy to release Hg species from coal. Such coal should be characterized in advance of its combustion to identify the most suitable pre-treatment temperature for removal of Hg 26,27 .
South Africa, as a signatory to the Minamata Convention is required (under Article 8), to reduce Hg emissions from coal-fired power plants during electricity generation 11,28 . The present study was initiated to develop analytical methods capable of identifying coals that contain Hg species having low volatilization temperatures such that they can undergo appropriate pre-combustion treatment. A new coal classification system could also be created based on the temperature of release of various Hg species. Such a classification system would be useful in identifying the effective pre-combustion and pre-gasification coal treatment method for reduction of Hg in coal prior to its combustion.

Experimental
Instrumentation. A RA-915 + Zeeman mercury analyzer (Lumex, St. Petersburg, Russia) was used for the quantification of various Hg species in coal samples. The working principle of the Zeeman mercury analyzer has been detailed by Sholupov et al. (2004) 29 and is based on the thermal desorption of Hg from solid coal samples followed by the detection of Hg atoms by atomic absorption. Background absorption is eliminated by the use of a high frequency Zeeman correction system 30 . The concentration of Hg in the sample is determined from a calibration curve plotted as integrated analytical signal (arbitrary units) versus absolute mass of Hg (ng). Realtime measurement of Hg during its thermal release from samples is accomplished within 60-100 s at a resolution of one second (1 s).
Samples. Samples of coals (n = 57) were collected from coal mines of Gauteng, Limpopo, Mpumalanga, Free State, KwaZulu-Natal and Eastern Cape Provinces of South Africa following a standard method for obtaining representative samples 31 . Coal production in South Africa is concentrated in the Highveld region of Mpumalanga Province where the Witbank, Highveld, and Ermelo coals are produced 32,33 . The mined coal is usually supplied to Eskom (Electricity Supply Commission) power stations for power generation and Sasol (abbreviated from Afrikaans Suid Afrikaanse Steenkool en Olie Maatskappy, literally translated as South African Coal and Oil Company in English) for use as feedback for the production of liquid fuel and chemicals and the surplus stock of coal is exported. The coal reserves in Witbank, Highveld and Waterberg (Limpopo Province) constitute approximately 70% of South Africa's recoverable coal reserves 34 . The Witbank and Highveld coals are laterally contiguous as coal types as different characteristics coexist in the same coal beds.
Analytical procedure for determination of total Hg. For the determination of total Hg, thermal decomposition of pulverized accurately weighted (to nearest mg) sub-samples was undertaken. The weighed sample was placed in a pre-cleaned quartz sampling boat and inserted into the furnace of the Hg analyser. The exact weight of the sample, ranging between 200 and 300 mg, was recorded using the RAPID software. Determination of total Hg was accomplished by heating samples until complete evaporation of Hg was ensured, usually within 60 to 100 s at 0.8 °C/s from ambient to 720 °C. Software permits temporal evolution of the analytical peak (Hg absorption) to be followed with the area under the peak, the maximum absorbance value and calculated concentration to be displayed. Each sample of coal was analysed in triplicate and results reported as a mean ± standard deviation of Hg concentration.

Results
Calibration of the mercury analyser. The mercury analyser was calibrated using certified reference materials (CRMs) containing Hg content in the range covered in the coal samples. Standard SARM 20 (MINTEK, South Africa, 250 ± 30 ng g −1 ) and PACS-2 (National Council of Canada, Canada, 3040 ± 99 ng g -1 ) were used for instrument calibration while CRM 7002 (Light sandy soil; Czech Republic, 90 ± 5 ng g -1 ) and CRM 024-050 (Loam soil 1, 710 ± 50 ng g -1 ) were used for method validation.. A calibration curve, plotted as integrated absorbance (peak area, arbitrary units) versus absolute mass of Hg (ng) defined by a typical regression equation such as y = 350x − 71 was obtained and used for the quantification of Hg. The calibration curve was linear (R 2 = 0.998) up to 800 ng of Hg. This absolute value of 800 ng Hg indicates that the relative Hg concentration that can be determined in coal with acceptable precision and accuracy is 3200 ng g −1 for a sample mass of 250 mg. Therefore,

Scientific Reports
| (2020) 10:19282 | https://doi.org/10.1038/s41598-020-76453-z www.nature.com/scientificreports/ this analytical approach based on thermal decomposition to determine the Hg content in coal is more favourable than chemical decomposition methods as it is more rapid and offers high efficiency. In general, results for determination of Hg in the CRMs were in good agreement with certified concentration values at 95% level of confidence as the found values in SARM 20, PACS-2, CRM 7002 and CRM024-050 were 248 ± 7 ng g -1 , 2990 ± 110 ng g -1 , 94 ± 8 ng g -1 and 718 ± 14 ng g −1 , respectively. To check the reproducibility of the analytical method, replicate samples of coal were analysed as shown in Fig. 1. The long-term results of Hg measurements (n = 33) were within the confidence intervals of certified values at the 95% confidence level.
Limit of detection and limit of quantification. Due to a lack of "blank" coal samples on the market, the determination of the limits of detection (LOD) and quantification (LOQ) were calculated from the regression line of the calibration curve presented in a general form: y = a + bx . Numerical calculations were performed using the following formulas: LOD = a + 3S a b and LOQ = a + 10S a b , where S a is the standard deviation of the response y and b is the slope of calibration curve 35 . The LOD and LOQ were found to be 0.3 ng and 1.0 ng, respectively. These values indicate that for a coal sample of 250 mg the relative LOD is 1.20 ng g −1 and the LOQ is 4.0 ng g −1 , illustrating that the methodology is capable of determining Hg content in coals of any commercial use.
Influence of coal particles size on the results of Hg determination. The first factor to be evaluated during method development was the influence of particle size on the accuracy and precision of the results. This was achieved by analysing coal samples with particle sizes in the range 1-3 mm (1000-3000 µm) and pulverized coal with particle sizes of less than 50 µm. It was found that the results of the determination mercury in coals having a uniform distribution of Hg were not influenced by the particle size, as evident in Table 1.
This conclusion follows from a comparison of values of the relative standard deviation (RSD), which in the case of large particle size, RSD ranged between 6.4% and 42.8%, while for samples with small particle size, it was 1%-5.9%. Coal samples with inhomogeneous distribution of Hg can be analysed with good precision (RSD = 6%) only if particle size is first reduced. For samples having particle size 1000-3000 µm or higher, an RSD of as high as 42.8% could be reached. Results of this study suggest that coal samples should be ground to a particle size of 50 µm or lower to achieve higher accuracy and precision for coal with any kind of Hg distribution.   www.nature.com/scientificreports/ 251 ng g −1 , 254 ng g −1 and 245 ng g −1 , respectively, with a mean Hg concentration of 250 ± 5 ng g −1 irrespective of rising sample masses. Similarly, quantitative results of Hg determination in four coal samples analysed in assessment of coal particle confirmed that sample mass of 50 mg to 300 mg has no influence on either the reproducibility of Hg concentration as shown by relative standard deviations that ranged between 1.0 and 4.2% (Table 2). Therefore, it can be concluded that the results for determination of total Hg in coal are independent of sample mass taken for analysis.

Total Hg concentration in coals.
Results for determination of total Hg in individual coals samples from South African Provinces, viz., Gauteng, Limpopo, Mpumalanga and KwaZulu-Natal and Free State, show high variability in concentration and range from 10 to 493 ng g −1 as shown in Table 3.
The mean concentration varied from 68 to 230 ng g −1 . Thus, the average Hg concentration in Mpumalanga's coal fields of Highveld and Witbank were found to be 158 ± 22 ng g −1 and 115 ± 60 ng g −1 , respectively. The mean Hg content in coal from Highveld deposit is in agreement with the average of 150 ± 50 ng g −1 reported by Wagner and Hlatshwayo 36 . The mean values of Hg content in coals Gauteng coal mines were found to be 180 ± 150 ng g −1 and compare well with the average of 200 ng g −1 found in Gauteng coals as reported by Bergh 37 . The lowest mean Hg content of 68 ng g −1 was determined in Eastern Cape coals and the highest mean Hg content of 230 ng g −1 was found in Free State Province. An abnormally high Hg concentration of 2514 ng g −1 was measured in one coal sample from Mpumalanga Province and is statistically different from all other results, but matches the result of 2430 ng g −1 measured in coal from Vryheid formation 38 . The mean value of Hg concentration in coal samples from all South African Provinces was found to be 150 ± 53 ng g −1 .  www.nature.com/scientificreports/ Total Hg in density fractionated coal samples. Density fractionated coal samples were analysed in order to identify the fraction of coal with reference to its specific gravity that has the highest Hg content. Two batches of selected coal samples from Witbank and Gauteng mines were used for study. Fractions of coals, separated by densities ranging between 1.40 g cm −3 and 2.00 g cm −3 , were analysed for the Hg concentration and results summarized in Table 4. For Witbank coals, the batch consisted of coal float samples, viz., WC-f1.40 to WC-fl2.0 and WC-Sink. It was found that low density float fractions contain more Hg (371 ± 24 ng g −1 , 561 ± 20 ng g −1 and 817 ± 12 ng g −1 ) with total average of 583 ± 224 ng g −1 bound to organic fractions. Coal sink sample, WC-Sink, from this batch was found to contain only 79 ng g −1 . For Gauteng coals, an average Hg content in coal float samples of low densities bound to organic compounds were found to be 34 ng g −1 while the mineral fraction of sink sample presented 1340 ng g −1 . Such material was dominated by kaolinite minerals. These results show that the contribution of Hg bound to the organic fraction is significant compared to the total Hg content 39 .
Results for the determination of thermospecies of Hg in coal.. The procedure for determination of thermospecies of Hg developed by Mashyanov et al., 2007 40 demonstrated good reproducibility of results at over selected temperature intervals, as noted in Fig. 3. After analysis of all samples for the presence of Hg peaks, it was established that some coal samples may contain only one Hg species, as indicated by the presence of only a single peak over the full temperature range investigated, whereas other samples may contain two to four peaks that are associated with different chemical forms of Hg in the coal specimen (cf. Fig. 3). The total concentration (ng g −1 ) and percentage (%) of various thermospecies of Hg in selected samples from each of the South African provinces are shown in Table 5. These include coal samples with the smallest, close to the mean, and highest concentrations of Hg. It was found that out of twenty one (21) analysed bituminous coal samples (excluding pyrite coals and SARM 20), sixteen (16) coal samples or 76% coal samples contained thermospecies of Hg that are released

Discussion
Results of this investigation indicate that correct measurement of total Hg concentration in coal samples after thermal evaporation is constrained by achieving a number of conditions. One of them is the nature of the distribution of Hg in the coal which influence the reproducibility of determination of total Hg. In the case of a homogeneous distribution, the particle size is not an important factor, but if this is not satisfied and Hg is present in the form of minerals such as cinnabar (HgS) or other compounds high in Hg concentration, the RSD of results can be very high, leading to poor precision. Thus, analysis of coal samples having an uneven distribution of Hg present in particle sizes ranging from 1000 to 3000 µm could result in RSDs as high as 40%, whereas when using pulverized coal having particle sizes below 50 µm yields only 2% (cf Table 1). The extremely uneven distribution of Hg in samples from Vaal coal mine (northern Free State) could be connected with the geologic structure of the seam, wherein a combination of chemical and physical weathering has resulted in a highly undulating floor 33 . To achieve higher accuracy and precision, it is recommended that all coal samples be ground before analysis. Another possible factor affecting reproducibility and accuracy of results during Hg determination in coal could be associated with sample mass taken for analysis. This study revealed that the Hg concentration in coal standard SARM 20 and coal G-1 derived from sample masses ranging from 50 to 300 mg were identical at the 95% level of confidence ( Table 2). The mean total Hg in SARM 20 was determined to be 250 ± 5 ng g −1 , in agreement with the certified value of 250 ± 10 ng g −1 . The total concentration of Hg in coal G-1 was 432 ± 5 ng g −1 , independent of sample mass in the range 50-300 mg, yielding an RSD of 1.0%. These data show that results for determination of Hg in coals are not influenced by mass of sample taken for analysis.
The last factor which can influence the results is the stability of Hg as regards evaporation from the samples. Stability of thermos peaks of Hg could be identified by the shape of resultant absorbance signals recorded during analysis (cf Fig. 1). Various shapes were obtained are shown in Fig. 3. They reflect the presence of different modes of occurrence of Hg in coals. Detailed studies of Hg thermo peaks indicate that coal contains various forms of Hg that are released over specific temperature ranges, viz., 20-180 ºC, 180-360 ºC, 360-540 ºC, 540-700 ºC, generating very stable and reproducible results, as summarized in Table 5. The mean Hg concentration in coals of South African Provinces was found to be 150 ± 53 ng g −1 while the global average Hg content of coal, expressed on a whole-coal basis (the Clarke value for Hg in coal) is 100 ± 10 ng g −1 and is the same for bituminous, subbituminous, and lignite rank coals. This result confirms previous findings that concluded that the mean value of Hg in South African coals is higher than the global average 42 .
Analysis of density fractionated coal samples showed that such samples can be used to identify the coal fraction that concentrates the Hg (cf Table 4). In general, it was found that in some coal samples, Hg accumulates in organic compounds in low density coal fractions whereas in other samples of coal Hg is concentrated in the mineral sink and the highest content found was 1340 ± 35 ng g −1 in contrast to mineral bound coal fractionated sink sample of batch 1 (79 ± 2 ng g −1 ). The results of pyrolysis of the density-fractionated samples lead to the conclusion that the method of thermal speciation is more efficient for the selection of coal for Hg removal by mild pyrolysis before coal combustion. Thermospeciation of Hg in coal prevents direct application of pyrolysis technology to raw unsuitable coal with a high content of pyrite-bound Hg because the high pyrolysis temperature demands high energy consumption, thereby degrading overall coal heating efficiency.
In spite of the fact that most of South African coals are bituminous grade, 2% being anthracite grade and 1.6% being coal of metallurgical quality, several thermospecies of Hg were detected in most coal samples due to their different binding energy to coal matrix.
Coal may contain one or several Hg species that are released at variable temperatures (cf Table 5). The majority of such species were released in the range 20-180 ºC and can be attributed to elemental Hg as it is known to be vaporized in this range of temperature. The second dominant thermospecies, released between 180ºC and 360ºC, could be connected to the presence of organobound-Hg compounds, whereas the least amount of Hg is released between 360 and 540 ºC and may be connected with a pyrite bound Hg fraction 46 . Identification of the volatility temperature range of Hg in coal in advance of use is a necessary step required to select an appropriate method for removal of Hg during coal combustion for electricity generation.

Conclusions
This study shows that the accuracy and precision of determination of total Hg in coals having an inhomogeneous distribution of Hg is dependent on the particle size of the samples taken for analysis. Particle sizes below 50 µm are recommended for analysis to ensure accurate and precise results. It was also found that the results of analysis are independent of the mass of the samples in the range of 50-300 mg. Results for determination of total Hg show that the mean value of Hg concentration in coal samples from all South African Provinces is on the level of 150 ± 53 ng g −1 . The presence of Hg species of various thermal stability in coal was demonstrated through detection of multiple desorption peaks, dominated by low temperature evaporation of elemental Hg released at 20-180 °C. The specific temperature range of Hg peaks creates a basis for selection of coals for preliminary cleaning prior combustion at electrical power stations according to the thermal stability of the Hg species. Identification of thermospecies is an analytical problem that requires further studies.
Analysis of density fractionated coal samples showed that Hg can be concentrated in both organic components of coals, as well as in mineral fractions. This study presents new insights into our knowledge of the forms of Hg present in South African coals.