3.1 ATR-FTIR Spectra Analysis
ATR-FTIR spectra for all coal samples verified the PCA analysis. Bands that are present between 1100-600 cm-1 correspond to mineral matter such as illite, montmonrollite, quartz and aluminosilicates. The band specifically around 1030 cm-1 shows the presence of Si-O bonds (ash) in the system. The raw coal and residue clearly show the presence of ash, whereas the intensity of the band around 1030 cm-1 diminishes in the SCCs (Balachandran 2014). The aromatic absorption bands at 750 and 815 cm-1 in the raw coals and the SCCs are due to out-of-plane vibrations of one isolated and two adjacent C-H aromatic groups, respectively (Table I). The spectra of the SCCs showed one such aromatic band at 750 cm-1. Absorption bands at 1600 cm-1 in SCCs are attributed to the presence of C=C stretching accentuated due to H-bond containing functional groups (Table I) (Cooke et al. 1986). These aromatic bands were found to be more intense in SCCs showing effective action of the e,N solvent in coal cleaning (Sun et al. 2011).Pandra coal shows an increase in aromatics, especially triaromatics while, Bhagabandh shows increase in diaromatics and triaromatics, which is also evident by the increase in the Har/Haliratio.
3.2 Py-GCMS and PCA of OCs and SCCs
3.2.1 Py-GCMS
Py-GCMS studies were performed for the five coals and their solvent extracted SCCs. The basic characterization of these coals (Elemental, proximate analysis, TG curve) is already reported by the authors (experimentalists in this group, Dhawan et al.) in a previous published work (Sharma et al. 2019).
Table I Structural parameters deduced from the FTIR peak area measurements
Coal
|
Har/Hali
|
Har(wt%)(2700-3000 cm-1)
|
Hali(wt%)(700-900cm-1)
|
Car(wt%)(1600-1800 cm-1)
|
Cali(wt%)(1375-1450cm-1, 2850-2990 cm-1)
|
Car/Cal
|
-C=O (wt%)(2695-2830 cm-1)
|
CH3(deformation)(1300-1400 cm-1)
|
Har/Car
|
Bhagabandh OC
|
1.34
|
8.03
|
5.99
|
6.23
|
6.20
|
1.00
|
4.32
|
2.95
|
1.28
|
Bhagabandh SCC
|
1.55
|
9.29
|
5.98
|
6.82
|
5.73
|
1.19
|
3.78
|
2.55
|
1.36
|
Bahula OC
|
1.35
|
7.95
|
5.87
|
5.71
|
6.21
|
0.92
|
4.09
|
2.98
|
1.39
|
Bahula SCC
|
1.59
|
9.23
|
5.77
|
5.93
|
5.98
|
0.99
|
3.93
|
2.63
|
1.55
|
Moonidih OC
|
1.46
|
8.45
|
5.76
|
6.12
|
5.97
|
1.02
|
3.84
|
3.65
|
1.38
|
Moonidih SCC
|
1.66
|
9.71
|
5.83
|
5.75
|
6.21
|
0.93
|
3.90
|
3.07
|
1.68
|
Pandra OC
|
1.43
|
9.46
|
6.59
|
6.90
|
7.72
|
0.89
|
4.66
|
14.34
|
1.37
|
Pandra SCC
|
1.58
|
9.50
|
5.99
|
5.79
|
7.28
|
0.79
|
3.84
|
3.00
|
1.64
|
NL OC
|
1.53
|
8.76
|
5.74
|
5.85
|
5.93
|
0.98
|
3.54
|
3.00
|
1.50
|
NL SCC
|
1.47
|
8.82
|
5.97
|
6.30
|
7.01
|
0.89
|
3.74
|
2.95
|
1.4
|
Each sample gave rise to more than 50 identifiable chemical compounds during pyrolysis. The compounds were categorized according to triaromatics, diaromatics, monoaromatics, cycloalkanes, n-alkanes, alkenes and other structures based on matches with the NIST mass spectral library. The main aliphatic compounds present in the coal tar (i.e., pyrolysis vapour) were paraffin hydrocarbons, while the main aromatic components were monoaromatics such as benzene, toluene, o-xylene, phenols and substituted phenols. Polycyclic aromatic hydrocarbons (PAHs) that were identified included naphthalene, substituted naphthalenes, phenanthrene, substituted phenanthrenes, anthracene, pyrene, etc, with naphthalene and its derivatives constituting the main constituents. The coal tar also contained some oxygen compounds, such as acids, and small amounts of aromatics containing nitrogen (Makan et al. 2017).
The mass spectra of the whole sample (OCs and SCCs) pyrolysis vapors show a complex series of ions ranging from m/z 100 to m/z 450, indicating the presence of polyaromatics such as fluoranthene and pyrene. Other significant products that were obtained through the analysis of the NL OC were methylpyrenes (m/z 216), methyl phenol (m/z 220) and toluene (m/z 92) and in NL SCC, a high intensity of m/z 220 and higher alkenes was observed.
An intense anthracene signal (m/z 178) was observed for Pandra SCC while the product mixture from
Pandra OC was mainly dominated by alkanes and alkenes (Fig. 2). An intense signal corresponding to o-xylene (m/z 106) was observed in the case of both Moonidih OC and SCC (Fig. 3). Alkane peaks from C17 to C28are observed at the end of the chromatogram. The narrow peaks (for C17 to C28 alkanes and alkenes) vary from coal to coal, showing higher intensity in the case of high ash Pandra coal and low intensity in the case of low ash Bhagabandh and Moonidih coals (Fig. 2 & 3). Toluene was found to be a major product from Bahula OC. Indeed, in general the OCs were found to be good sources of monoaromatics such as benzene, toluene and o-xylene (Islas et al. 2000).
Average aromatic content in the Moonidih OC is marginally higher than the corresponding SCC (Fig. 3). In the case of the other four coals studied – Pandra coal (high ash non-coking coal), Bahula coal (high ash non-coking coal), Bhagabandh coal (low ash coking coal) and Neyveli lignite (low ash lignite), the average content of the aromatics in SCCs as compared to OCs is higher than the aliphatics. Aromatics larger than naphthalene could have been formed from char or some other non-volatile products during the pyrolysis process or through numerous reactions involving alkylated derivatives (Sharma et al. 2019).
As explained by Radenovic (Rađenović 2006), the pyrolysis conditions result in the formation of free radicals via the cleavage of associations between the primary and secondary units of the condensed polyaromatic coal macromolecular network.
The SCCs obtained were found to have more volatile matter (VM) and easily degradable units compared to the OCs, resulting in increased tar formation from the SCCs. The various reactions that result in tar formation include depolymerisation, dehydrogenation, decarboxylation, hydrogenation and the stabilization of the heavy weight compounds obtained through secondary cracking reactions from the primary structure of coal. Thus, some information could be derived about the action of the e,N solvent system (NMP containing a small amount of EDA) studied through the Py-GCMS analysis of the samples. NMP, a polar diprotic solvent that has a good affinity for aromatics, is widely used in petroleum refineries. EDA has been found to be a good coal swelling and extracting solvent through the breaking of hydrogen bonds (Pande 2000; Pande and Sharma 2001, 2002). The synergistic action of the two solvents when used together, i.e., EDA, that cleaves the H-bonds, and NMP, that extracts aromatics, and leaves SCCs with negligible ash content. Fig. 4 summarizes the different compounds obtained from the OCs and SCCs. The highest amounts of mono-aromatics were obtained from the Bahula and Moonidih coals (both OC and SCC). NL OC gave predominantly mono-aromatics and alkenes, whereas NL SCC afforded relatively more n-alkanes and alkenes. The amounts of mono-, di- and tri-aromatics obtained during pyrolysis of the Bhagabandh SCC was relatively high as compared to the other coals (OCs and SCCs) indicating the effectiveness of the e,N solvent in extracting aromatics from this sample. Bhagabandh gave the highest extraction yield of all the coals. Based on these results, these coals and their SCCs could constitute a good source for the recovery of mono-aromatics such as benzene, toluene, xylene and other substituted aromatic compounds.
When the area % of the five categories of the compounds were compared for their composition for a coking coal (Moonidih coal), and a non-coking high ash coal (Pandra coal), it was observed that the ash contents and the coking characteristics of coals effect the formation of certain coal macromolecules (Fig. 5(a) and (b)). The SCC (Fig. 5(b) obtained from the extraction of a coking coal Moonidih shows more monoaromatic content whereas that from the non-coking Pandra coal shows more alkenes. The tars of the SCCs obtained through the e,N solvent system show that the e,N extraction is largely aimed at the extraction of the specific moieties in coal, earlier studies have shown that the SCCs of the non-coking coals showed coking behavior extending their applicability in the steel industries (Pande 2000). Thus, the solvent extraction of the coals was found to enhance their coking properties by removing mono-aromatics and certain triaromatics, i.e., the use of mixed solvents may improve the coking properties of coal as well as significantly reducing the ash content.
3.2.2 PCA Analysis
PCA was carried out for the OCs and SCCs of the five coals to identify structural similarities and differences between the samples before and after solvent extraction. Additionally, through PCA analysis, it was possible to detect zones showing variability in the data and detect any outliers/abnormalities (Melendez et al. 2012). PCA was performed with coal samples as samples (rows) and compound groups/compounds as features/variables (columns). Principal Component 1 (PC1) with an eigen vector corresponding to the largest Eigen value of 354.35 captured 61.88% of variance in the data. Similarly, Principal Component 2 (PC2) with an Eigen vector corresponding to the second largest Eigen value of 120.31 captures 21.01% of variance in the data. Thus, the samples were represented on a 2D dimensional plane to visualize the spatial distribution relative to each other with axes PC1 and PC2 capturing total 82.89% of variation in the data, when total peak area corresponding to compounds grouped as in Fig. 4, were considered as features. The loadings plot represents weight coefficient of each compound group, denoting the contribution of original features (dimensions) to the directions of principal axes of variation in data (Principal Components/Eigen Vectors).
In Fig. 6 (b), n-alkanes and n-alkenes make the major contribution to PC1 in the positive direction and the other compounds as a whole make significant contributions in the negative direction. Aromatic compounds have smaller weights as compared to alkanes, alkenes and other compounds, and they contribute to the negative side of PC1.Cycloalkanes lie close to the origin and have a small contribution in the positive PC1 direction. Aromatics have predominant contributions to PC2 with monoaromatic compounds having a high positive weight and diaromatics, triaromatics having negative weights. Alkenes and alkanes have smaller contributions in PC2 as compared to aromatics, with nearly equal weights but in opposite directions, i.e., positive and negative PC2 respectively.The original coals cluster together on the positive side of the PC1 axis, whereas, the super clean counterparts are spread on the negative side of PC1 with the exception of Neyveli Lignite SCC, which shows significant similarity to the original coals in the direction of maximum variance in the data.
After removing Neyveli lignite from the analysis, in the scores plot Fig. 7(a), the SCCs of Bahula, Moonidih, Bhagabandh and Pandra flip to the positive side of the PC1 in a pattern that is almost a mirror image about the PC2 axis. Since in the loadings plot, the n-alkanes, n-alkenes, monoaromatics and others have flipped to opposite sides too, the nature of the PC1 axis and its distinguishing characteristics based on the relative composition of samples is retained. The PC2 axis is still composed of a high positive weight of monoaromatics and negative weights of diaromatics and triaromatics. Diaromatics and triaromatics now have negligible loadings on PC1 and hence the specificity of PC2 increases and its nature is retained. Fig. 7 would be considered for further analysis.
PC1 distinguishes samples based on aliphatic and monoaromatics content whereas PC2 primarily differentiates samples based on aromatic content. Original coals have negative scores on PC1 and thus have high aliphatic content, comprising n-alkanes, n-alkenes and cycloalkanes, as compared to solvent extracted super cleaned counterparts. For all samples, the major shift on PC1 as compared to a mild shift in PC2 scores supports the conclusions drawn by Sharma (Sharma et al. 2019). The general trend of decrease in Aliphatic/Aromatic content scores post e,N solvent extraction for both non-coking high ash and coking low ash coals. Intense aromatics bands in SCCs signifying increase in aromatics as pointed out in ATR FTIR analysis of samples, can be observed from the fact that monoaromatics have a positive loading on PC1 and coals’ super cleaned samples having a higher negative PC2 score (shift towards more di- and triaromatics content) as compared to their original ones (Dhawan and Sharma 2019). This transformation was found to be more pronounced for non coking coals, i.e Bahula and Pandra Coals, for which average aromatic content increases as well as for all other aromatic groups, as reported by Sharma et al. (Sharma et al. 2019) . Significant increase in triaromatics content of Pandra Coal post solvent extraction is the highlight among this trend.
A total of 91 compounds can be identified through Py-GCMS of all samples together. The loadings and scores plot when PCA is done without clubbing compounds into groups, is presented in Fig. 8. Compounds having loadings more than 0.1 on either PC1 or PC2 have been labelled, as their loadings is considered significant. Original and super cleaned coals form clear distinct clusters on negative and positive halves of PC1, respectively. The all compound plots are found to be consistent with the analysis drawn from grouped compound plots, although specific deductions are difficult to make, given the large number of weights represented in loadings plot yet, in PC2 which comprises of significant loadings from both aromatics and aliphatic compounds, the distinguishing characteristics are not retained with both effects balancing each other making respective clean samples almost parallel to their raw ones. However, Bahula coals with significantly high content of phenols has been clearly highlighted. For low ash coking coal Moonidih, having highest content of polyaromatic structures, total aromatics content has been reported to be comparable before and after solvent extraction, whereas, distinct increase in other structural units like pyrrolidinones makes it super cleaned sample acquire high positive score on PC1.
3.2.3 Gaussian Contour Plots
As observed in Fig. 9, both contours are independent. The clusters are well separated and thus, the shape of mixture contours do not differ much from the component contours, given that the weights (number of sample points) are comparable. Both SCCs and OCs are positively correlated (ellipsoidal shape tilted to the right) and their PC1 and PC2 scores co-vary in the given 2D plane. Positive correlation implies that for both OCs and SCCs, as the positive score increases in the PC1 axis, the positive score in PC2 increases. Moonidih, Bhagabandh and Pandra OCs lie on the same contour curve and have the same probability density value. They belong to the original coals cluster under the same confidence limits.