Pyrite–Coal Depressants Interactions During Coal Reverse Flotation

Abstract

This work investigates coal desulfurization by using reverse flotation. In this method, pyrite (the only source of sulfur in the studied coal) was separated from the meta-bituminous coal by using three different coal depressants (starch, dextrin and humic acid). A novel variable elimination approach was used to determine the contribution of the depressant type and the depressant concentration on the desulfurization performance. The results showed that the pyrite recoveries are influenced by the depressant type while the highest pyrite recovery was achieved in the presence of humic acid. Therefore, humic acid should be used in flotation rougher and scavenger cells in which the aim is to achieve high pyrite recovery. By contrast, the pyrite grades are affected significantly by the depressant concentration. Considering that the aim in flotation cleaner cells is to achieve high pyrite grade, any of the studied coal depressants can be successfully used but at high concentrations. This work demonstrated that the selection of flotation depressants depends on the type of flotation cells used in coal desulfurization.

1. Introduction

Sulfur is a detrimental impurity when present in coal and is a typical problem in coal deposits in many parts of the world [1,2]. It can be present in coals in inorganic, organic or sulphatic forms [3]. Sulfide minerals such as pyrite contribute significantly to the inorganic Sulfur in many coals. The removal of pyritic minerals from coal leads to increased economic benefits. In addition, the reduction in sulfur content in combustible coal reduces SO₂ emission into the environment, and the subsequent acid rain [4]. Sulfuric acid is produced when pyrite present in tailings is oxidized in the presence of oxygen and water, leading to acid mine drainage which is considered as the main cause for the release of toxic elements from tailings into the environment [5]. Therefore, the desulfurization of coal prior to its combustion has significant economic and environmental implications, as well as the sustainability of coal mining operations.

Reverse flotation is commonly used to reduce either the Sulfur or ash content of coals and other minerals [6,7]. In this process, the gangue minerals containing Sulfur or ash are floated while the flotation of coal is depressed. The gangue minerals that lead to ash in coal have significantly different flotation properties to those of pyritic minerals (i.e., sulfur-containing minerals). Thus, reverse flotation for ash reduction and Sulfur removal could not be performed in a single process. A number of studies have been conducted to investigate the performance of reverse flotation to lower the ash content of coal [8,9,10,11]. However, studies conducted on pyritic minerals flotation for Sulfur reduction in coals are very few [12,13].

Since coal is naturally floatable, depressants play a critical role in rendering the coal hydrophilic and thus enhancing the selectivity of the coal’s reverse flotation process. The most commonly used coal depressants are either organic colloids such as dextrin, starch, quaternary amines, tannic acid or humic acid [10,12,14]. Dextrin is derived from starch through heat degradation and thus is a smaller but highly branched molecule as compared with starch molecules which are larger and consist of a water-soluble linear (amylose) and water-insoluble branched (amylopectin) components [15,16]. Coal flotation depression by starch and dextrin occurs through hydrophobic bonding between the non-polar groups of starch/dextrin and the coal surface. Humic acids are macromolecular, negatively charged polyelectrolytes that are composed of carboxylic and phenolic functional groups, which adsorb on coal surfaces and reduce the number of active sites on minerals that can be adsorbed by collectors [17].

Although these depressants were proven to be effective in depressing coal flotation, it is important to note that the gangue floated in these studies were silicates and carbonates (e.g., calcite, dolomite, and silica) with the aim of reducing ash content in coal [18]. Thus, it is important to investigate whether these coal depressants affect the flotation of gangue pyritic minerals from coal. For example, some depressants such as dextrin can also depress pyrite at high concentrations [19,20]. Dextrin was found to be an effective depressant of pyrite above pH 4 even in the presence of an xanthate collector [19]. However, the influence of different depressants that cannot cause harm to the environment (i.e., green chemical compounds) such as dextrin, starch and humic acid on the flotation performance of pyrite from meta-bituminous coal has not been studied. More importantly, the impact of the type of depressant and their concentration have also not been systematically evaluated. It should be noted that considering that dextrin is a derivative of starch, it is possible that starch may possess some pyrite depression behavior.

In the present study, the efficiency of reverse flotation in the separation of pyrite from meta-bituminous coal in the absence and presence of three different depressants (starch, dextrin and humic acid (HA)) was investigated. The flotation kinetics of pyrite from meta-bituminous coal with different depressants at various concentrations were compared. The flotation performance in a microflotation cell was characterized by pyrite recovery and the pyrite grade of the concentrates. The variable elimination approach method, recently developed by Albijanic et al. [21], was used to quantify the influence and systematically analyze both the depressant type and its concentration on the desulfurization of the meta-bituminous coal to suggest the best depressant for different stages of a flotation circuit (i.e., rougher, scavenger and cleaner stage).

2. Materials and Methods

2.1. Materials

The pyrite samples utilized in this study were sourced from Navajun, La Rioja, Spain with a purity >95%. The total Sulfur content in the samples was analyzed by XRD and was 52.4% [22]. The meta-bituminous coal samples were collected from the Shanxi province in China [22]. Pyrite was first dry ground with a ceramic mortar and pestle and then screened to obtain a −212 + 150 μm size fraction. The coal sample was crushed and screened to obtain a −0.5 + 0.25 mm size fraction. The pyrite sample was ground to a lower size distribution than that of the coal sample since pyrite particles are embedded in the raw coal and the size distribution of pyrite is much lower than the coal particles [23]. The moisture content (M_ad), the volatile content (V_ad), the fixed carbon content (FC_ad), the ash content (A_ad) and the total Sulfur content (S_t) in the coal are shown in

Table 1. Analysis of air-dried meta-bituminous coal [24].

Mad (%)	Aad (%)	Vad (%)	FCad (%)	St (%)
1.5	26.9	18.3	53.3	0.39

. A detailed mineralogy of the coal used in this study has been published elsewhere [24]. The main minerals in the meta-bituminous coal sample were kaolinite (39.3%), quartz (29.1%), montmorillonite (6.4%) and calcite (4.6%). The total Sulfur content was negligible in the coal samples.

Potassium amyl xanthate PAX 99% (Chem-Supply, Gillman, Australia) and copper sulfate 99% (Chem-Supply, Australia) were used as pyrite collectors and activators, respectively. Methyl isobutyl carbinol MIBC 99% (Orica, Melbourne, Australia) was used as a frother. Soluble potato starch 99% (Sigma-Aldrich, St. Louis, MI, USA), potato dextrin 99% (Chem-Supply, Australia) and humic acid 99% (Sigma-Aldrich, USA) were used as depressants. Starch and dextrin solutions were made by preparing a paste of 1 g of starch or dextrin with cold water and then adding the paste to 100 mL of boiling water. Double distilled water 18.2 MΩ·cm was used.

2.2. Microflotation Tests

The separation of pyrite from meta-bituminous coal was analyzed through micro-flotation tests. In each flotation test, 1 g of pure pyrite particles were fully mixed with 1 g of coal particles. The 50% pyrite–coal mixture was used to enhance the analysis sensitivity to detect changes in the pyrite removal and to help understand the mechanisms of pyrite depressant interactions. Pyrite and coal were mixed in the flotation cell at a 1% solid concentration, i.e., 1 g of pyrite-coal mixture and 99 g of water. Before the micro-flotation tests, samples were deslimed in an ultrasonic cleaner (Derui, China) to reduce the effect of fine particle entrainments at 40 kHz for 1 min. The deslimed samples were conditioned for 5 min using a magnetic stirring plate with potassium amyl xanthate collector (1 × 10⁻⁴ mol/L), starch, dextrin or humic acid depressant (1 kg/t, 2 kg/t, 3 kg/t or 4 kg/t), CuSO₄ activator (20 g/t) and MIBC frother (2 kg/t). The conditioned pulp was adjusted to pH 6 and was then transferred into a 300 mL flotation column (FLSmidth, Australia), which was used to conduct the micro-flotation experiments [22]. The air flow rate was 0.5 L/min and the revolutions used were 700 rpm. The flotation performance was characterized by the recovery and grade of the pyrite determined by Sulfur analysis using a LECO SC632 Sulfur analyzer. Each experiment was conducted in triplicate and the experimental error was found to be less than 5%. The flotation experiments were conducted until no froth was produced (ca. 4 min). The flotation concentrates were collected after 1, 2 and 4 min.

2.3. Data Analysis—A Variable Elimination Method

The effect of the depressant type and its concentration was analyzed through a variable elimination approach [22]. A regression model was proposed, as in Equation (1)

$$y=f\left(x,\beta \right)$$

(1)

where $\beta =\left[{\beta }_1,{\beta }_2,\cdots ,{\beta }_m \right]$ and $x=\left[x_1,x_2,\cdots ,x_n \right]$ are the vectors showing regression coefficients and input variables, respectively; while $y$ is a response variable. The following two hypotheses are tested to determine the significance of ${\beta }_i$.

(1) $H_0\left({\beta }_i=0\right)$ ${\beta }_i$ is not significant,

(2) $H_1\left({\beta }_i\ne 0\right)$ ${\beta }_i$ is significant.

An elimination of ${\beta }_i$ from the model shows the significance of ${\beta }_i$ on the calculated $y$. In other words, there are two cases:

Case I: All coefficients are included in the model (see Equation (1)).

Case II: One coefficient is excluded from the model, i.e., one of the coefficients is zero $\left({\beta }_i=0\right)$, while other coefficients are unchanged (see Equation (1)).

Therefore, the coefficient ${\beta }_i$ is significant if the elimination of this coefficient (i.e., ${\beta }_i=0$) from the model has a significant influence on the calculated $y$ and thus the hypothesis $H_1$ is valid. As a result, there is a significant difference between the quality of curve fitting of Case I and Case II.

The quality of curve fitting is determined by calculating and average relative error $\left(\delta \right)$, which is for Case I:

$$\delta ={\displaystyle \frac{1}{n}}{\displaystyle \sum }_{j=1}^n\left|{\displaystyle \frac{y_e^j-y_c^j}{y_e^j}}\right|$$

(2)

where $n$ represents the number of experimental data; $y_e^j \mathrm{and} y_c^j$ are the $j^{th}$ experimental and calculated values, respectively. The calculated average relative error $\delta \left({\beta }_i=0\right)$ for Case II is:

$$\delta \left({\beta }_i=0\right)={\displaystyle \frac{1}{n}}{\displaystyle \sum }_{j=1}^n\left|{\displaystyle \frac{y_e^j-{\left(y_c^j\right)}_{{\beta }_i}}{y_e^j}}\right|$$

(3)

where ${\left(y_c^j\right)}_{{\beta }_i}$ is the $j^{th}$ calculated values for $y$, when the regression coefficient ${\beta }_i$ is excluded from the model and other regression coefficients are unchanged. A higher average relative error indicates a significant influence from the input variable x_i associated with that regression coefficient β_i.

3. Results and Discussion

3.1. Desulfurization Process

The desulfurization process of the meta-bituminous coal was studied by determining the removal of pyrite through flotation in the presence of three depressants (starch, dextrin and humic acid (HA)); the only source of Sulfur in the meta-bituminous coal is pyrite. Figure 1 shows the pyrite grade after flotation with starch, dextrin and humic acid. It was shown that the pyrite grade increased with the increase in concentration for all depressants, which suggests that each of them was effective in depressing the floatability of coal particles. This may be due to all depressants adsorbing into pyrite through –COOH and –OH interactions with the ferric hydroxide groups of pyrite [25]. Humic acid showed the lowest pyrite grade at concentrations below 3 kg/t, while starch and dextrin showed an almost constant high pyrite grade (>90%) regardless of concentration. This could probably be due to humic acid being more hydrophobic when compared to starch and dextrin, and therefore decreasing humic acid–pyrite adsorption at low concentrations [26].

The classical first-order kinetic model [27] was used to study pyrite recovery ε, see Equation (4).

$$\epsilon ={\epsilon }_{\infty }\left[1-exp\left(-kt\right)\right]$$

(4)

where ε_∞ is the ultimate recovery in %, t is the flotation time in min, and k is the flotation rate constant in 1/min. The fitting of Equation (4) was carried out by using the least square method and the Solver function in Microsoft Excel.

Table 2. Pyrite flotation performance results in the presence of different depressants at 10⁻⁴ mol/L PAX solution.

Depressant Types	Concentration (kg/t)	ε∞	k (1/min)	R2	Coal Recovery
Starch	0	99.38	8.38	1.00	78.69
	1	96.09	5.50	0.99	13.45
	2	95.11	5.44	0.99	7.13
	3	94.96	5.24	0.99	3.29
	4	67.81	4.80	0.99	3.18
Dextrin	0	99.38	8.38	1.00	78.69
	1	97.58	6.14	0.99	4.26
	2	96.25	5.85	0.99	3.32
	3	95.86	5.78	0.99	2.12
	4	76.15	4.12	0.99	1.34
HA	0	99.38	8.38	1.00	78.69
	1	99.27	7.12	0.99	24.24
	2	98.62	7.03	0.99	19.79
	3	97.48	6.98	0.99	5.01
	4	96.18	6.86	0.99	1.26

shows the results for the ultimate recoveries, the flotation rate constants and the regression coefficient for different concentrations of the three depressants. As seen in Table 2, the classical first-order kinetic model successfully predicted the flotation performance of separation of pyrite from meta-bituminous coal with various types of depressants (R² was at least 0.99). The ultimate recoveries and the flotation rate constants decreased with the increase in the depressant concentration. However, the use of humic acid resulted in the highest ultimate recovery independent of concentration (>95%). This might be due to the higher hydrophobic properties of humic acid benzene rings than those of the polysaccharides of starch and dextrin. At higher concentrations starch and dextrin may increase the hydrophilicity of pyrite particles while humic acid keeps the hydrophobicity of pyrite particles.

3.2. Variable Elimination Approach

To better understand the desulfurization process, the variable elimination approach was used. In this method, the correlation was proposed for predicting the flotation recoveries, the pyrite grades and the flotation rate constants:

$$Y={\beta }_1{\left(1+\mathrm{T}\right)}^{{\beta }_2}{\left(1+{\mathrm{C}}_{DEP}\right)}^{{\beta }_3}$$

(5)

where β₁, β₂ and β₃ are the regression coefficients in the proposed correlation, Y is the corresponding predicted variable (the flotation recovery, the flotation rate and the pyrite grade), and T is the type of the depressant, i.e., for starch (T = 1), for dextrin (T = 2) and for HA (T = 3); C_DEP (kg/t) is the depressant concentration. The regression coefficients (β₁, β₂, and β₃) were obtained by using the LAB fit software V 7.2.50c [28] in which the Levenberg–Marquardt algorithm was used.

Table 3. Regression coefficients and model validation.

Predicted Variables	β1	β2	β3	δ
Y	(t Value)	(t Value)	(t Value)	(%)
ε∞	90.90 ± 8.14	0.11 ± 0.07	−0.09 ± 0.04	6.21
	(11.17)	(1.44) *	(−2.38)
k	6.00 ± 0.72	0.28 ± 0.10	−0.26 ± 0.05	7.97
	(8.34)	(2.81)	(−5.61)
Pyrite grade	73.66 ± 5.99	−0.08 ± 0.07	0.25 ± 0.04	5.40
	(12.30)	(−1.30) *	(6.86)

* Insignificant coefficients (95% confidence).

shows the calculated regression coefficients, their standard errors and t-values for each coefficient. The experimental and predicted variables using Equation (5) are shown in Figure 2.

A variable elimination approach, described in Section 2.3, was used to quantify the contribution of the depressant type and its concentration on the desulfurization process (the flotation recovery, the flotation rate constant, and the pyrite grade). As seen in Equation (4), the influence of the depressant type on the desulfurization process determines β₂ while the influence of the depressant concentration on the desulfurization process reveals β₃. The coefficient β₁ does not represent the effect of either the depressant type or the depressant concentration and thus β₁ is not used in this analysis i.e., β₁ is kept constant. More precisely, for example in the case of the pyrite grades, two cases were considered:

(1) β₁ = 73.66, β₂ = 0 and β₃ = 0.25, i.e., the elimination of β₂ from the model (see Table 3 and Equation (4)) determines whether this coefficient affects significantly the pyrite grade prediction.

(2) β₁ = 73.66, β₂ = −0.08 and β₃ = 0, i.e., the elimination of β₃ from the model (see Table 3 and Equation (4)) determines whether this coefficient affects significantly the pyrite grade prediction.

Figure 3 shows the parity plot diagrams and calculated δ(β_i = 0) for the ultimate recovery, the flotation rate constant and the pyrite grade. Considering that δ(β₁ = 0) is always zero, it is not shown in Figure 3. As seen in Figure 3, in the case of the pyrite grade and the flotation rate constant, δ(β₃ = 0) is higher than δ(β₂ = 0) which indicates that the depressant concentration is a more significant variable than the type of depressant. The opposite is true in the case of the ultimate recovery since δ(β₃ = 0) is lower than δ(β₂ = 0). The effect of the type of depressant on pyrite recovery may be due to the differences in the hydrophobicity of depressant, while starch and dextrin contain many hydrophilic functional groups, humic acid contains benzene rings increasing its hydrophobicity. In the case of the effect of depressant concentrations on pyrite grade, it may be due to an optimum concentration, while insufficient depressant might not fully inhibit coal, leading to a lower grade, while excessive depressant can also affect the flotation process, reducing overall efficiency.

Therefore, it is recommended to use humic acid rather than starch and dextrin in rougher and scavenger coal flotation circuits, as recovery is highly affected by the depressant type, and the aim in these circuits is to achieve the highest possible recovery. As regards the pyrite grade, the depressant concentration influences more pyrite grade rather than the depressant type (δ(β₃ = 0) > δ(β₂ = 0)). It means that considering that the aim in coal flotation cleaner cells is to achieve high pyrite grade and not high pyrite recovery, any of the studied coal depressants can be successfully used but at high concentrations.

4. Conclusions

In this paper, a variable elimination approach was used to quantify the contribution of the depressant type (humic acid, starch and dextrin) and the depressant concentration on the desulfurization of meta-bituminous coal. It was found that the pyrite recoveries are significantly more influenced by the depressant type rather than the depressant concentration and thus it is recommended to select humic acid to achieve the highest pyrite recovery. In contrast, the pyrite grades are affected significantly by the depressant concentration and thus any of the studied coal depressants can be successfully used but at high concentrations. The corresponding pyrite recovery and grade using humic acid as a depressant are both over 96% at higher concentrations (4 kg/t). When the depressant concentration is more important than the depressant type, the least expensive depressant should be selected. This paper demonstrated that the variable elimination approach is a reliable method when selecting the reagents for the desulfurization of the meta-bituminous coal, which indicated that the type of depressant is a more significant variable than the depressant concentration and it is recommended to use humic acid rather than starch and dextrin to achieve high mineral recovery. This is very important considering that maximizing flotation plant efficiency and reducing operating costs are essential during the coal desulfurization process. A significant amount of further work concerning the flotation kinetics and corresponding restriction factors is required to investigate the desulfurization process for samples with different mixtures of coal and pyrite.