A Study of Coal Extraction with Electron Paramagnetic Resonance and Proton Nuclear Magnetic Resonance Relaxation Techniques

Energy & Fuels 1994,8, 907-919 907

A Study of Coal Extraction with Electron Paramagnetic Resonance and Proton Nuclear Magnetic Resonance

Relaxation Techniques D. C. Doetschman,*tt R. C. Mehlenbacher,? and 0. It02

Department of Chemistry, State University of New York, Binghamton University, Binghamton, New York 13902-6000, and Institute for Chemical Reaction Science, Tohoku

University, Katahira 2-Chome, Aoba-ku, Sendai 980, Japan

Received September 13, 1993. Revised Manuscript Received March 14, 1994"

An electron spin and proton magnetic relaxation study is presented on the effects of the solvent extraction of coal on the macromolecular network of the coal and on the mobile molecular species that are initially within the coal. The eight Argonne Premium coals were extracted at room temperature with a 1: 1 (v/v) N-methylpyrrolidinone (NMP)-CSz solvent mixture under an inert atmosphere. As much solvent as possible was removed from extract and residue by treatment under vacuum oven conditions Torr a t 145-150 "C) until constant weight was achieved. The extraction, without further washing with other solvents, results in substantial uptake of NMP, apparently by H-bonding or acid-base interactions. The NMP uptake tends to be higher, and the NMP tends to be more tightly bound in coal matter with higher heteroatom (N, 0, S) content. The molecular material in the medium rank bituminous coals is more aromatic and heteroatom-poor than the macromolecular material and is mobilized by the extracting solvent. Likewise, the more aromatic and heteroatom-poor molecular free radicals are also extracted. However, mobilization of the molecular free radicals by solvent and the exposure of free radicals in the macromolecular matrix to solvent or species dissolved in the solvent result in preferential reactions of the more aromatic and heteroatom- poor free radicals. Greater losses of extract free radicals, being the more aromatic, occur than residue free radicals. As a consequence, the surviving extract radicals exhibit a greater heteroatom content than the original whole coals, as determined from EPRg value changes. The electron paramagnetic resonance (EPR) spin-lattice relaxation (SLR) of these coal free radicals has previously been inferred (Doetschman and Dwyer, Energy Fuels 1992,6,783) to be from the modulation of the intramolecular electron-nuclear dipole interactions of the CH groups in a magnetic field by motions of the radical in the coal matrix. Such a modulation depends on the flexing amplitude and frequency and to a lesser extent upon the electron spin density a t the CH groups in the radical. The observed EPR SLR rates decrease with coal rank in agreement with the smaller spin densities and the lower rocking amplitudes that are expected for increasing aromaticity with rank and increasing polycondensation at the highest ranks. The EPR SLR rates are found to be generally faster in the extracts (than residues) where the molecular species would be expected to be smaller and more flexible than in the cross-linked, polymeric, macromolecular matrix of the residue.

Introduction The extraction of molecular species by solvent from the

macromolecular matrix of the organic part of the coal depends on a number of factors that are relevant to practically all coal utilization or derivitization processes. The molecular structure of the coal is important, par- ticularly in regard to its flexibility and permeability to smaller molecular species. One may regard the problem in terms1p2 of a hydrogen-bonded, polymeric, or polycondensed network. It is desirable that the smaller molecular species within the network be mobile in order for extraction to occur. The extraction process takes place as a result of mobile solvent molecule penetration into the network and as a result of the subsequent mobilization and

t State University of New York. t Tohoku University.

Abstract published in Advance ACS Abstracts, May 1, 1994. (1) Given, P. H.: Marzec, A.: Barton, W. A.: Lynch, L. J.; Gerstein, B.

C . Fuel 1986,65,155. (2) Derbyshire, F.; Marzec, A.; Schulten, H.-R.; Wilson, M. A.; Davis,

A.; Tekely, P.; Delpuech, J.-J.; Jurkiewicz, A.; Bronnimann, C. E.; Wind, R. A.; Maciel, G. E.; Narayan, R.; Bartle, K.; Snape, C. Fuel 1989,68, 1091.

0aa7-0624/94/250a-0907$04.50/0

dissolution of small coal molecular species by the solvent in the network. According to this hypothesis,'P2 coals comprise an intractable molecular matrix and smaller molecular species. This hypothesis was put forth most recently1p2 to explain nuclear magnetic resonance (NMR) relaxation in coals.- The language of this hypothesis will be employed extensively in this paper. Yokono and Sanada3 studied the proton NMR spin-spin relaxation of several whole coals and detected multiple decay components in them. They also examined the proton NMR spin- lattice relaxation (SLR) times in whole coals and their pyridine extracts and residues, finding a different, but single, decay component in each coal, extract, and residue.3 Jurkiewicz et al. in the debate of Derbyshire et al. concerning the macromolecular-molecular hypothesis also discussed the proton NMR relaxation in pyridine-swollen

(3) Yokono, T.; Sanada, Y. Fuel 1978,57, 334. (4) Jurkiewicz, A.; Marzec, A,; Pislewski, N. Fuel 1982, 61, 647. (5) Marzec, A.; Jurkiewicz, A.; Pislewski, N. Fuel 1983,62, 996. (6) Barton, W. A.; Lynch, L. J.; Webster, D. S. Fuel 1984,63, 1262. (7) Jurkiewicz, A.; Marzec, A.; Idziak, S. Fuel 1981,60, 1167. (8) Yang, X.; Larsen, J. W.; Silbernagel, B. G. Energy Fuels 1993, 7,

439.

0 1994 American Chemical Society

908 Energy & Fuels, Vol. 8, No. 4, 1994

coals and pyridine-swollen, pyridine-extracted coal residues,4,6-7 finding distinguishable macromolecular and small molecular spin-spin relaxations. In a high-resolution, solid-state NMR study, two exponential inversion recovery components with different T1 were found to fit the SLR data for the Argonne coals.g The NMR relaxation results on coals undergoing thermal softening show that the coals contain mobile regions or structures, as well as immobile or rigid regions or structures,'OJ' results that also conform to the macromolecular-molecular hypothesis. Takano- hashi and Iino12 have made an EPR study of the line shapes of whole coals and of the residues and extracts from a three-step, multiple-solvent extraction of the coals, which showed considerable changes in the line widths, g values, and spin concentration upon extraction.12 They observed an increase in the g value in an acetone-soluble fraction. The g values suggested that there was a preferential solubility of heteroatom-bearing free radicals. The line widths of the acetone-soluble fractions suggest that they have decreased aromaticity.12 Evidence was also found that the total number of free radicals in the original whole coals was not conserved in the course of extraction.12 These results, together with recent ~haracterizations~~J~ of whole Argonne Premium coals by EPR relaxation methods, suggested that a systematic study should be done of the EPR and NMR relaxation characteristics of this representative, well-characterized set of whole coals and their extracts and residues. In particular, the EPR SLR in the maceral constituents of the whole Argonne coals was evidently from the modulation of the electron-proton magnetic dipolar interaction.13 Thus, EPR SLR is sensitive to the flexing of the matrix or host of the free radical13 as well as to the chemical structure of the free radical that governs the magnitude of the interaction.I6J7 Thus, one would expect EPR SLR to give information about the dynamics and chemical characteristics of the free radicals in the molecular matrix and in the molecular species that survive the extraction process.

Within the macromolecular-molecular hypothesis, the primary role of the extracting solvent is to present the molecular species with a milieu in which they are more soluble than in the coal macromolecular matrix. However, a secondary role of the solvent is to enter the macromolecular matrix and, if possible, alter its structure, so as to expose the coal molecules to solvent. Given the common chemical origins of the coal molecular species and the macromolecular matrix, a good solvent molecule may be attracted to the macromolecular matrix almost as strongly as it is to the molecular species. Thus, coal swelling is a well-known characteristic, being greatest in the best coal-

(9) Dela Rosa, L.; Pruski, M.; Lang, D.; Geratein, B. Energy Fuels 1992, 6, 460. (10) Lynch, L. J.; Sakurovs, R.; Webster, D. W.; Redlich, P. J. Fuel

1987,67, 1036. (11) Sakurovs, R.; Lynch, L. J.; Barton, W. A. In: Chapter 11 of

Magnetic Resonance of Carbonaceous Solids; Both, R. E., Sanada, Y., Eds.; ACS Adv. Chem. Ser. 1993,229,229.

(12) Takanohashi, T.; Iino, M. Energy Fuels 1990, 4, 452. (13) Doetachman, D. C.; Dwyer, D. W. Energy Fuels 1992,6,783. (14) For example, see: (a) Tsvetkov, Y. D.; Dzuba, 5. A.; Gulm, V. 1.

In: Magnetic Resonance of Carbonaceous Solids; Botto, R. E., Sanada, Y., Eds.; ACS Ado. Chem. Ser. 1993,229; Chapter 23, p 443. (b) Chen, X.; McManus, H.; Kevan, L. Chapter 24, p 452. (c) Clarkson, R. B.; Wang, W.; Brown, D. R.; Crookham, H. C.; Belford, R. L. Chapter 27, p 507 (d) Silbernagel, B. G.; Gebhard, L. A.; Bemardo, M.; Thomann, H. Chapter 29, p 539. (e) Bowman, M. K. Chapter 32, p 605.

(15) Vorres, K. 5. Energy Fuels 1990,4, 420. (16) Retcofsky, H. L. Magnetic Resonance Studies of Coal. In Coal

Science; Gorbaty, M. L., Larsen, J. W., Wender, I., Me.; Academic Press: New York, 1982; p 43.

(17) Revofsky, H. L.; Hough, M. R.; Maguire, M. M.; Clarkson, R. B. Chapter 4 in Ado. Chem. Ser. 1981, No. 192,37.

Doetschman et al.

solvent combinations.18 A number of NMR studies of pyridine bound to or sorbed onto coals reveal a pyridine whose rotational freedom is hindered in ways that depend on the type of ~ o a l . ~ " ~ ~

This interaction is thought to be through Lewis acid- base or H-bonding interactions with the N lone pair of pyridine.23 In their study of the extraction of the Argonne Premium Coals with NMP and CS2, with a slightly different procedure than we employ, Takanohashi and Enol2 report extract and residue elemental N analyses that usually equal or exceed the analyses of the whole coals, consistent with the retention of some NMP. In a Fourier transform infrared (IR) study, the high N analyses in NMP extraction were noted and were shown by IR to correspond to 1-3 76 NMP retention in the extractaZ4 Hall and Larsen have shown that NMP forms a gel of fixed composition in the swelled Illinois No. 6 coal from which the tightly bound NMP can be removed only by an activated pr0cess.2~ In the present study, similar information is obtained about NMP retention from our extraction yields, extract characterization, and NMR studies.

The organic free radicals observed by EPR in coals and other carbonaceous materials have properties consistent with stable, odd-alternate, polycondensed, ring systems.16J7 Less clear, however, is the chemical nature of the heteroatom-containing species in coals known to impart positive EPR g value shifts that correlate with oxygen, nitrogen, and sulfur content.16J7 It has been argued that the unusual stability of the odd-alternate hydrocarbon system is responsible for the persistence of odd-alternate, polycondensed free radicals over the long coal lifetime. However, this argument might be questioned for the apparently highly reactive phenoxy, semiquinone, etc., radicals proposed for g-shifted species in C O ~ S . ~ ~ ~ ' ~ Thus, further clarification is needed as to whether (a) heteroatom incorporation into more stable odd-alternate hydrocarbon ring systems gives the observed g shifts, whether (b) immobilization of reactive, heteroatom-containing free radicals leads to their longevity, or whether (c) these reactivity assumptions are unfounded.

When coal is treated with an extracting solvent, the molecular species are mobilized by the solvent, and while the macromolecular matrix may be mobilized to a much lesser extent, it is exposed to other coal molecular species and to solvent molecules at previously protected regions. There have been conflicting reports about the conservation or nonconservation of the numbers of spins in coal samples upon solvent treatment or extraction.12~w28 Some authors report that the numbers of spins decrease in room temperature one reports little change upon NMP extraction at 202 OC,30 and another reports that NMP extraction at room temperature leads to small

~

(18) Iino, M.; Takanohaehi, T.; Ohsuga, H.; Toda, K. Fuel 1988, 67, 1639.

(19) Silbernagel, B. G.; Ebert, L. B.; Schlosberg, R. H.; Long, R. B. In Coal Structure; Gorbaty, M. L., Ouchi, K., Ma.; Adv. Chem. Ser. 1981, 192, 23.

(20) Vassallo, A. M.; Wilson, M. A. Fuel 1984, 63, 571. (21) Ripmeeeter, J. A.; Hawkins, R. E.; MacPhee, J. A.; Nandi, B. N.

Fuel 1986,65,740. (22) Vassallo,A. M. Chapter9inMagneticResonanceof Carbonaceous

Solids; Both, R. E., Sanada, Y., Eds.; Adv. Chem. Ser. 1993,229,202. (23) Green, T. K.; Larsen, J. W. Fuel 1984, 63, 1538. (24) Cai, M. F.; Smart, R. B. Energy Fuels 1993, 7,52. (25) Hall, P. J.; Laraen, J. W. Energy Fuels 1993, 7, 47. (26) Yokokawa, C. Fuel 1969,48,29. (27) Duber, S.; Wieckowski, A. B. Fuel 1984, 63, 1641. (28) Wind, R. A.; Jurkiewicz, A.; Maciel, G. E. Fuel 1989, 68, 1189. (29) Jurkiewicz, A.; Wind, R. A.; Maciel, G. E. Fuel 1990,69, 830.

A Study of Coal Extraction

increases in the total number of spins in some coals but to minor decreases in others.12 The present NMP extractions have been performed at room temperature, but the resulting coal fractions were treated a t 145-150 "C under vacuum oven conditions, for as much as 24 h, in order to drive the solvents off. Both processes were performed under strictly oxygen-free conditions. Present results show that the total number of free radicals in the original whole coals is not conserved in the course of extraction. Takanohashi and Iino have also performed elemental analyses of the Argonne Premium coal extracts and residues12 that also bear on the relative heteroatom contents and degrees of aromatic ring condensation. There were previous indications from g value changes upon extraction that there is a preferential solubility of heteroatom-bearing free radicals.lZ The present results point to a preferential solubility of more aromatic, heteroatom- poor material. The material then suffers a preferential loss of aromatic, heteroatom-poor free radicals by free radical reactions. The material finally reaches a point where it contains proportionately more heteroatom-rich free radicals than the original whole coal.

It is very widely recognized that molecular oxygen causes complications in the observation of free radical EPR by its reactivity with free radicals and by its paramagnetic interactions with the free radicals. It has been recently pointed out that coal NMR SLR rate measurements may likewise be very sensitive to the presence of oxygen.=tm Processes in the present study have been performed with the strictest possible exclusion of molecular oxygen, consistent with the nature of the processes. Demineral- ization procedures were omitted in this study in order to eliminate the difficulty of doing them in as strict absence of oxygen as the other procedures. We were also concerned about possible uncontrolled chemical effects of demineralization on reactive, coal free radicals and physical effects on coal porosity in the extraction process. Possible complications in the measurement of EPR parameters in the presence of mineral matter are discussed in the Discussion section.

Energy & Fuels, Vol. 8, No. 4,1994 909

centrifuge and decanted, fiitered, and washed through l-pm Millipore filters in the glovebox. The extractions of the solids were repeated in like manner until the extract achieved an arbitrarily chosen optical density <2 at 505 nm, collecting all washings and extracts in a large flask. The CSZ and NMP solvents in the preweighed flasks of extract or residue were removed by rotoevaporation to constant weight with a procedure identical to the initial drying. Final weights were recorded. Constant weight was achieved in both water and solvent removal in 12 h or less. Extraction is believed to be quantitative except in the case of the Beulah Zap coal where material loss is encountered due to the "explosion" of the sample caused by the release of adsorbed gases when the sample is lowered into the rotoevaporator heat bath.

EPR and NMR sample tubes, preweighed with caps, were loaded respectively with approximately 40 and 150 mg of whole coal, extract, and residue in the glovebox. The capped, loaded tubes were reweighed, and then, in the glovebox, silica wool wads were pushed into the tubes over the samples in order to prevent dispersal of the sample during the subsequent evacuation. The samples were attached via O-ring fittings to a diffusion pump-equipped Kontes vacuum line and pumped to less than 10-9 pmHg pressure. Finally, the samples were flame-sealed under vacuum and stored in the dark at about -10 OC except when being examined. The remainder of the samples was stored, as required for other analyses, in vials under nitrogen atmosphere in the glovebox.

The extract samples were characterized beyond the work of Takanohashi and Iino12 by the following methods. Transmission IR spectra of the extracts, pressed into KBr pellets (6.3 mg of extract and 1 g of KBr), were run (600- 4000 cm-l) on a Perkin-Elmer 1420 IR spectrometer. Spot checks of the elemental analyses, in addition to those performed by Takanohashi and Iino,lZ were done as follows: Upper Freeport extract and residue C, H, N, ash; Illinois No. 6 residue N, S; Illinois No. 6 extract and Pocahontas No. 3 and Wyodak Anderson extract and residue N; Blind Canyon, Lewiston Stockton and Pitts- burgh No. 8 residue S. The C, H, N, and ash analyses for the high yield, Upper Freeport extract are in close agreement with literature v a l u e ~ ~ ~ J ~ for the whole coals. The N analyses are uniformly higher than the whole coal values,15 except for the unrepresentative Upper Freeport residue. The residue S analyses are not out of line with dry ash freelZ and dry15 basis whole coal literature values, exceeding literature values in all cases but the Illinois No. 6 residue. Solution proton NMR spectra of solutions of known extract concentrations by weight in pyridine-da were examined and compared quantitatively with a standard proton NMR spectrum of an NMP solution of known concentration in pyridine-& (Merck Isotopes). NMR spectra were acquired on a Bruker AM360 360-MHz spectrometer with 7r/2 pulses (12 ps) at repetition times of 5 s, by calculating the Fourier transform (FT) of the resulting free induction decay (FID). A solid-echoal proton NMR (r/2,-m/2,) technique was employed to examine and compare the solid extract and whole coal spectra. The spectra were obtained on the spectrometer described above with 12-ps 1r/2 pulses and intervals 7 = 10 ps with a repetition period of 20-30 s. In order to be able to compare the retained NMP solid-echo spectra of the solid extracts on a relative basis with FID-FT spectra, the same length,

Experimental Section

The eight Argonne coals were extracted repeatedly with 1:l (v/v) NMP-CSz solutions at room temperature. After thorough mixing, approximately 5-g samples of the coals were removed in a pre-pure grade Nz(g) atmosphere glovebox from the vials in which they were shipped17 and placed in preweighed round-bottom flasks. The samples in the stoppered flasks were weighed and then dried by careful rotoevaporation (Buchi/Brinkman RE111) (in UQCUO with Nz(1) trap) to a constant weight in a bath at 145-150 OC. The rotoevaporator was outfitted with a pre- pure grade N2(g) tank for flushing the system at the times of unstoppering/flask attachment and flask removal/ stoppering. To the flasks of dried coal in the glovebox was added 100 mL of the NMP-CS2 solution. The NMP was Aldrich HPLC grade, and the CSZ was Aldrich spectrophotometric grade. The flasks were stoppered and placed in the initially room temperature bath of aQuantrex 6210 ultrasonic agitator for 30 min. The temperature of the bath typically rose 5 OC in the course of the sonication. The stoppered coal and extract solutions were centrifuged for 30 min at 15 OOO rpm (17 540 RCF) in a Sorvall RC-5B

(30) Seehra, M. S.; Ghosh, B.; Zondlo, J. W.; Mintz, E. A. Fuel Process Technol. 1988,18, 279.

(31) Boden, N.; Gibb, M.; Levine,Y. K.; Mortimer, M. J. Magn.Reson. 1974, 16, 471.

910 Energy &Fuels, Vol. 8, No. 4,1994

single 812 pulse was employed with a repetition period of 1 s in both cases. In all cases, the FT of the FID or the solid echo was calculated to obtain the NMR spectra of the coal, coal extract, or retained NMP.

The residues have been further characterized in the following experiments. Weighed amounts of the residues were washed by sonication with acetone-de (Merck Iso- topes) and filtered, and then the solution proton NMR spectra of the washings were examined. The amount of washable NMP retained in the residues was determined by comparison with a proton NMRspectrum of a standard solution of NMP in acetone-&. Solid-echo and FID experiments, which were similar to those done on the extracts, were performed on the solid residue samples.

The NMR SLR times of the solid extract and residue samples were measured from the recovery of (the FT of) the solid echo described above, as a function of delay time after a K (24 ps) inversion pulse. The repetition times were 20-30 s. The T1's were determined from a least- squares fit of the parameters a, b, and TI in the equation

(1)

to the FT intensity, as a function of delay time T a t a particular chemical shift in the spectrum, chosen so as to avoid the NMP peak. TI'S were otherwise independent of the choice of chemical shift. No evidence of multiexponential or multiphasic recovery was found away from the NMP line to be discussed in the Results section.

Several types of CW and pulsed EPR measurements were made on the extract and residue samples. Measure- ments made on the whole Argonne Premium coals have been published separately in a paper13 that describes the experimental details of each type of measurement. All experiments were performed on a Bruker ESP-380 pulsed EPR spectrometer. The numbers of spins in the broad- line components of the coal fractions were determined, with reference to a standard,l3 from the integrated intensities of the magnetic field sweeps of the two-pulse echo, corrected for spectrometer dead time from the decay of the echo. The numbers of spins in the narrow-line (inertinite) components were determined, with reference to a standard,l3 from the integrated FID FT spectra, similarly corrected for spectrometer dead time from the FID. Broad-lineg values were determined from the centers of the CW EPR spectra a t 90-dB power attenuation, as were the broad-line line widths from the widths. The narrow-line g values were determined from proton NMR gaussmeter (Micronow 515-1) field measurements at the center of the FID, as determined by visual oscilloscope examination of the FID. Both determinations employed microwave frequency counter measurements. Narrow- line Lorentzian line widths (and 2'2 values) were determined from the decay time of the FID. Broad-line spin- spin relaxation times TZ were determined from the lower limit of the two-pulse echo decay coherence memory time, TM, versus the magnetization turning angle (8) of the pulse from an appropriate plot of ~ / T M versus the appropriate function of 8. Broad-line SLR times TI were determined from the two-pulse echo saturation recovery, which followed a picket-fence, multipulse saturation sequence. The whole pulse sequence is designed to achieve saturation over a broad frequency range in order to avoid spin diffusion contributions to the recovery. The narrow-line TI were determined directly from the inversion recovery of the on-resonance FID decay a t a particular decay time in the FID without intermediate Fourier transformation.

I = a - be-TIT1

Doetschman et al.

Extract Yield UF

n 100

I

80 1 s i

0 ' 70 75 80 85 90 95 100

Percent Carbon (dmmf) Figure 1. Equivolume NMP-CS2 extraction yields by weight from the Argonne Premium Coals versus C content in the whole coal (coal rank). The results have been corrected for retained NMP and are presented on a dry-mineral-matter-free (dmmf) basis. The Beulah Zap coal extraction was not quantitative and is omitted. See text for abbreviations. It is estimated that no actual loss of material exceeded 5 7%. See text for other indirect indications of the uncertainty in the values presented using an NMR-based estimate of NMP content.

Results

Extraction. The extract yields by weight from the Argonne Premium Coals are shown as a function of rank in Figure 1. The data were determined from the combined weights of extract and residue, corrected for the propor- tioning of the NMP retained in them as determined from the extract NMP solution NMR (see next section). All mineral matter was assumed to remain with the residue, as was confirmed by a spot check of the ash content (>59% ) of the Upper Freeport (UF) residue (90% extraction yield), and negligible materials losses were assumed. NMP NMR internal consistency checks regarding total retained solvent show materials losses to be no more than about lo%, except for the omitted Beulah Zap (BZ) results (see Experimental Section). Except for the very high solubility of the UF coal, the yields all lie between 20% and 55%. While we do not have quantitative results, it should be noted that the BZ extraction yield qualitatively appears to be even less than the Wyodak Anderson (WA) coal. The BZ yield was so low, in fact, that some of the measurements that were made on other extracts were not possible for the BZ samples. While the trends with rank are consistent with the results of Iino et aZ.,13 there are too few data to define clearly the rising yield with rank, the peaking a t 85-87 % C, and the rapidly decreasing values above 87% C that were demonstrated by them. A plot of extract yield versus heteroatom conten@ (N, 0, S) in the organic fractions of the coals reveals essentially no more than the reverse of this trend with coal rank.

The IR spectra of the extracts display the kinds of trends with coal rank that are generally expected. The aromatic C-H out-of-plane bending and C-C stretching bands are observed to increase with the rank of the parent coal above about 80% C but are weak and not clearly identifiable in the extracts of coals of lower rank. This is in accord with increasing aromaticity of the parent coal with increasing rank. The methyl and methylene C-H stretching peaks in the extracts of all eight coals are observable and have intensities that decrease with parent coal rank. Also noteworthy is the fact that the ratio of methyl to methylene C-H stretching peak intensities appears to increase with coal rank. This suggests that, as the aliphatic nature of


the coal extracts increases with decreasing parent coal rank, there is a tendency toward longer chains with relatively more methylene groups per chain (usually terminated with a methyl group). A broad band of C-O stretching modes, superimposed by that of NMP" in some of the coal extracts, is found between 1550 and 1750 cm-' throughout the range of extracts. The broad band exhibits relatively subtle changes in shape from one sample to another and is probably from a variety of functional groups that contain the C=O group. The overall intensity of the broad C-0 stretching band appears to be more a function of the particular coal than a function of parent coal rank. The absence of the strong, sharp C-0 stretching peak characteristic of coal air oxidation is missingin all samples, presumably because of the oxygen-free extraction and sample preparation. The fact that the broad, weak C-O peaks are observed in the extracts of all of the coals probably reflects some ability of the NMP in the solvent to extract the molecules with polar functional groups across the range of coal ranks.

The solution proton NMR spectra of the extracts exhibit broad lines or bands in the aromatic (7-8 ppm) and aliphatic (2.2-3 ppm) regions. Because of the breadths of the bands and the small samples employed, the intensities are weak relative even to the impurity proton NMR peaks in the NMR grade deuterated solvents. Therefore, quantitative integrations of these bands was not feasible, permitting only qualitative observations. Surprisingly, the relative intensities of the aromatic and aliphatic bands did not differ markedly through the range of ranks of parent coals. The intensity of the aromatic region, which is somewhat less obscured by solvent peaks than the aliphatic region, increases from a negligible level below 75% C in the parent coal to a maximum at around 85% C in the parent coal and decreases again to an undetectable level again above 89% C in the parent coal. The constant relative NMR intensities of the aliphatic and aromatic proton bands may reflect the particular extracting character of the solvent across the range of coal ranks. The intensity change in the extract spectrum with coal rank suggests that, as the coals increase in aromaticity, there is a corresponding increase in numbers of aromatic protons up to around 85% C in the parent coal. Then, as one progresses from the more aliphatic to the more aromatic coal beyond 85% C in the parent coal, the extracted molecules suffer not only a loss of aliphatic protons but also a comparable loss of aromatic protons as a result of aromatic ring condensation.

Solvent Retention. There are many indications that the NMP from the extraction solvent is retained in the extracts and residues. First, NMP is observed in both the IR (by its prominent CO stretching band)" and solution proton NMR spectra of the extracts by comparison with authentic NMP or NMP added to the sample. Extracts that exhibit obvious NMP peaks at ca. 1670 cm-' (authenticated with the IR spectrum of an NMP sample) were the Illinois No. 6 (IL), Pittsburgh No. 8 (PI), Lewiston Stockton (LS), and Pocahontas No. 3 (PO) coals. The possible presence of NMP IR peaks in the BZ and WA extracts cannot be ruled out because of the presence of substantial C-O stretching intensity in this region from other sources. When the residues are washed with acetone- d6, the resulting solutions exhibit the proton NMR spectrum of NMP ( ~ c H ~ = 2.73; 6c1.1~ = 1.98, 2.17, 3.35). There is also an additional broad singlet peak in the proton NMR FID and solid echo spectra of both the solid extracts


NMP Retained in Extract and Residue from Extract & Residue Weights

M I D I /

0 5 10 15 20 25 Percent (0 + S) in Whole Coal

Figure 2. Total amounts by weight of NMP retained in the extracts and residues on the basis of weight increase per 100 g of organic matter in the original whole Argonne Premium Coals presented versus the combined 0 and S atom contents of the whole coals. The results presented here are underestimated, as a result of materials losses, by no more than 16 % . and residues whose chemical shift range overlaps those of the solution NMP peaks. This broad peak is missing in the proton NMR spectra of the dried, solid, whole coals, which have not been subjected to extracting solvent. Finally, there is a substantial increase in the total weight of extract plus residue compared with the original weight of the whole dry coal, and there are increases in N content of the extracts and residues relative to the whole coals.12 In view of the high CS2 volatility, the temperature (145- 150 "C) and pressure (-15 pmHg) "vacuum oven* conditions employed in the removal of solvent, and the absence of a CS2 peak in the 13C NMR spectra of the extract and residue wash solutions, we assumed that only NMP is retained from the extraction solvent. The S content of the residues that we spot-checked did not appear to be unusual in comparison with whole coal S content.l2Js However, complications due to mineral matter S in the residue and the lack of measurements of extract S content do not permit us to rule out the possibility of S uptake from CS2 definitively.

In Figure 2 are shown the total amounts of NMP retained in the organic matter in the extract and residue. The data are based on the difference between the sums of the weights of the extract and residue and the original weights of the whole coal. The calculations included a correction for the mineral matter content of the whole coal16 and assumed negligible material losses, except for the omitted Beulah Zap result (see Experimental Section). Thus, the results are underestimated by no more than about 15%. It could be argued on the basis of the results in Figure 2 that there is an increasing tendency for NMP to be retained with increasing (0, S) heteroatom content.

Figure 3 presents individually the amounts of NMP retained (a) in the extract and (b) in the residue that can be washed from it with acetone. The values (a) were determined from the integrated NMP NMFt intensities of pyridine-ds solutions of the extracts in comparison with a standard pyridine-da solution of NMP (6cb = 2.69; ~ C H ~ = 1.82,2.15,3.19). Thevalues (b) werelikewisedetermined from the integrated NMR intensities of the acetone-de washings of the residues in comparison with a standard acetone-d6solution of NMP. Based on the signal-to-noise levels of the integrated intensities of the NMP peaks, the largest estimated standard deviations in the values plotted are 0.14 g of NMP/lOO g of extract and 0.78 g/lOO g of residue organic matter. These measures of NMP retention

912 Energy & Fuels, Vol. 8, No. 4, 1994 Doetschman et al.

NMP Retained in Extract from Solution NMR I

IL J dl

0 _--I 0 20 40 60 EO 100

NMP Retained in Residue Extract Yield

from Acetone Wash

3 - \F i 6 ; L A 0 20 Extract 40 Yield 60 80 100

Figure 3. Individual amounts of NMP retained in the extracts and residues from the extracted Argonne Premium Coals, as determined from quantitative NMP proton NMR measurements of solutions of the extracts and from acetone washings of the residues, respectively. Results are presented on the basis of dry weight of organic matter in the fractions and plotted versus the yield of extract. Uncertainties are less than 0.14 g of NMP/100 g of extract and 0.78 g/100 g of residue organic matter. in extract and residue bear a good correspondence to one another, with the exception of the PO residue from which a relatively greater amount of NMP is washed than is retained in the extract. These trends also roughly parallel the total retention in Figure 2, with the exception of the WA coal. The rather different PO coal extract and residue NMP retentions in Figure 3 may, at first glance, appear not to be in agreement with the total PO NMP retention in Figure 2. However, the PO residue NMP retention predominates because of the low, ca. 20'31, extract yield and is in line with the total PO NMP retention. The sums of the weights of extract NMP and the weights of NMP washed from the residue agree well with the weight increase upon extraction for the PO, LS, and BC coals. The shortfalls, which represent NMP that cannot be washed from the residue, are about 8 (UF), 15 (PI), 20 (IL), and 40 g (WA) per 100 g of organic matter. This sequence of increasing postwash amounts of NMP retained in the residues is the same as the sequence of increases in the elemental N in the residues over the N content in the corresponding whole coals. The absolute increases in elemental N content, if attributed to NMP utpake, correspond to NMP uptakes of a few tens weight percent, in agreement with the observed differences by weight.

Figure 4 shows the LS example of the broadened NMP line that is observed on top of the broad spectrum of the coal fraction in the solid-echo and FID-FT proton NMR spectra of the solid extracts and residues. This is not observed for the whole coals. The spectrometer dead time in the FID-FT spectra may be too long to observe the full width of the coal spectrum accurately. On the basis of the behavior of pyridine adsorbed onto coal solids,19-22 we assume that the broadening is the consequence of the hindered rotation of NMP molecules in the samples. The

n

i ia u( in m i a m I -n -a + -m -I* -im -11 -in m Figure4. The 360-MHzprotonFTNMRspedrumofthenarrow line (expanded) that is assigned to retained NMP, shown on top of the broader coal extract NMR line of the solid Lewiston Stockton coal extract. The expanded scale, which is not shown, is f22.5 ppm. The small peak to the left of the expanded peak is a spectrometer background peak which was ordinarily removed by subtraction before integration. Magic angle spinning was not employed.

broadened lines attributed to NMP have distinctly different NMR SLR T1 times (0.69 s) from most of the rest of the coal NMR lines (0.05-1.5 8). Somewhat surprising to us is the fairly uniform 4-ppm line width that is observed for this feature over the whole range of coal extracts and residues. No other evidence of NMP lines of greater breadth was found either directly in the spectrum or in the form of multiexponential SLR behavior elsewhere in the spectrum. Similar looking NMR spectra have been observed for adsorbed water in coals in which multiple exponential NMR SLR recoveries have been observed.32-94 However, the absence of the peak in the dried whole coals and the fact that the "vacuum oven" conditions for the isolation of the extracts and residues were similar to those used for drying the whole coals seems to preclude natural or chemically generated water as the source of this NMR line. Drying of the Argonne coals does not appear to have a dramatic structural effect insofar as the coal proton NMR FID decay times are concerned.34

In Figure 5 are presented the relative NMP NMR intensities in the solid extracts and residues versus the s u m of the oxygen and sulfur contents in the organic matter of the fraction. The intensities of NMP were determined from the integrated intensity of each single, broadened NMP proton NMR line, as shown in the example in Figure 4. The residue values were corrected for the mineral content of the whole which was assumed to remain with the residue. Standard deviations estimated from the signal-to-noise ratio of the integrals are 0.6 on the arbitrary scale presented in Figure 5. It is clear by comparison with Figure 3a that for the solid PO extract one is observing at most 4 g of NMP/100 g of extract by solid NMR. It follows from the data in Figure 5 that at most about 0.4 g of NMP/100 g of organic matter is being observed by solid NMR for any of the other extracts and in all of the residues. Thus, the percentage of NMP observable by

(32) Dele Rosa, L.; Pruski, M.; Geretein, B. Chapter 19 in Magnetic Reclonance of Carbonaceous Solids; Botto, R. E., Sanada, Y., Eds.; Adv. Chem. Ser. 1993,229,359.

(33) Hayamizu, K.; Hayeshi, S.; Kamiya, K.; Kawamura, M. Chapter 16 inMagnetic Resonance of Carbonaceous Solids, Botto, R. E., Sanada, Y. . Eds.: Adu. Chem. Ser. 1993.229. 295.

'(34) Yang, X.; Garcia, A. R.; Larsen, J. W.; Silbernagel, B. G. Energy Fuels 1992,6, 651.


NMP Retained in Fractions Observable by Solid NMR

100 90

2 80 4 70 W

.G 60 g 50 3 40

8 ;: 10 0

0 5 10 1s 20 25 Percent (0 + S) in Fraction

Extract . Residue Figure 5. Relative amounts of retained NMP observable in the proton NMR spectra of the Argonne Premium Coal extract and residue solids. The data are plotted versus the combined 0 and S atom content in the individual coal fractions. Standard deviations are 0.6 on the arbitrary scale used.

solid NMR in the residues and extracts other than the PO and WA extracts can be seen from Figure 3 to be at most 30-40% of the total NMP present and as little as 3%! Even though relatively small amounts of NMP are retained in the PO extract (see Figure 3a), Figure 5 also shows that the amount of observable solid NMP is around 10-fold greater than in any other extract or residue and that the PO coal fractions have the smallest combined oxygen and sulfur content.12 Given that there has to be 40-45 g of NMP/100 g of organic matter in the WA residue and that at most 0.4 g of NMP/100 g of organic matter is observable, it follows that less than 1% of the WA residue NMP is observable by NMR. The WA residue has the highest combined oxygen and sulfur content.12

The relatively uniform 4-ppm widths of the solid NMR peaks in solid extract and residue spectra (see Figure 4) are assumed to result from motional narrowing of the powder pattern of fully immobilized NMP molecules. Our estimate of the average width of the powder pattern from the magnetic dipole interaction^^^ between the protons of the NMP methylene groups gives a width of 72 ppm. This estimate, together with the observed 4 ppm line width, gives an estimate35 of 1.35 X 10-6 s for the rotational correlation time rc of these observable NMP molecules in the solid fractions. We feel that NMP molecules, whose motional narrowing leads to lines broader than 80 ppm, would go altogether undetected under the conditions employed. (We find no SLR evidence for any NMP spectrum narrower than 80 ppm except for the 4 ppm line.) Since this unobservable NMP represents, in most cases, the vast marjority of the NMP, one concludes that, except for the small fraction of relatively unhindered NMP with rc - 1.35 X 10-6 s, the bulk of the NMP is sufficiently firmly bound to have rc > 2.7 X 106 s.

EPR Experiments. The numbers of electron spins per dry mineral matter free (dmmf) gram of whole coal before the extractionl3 are compared in Figure 6 with the numbers of spins found in the extracts and in the organic matter of the residues from the original (dmmf) gram of the whole coal. The latter were calculated from the actual EPR spin number measurements, the weights of the extract or residue samples, the yields of extract or residue, and the appropriate corrections for mineral matter15 in the residues. BZ extraction data are omitted because the

(35) Carrington, A.; McLachlan, A. D. Introduction of Magnetic Resonance; Wiley: New York, 1979.


Spin Concentrations

* 10 c $ 1 I

74 76 81 81 85 86 88 92 Percent Carbon

0 Ex. Res. Figure 6. Concentrations of unpaired electron spins in the Argonne Premium Coals and their extracts and residues on a dry-mineral-matter-free (dmmf) basis versus whole coal C content (rank). In order to make additivity comparisons, the spin concentrations in the fractions are given on the basis of the dmmf grams of original whole coal. The sums of the spins in the combined fractions are also shown. Spin concentrations are accurate to an estimated f5%.

Table 1. Concentrations of the Numbers of Spins per gram of Inertinite Found in the Narrow EPR Lines of the Whole

Argonne Premium Coals and Their Residues* sDins Der mam of inertinite x 10-19

coal whole coal* (dmmf) residue (dmmf) ~~~ ~ ~

wyociak Anderson Illinois No. 6 3.6 Blind Canyon Pittsburgh No. 8 Lewiston Stockton 6.5 Upper Freeport 10.5 Pocahontas No. 3 13.3

0.0044 0.19 0.17 0.36 0.11 0.064 0.47

a No appreciable narrow-line signal was found in the extracts. * From ref 13.

extraction was not quantitative, and these calculations could not be performed. The ratios of the sums of spins after extraction to the original numbers of spins, which range from 60 % to perhaps as little as 5 % in a generally decreasing trend with rank, point out that there are substantial losses of free radicals in the extraction and isolation processes. (The pulsed EPR method employed may overestimate the numbers of spins in the high spin concentration bituminous coals, LS and PO.l3) There are also substantial losses in the numbers of spins in the inertinite fractions, as shown in Table 1, which compares the numbers of spins in the whole coal and residues. The whole coal numbers of inertinite spins for the WA, BC, and PI coals were not measurable in our previous studyla and were not repeated in the present study under a set of conditions that is apparently more sensitive. There are usually more spins found in the extracts than are left behind in the residues, with the exceptions of the IL and LS coals, although the presence of mineral matter has been found to depress the number of spins measured in CW experiments (see Discussion). Practically no narrow line EPR signals are found in the extracts. The largest fraction of inertinite spins that survives is 5% in the IL coal and the smallest survival is as low as 0.6% in the UF coal, assuming that all of the inertinite maceral remains with the residue (i.e., negligible extraction occurs).

The EPR g values of the free radicals in the broad, non- inertinite, whole Argonne Premium coals and their extracts and residues are given in Table 2, together with the line

914 Energy & Fuels, Vol. 8, No. 4,1994 Doetschman et al.

Table 2. g Values and Line Widths of the Broad, Non-Inertinite EPR Line in the Whole Argonne Premium Coals and Their Extracts and Residues

g valueb mC (G) coal percent Ca (dmmf) extract residue whole coal extract residue wholed coal

Beulah Zap 74.0 2.0033 2.0039 2.0037 5.9 6.8 7.4 Wyodak Anderson 76.0 2.0037 2.0039 2.0037 5.9 5.8 7.9 Illinois No. 6 80.7 2.0033 2.0025 2.0034 6.4 6.1 6.0 Blind Canyon 81.3 2.0032 2.0037 2.0030 6.4 6.1 7.6 Pittsburgh No. 8 85.0 2.0033 2.0025 2.0027 5.9 6.2 4.7 Lewiston-Stockton 85.5 2.0032 2.0027 2.0029 6.6 6.2 4.8 Upper Freeport 88.1 2.0033 2.0033 2.0025 6.6 5.9 4.4 Pocahontas No. 3 91.8 2.0032 2.0032 2.0027 5.9 5.9 5.2

a Whole coal, rounded to nearest tenth of 1 ?4 . b Estimated uncertainties are O.OOO1. Measured from peak to trough of EPR derivative spectrum; 0.1-G estimated uncertainty. d From the selected values in ref 16.

Delta g values 1

0.5 0 0

x (d

2 0 0 0 c c(

8 -0.5

UF

-1 74 76 80.7 81.3 85 85.5 88.1 91.8

Percent Carbon m Extract m Residue

Figure 7. Changes in the ESR g values, Ag, of the Argonne Premium Coal extracts and residues from the g values of the corresponding whole coals are presented versus the C content of the whole coals. Estimated standard deviations are O.OO0 15.

widths, AH, of the broad, non-inertinite parts of the EPR lines of each. Figure 7 shows the differences between the g shift of the coal fractions and the g shifts of the whole coals. The g changes of the extracts tend to increase with whole coal rank from a substantially negative BZg change to positive values in the higher rank coals. On the other hand, there are significant g changes, mostly positive but also negative, in some of the residues. Some of the high rank coal residues have large positive g changes, except for the small negative values of PI and LS. The large negative g changes of the residue in the high ~ u l f u r ~ ~ J ~ IL coal is unique.

The EPR line width values, AH, in Table 2 exhibit extract values that range from less than or comparable to residue values at low rank to extract values comparable with or greater than residue values a t higher rank. The whole coal line widths range from values considerably greater than or comparable to the values of their fractions at lower rank to whole coal values that are markedly less than the values of the fractions at higher rank.

The EPR SLR 2'1 times are shown in Figure 8 for the whole coals and their extracts and residues. The non- inertinite and inertinite residues generally have SLR rates decreasing (2'1 increasing) with coal rank (from 25 to 30 ps at 75 % to >lo0 ps above 90 % C for the non-inertinites). The 2'1 of the residues are generally greater than or comparable with the extracts, a trend that goes with a fairly uniform extract 2'1 (30 f 10 ps) across the whole range of coal ranks. We note that there are individual differences in the SLR of the whole coals. The 2'1 of the UF (160 ps; 88% C), PO (130 ps; 92% C), LS (74 ps; 85.5% C), and IL (64 ps; 80.5% C) coals are considerably longer

200

150 n 0

7 - 100 t-.

w

% v

E 50

0

Spin-Lattice Relaxation

Broad Lines Narrow Lines UF n

1200

1000

800 5 600

400

200

0

h

C v

F:

74 76 81 81 85 86 88 92 90 90 91 92 Percent Carbon

Extract c] Residue Coal Figure 8. ESR spin-lattice relaxation times, TI (left-hand scale), of the broad, non-inertinite ESR lines in the Argonne Premium Coals and their extracts and residues. The data for the narrow, inertinite lines are also shown for the whole coals and their residues (right-hand scale). No appreciable narrow-line signal was found in the extracts. Data are shown versus C content (rank) of the whole coal or the inertinite in the whole coal. The largest relative uncertainty in the values presented here is 25 % and most are around 10%.

Table 3. Spin-Lattice Relaxation Times, TI, of the Narrow Inertinite EPR Lines in the Whole Argonne Premium Coals

and Their Residues

Tla ors) coals percent Cb residuec whole coald

Lewiston Stockton 90.2 720( 150) 500 Illinois No. 6 90.4 820(30) 390 Pittsburgh No. 8 91.2 lOlO(50) 480 Pocahontas No. 3 91.8 810(10) 490

a Standard deviations are given in parentheses. b Selected values in ref 16. C Traces of a narrow line signal are found in inappreciable amounts in the extract with identical 2'1. From ref 13.

than their extracts and residues while the PI (17 ps, 85 % C) and BC (15 ps, 81.5% C) are considerably less. We also note in Table 3 that the 2'1's of the inertinites in the residues (720-1010 ns) are longer than in the inertinites (390-400 ns) in the whole coals before extraction.

The EPR spin-spin relaxation time, T2, values in the broad, non-inertinite lines of the whole Argonne Premium Coals and their extracts and residues are presented in Table 4. The extract 2'2 values are greater than the whole coal with the possible exception of the WA case. The residue T2 values are all less than the extract 2'2 except for the UF case. The residue 2'2 values are in some cases greater and in other cases less than the values of the whole coal. We have shown that there is little correlation of spin-spin relaxation rate, 1/2'2, with the actual concentrations of spins in the extracts and residues, whereas a good


Table 4. EPR Spin-Spin Relaxation Times, T2, of the Non-Inertinite Macerals of the Whole Argonne Premium

Coals and of Their Extracts and Residues.

~ ~~

C d whole coal extract residue Beulah Zap Wyodak Anderson Illinois No. 6 Blind Canyon Pittsburgh No. 8 Lewiston Stockton Upper Freeport Pocahontas No. 3

1020(160) 1340(160) 464(35) 950( 100) 355.8(8.8) 1450(110) 390( 20) 507(54)

1590(45) 1140(110) 2840( 550) 1970(180) 3860(890) 1710(370) 1470(230) 1570(210)

355(14) 623(38) 1045( 80) 705(93) 1680( 260) 724(71) 24W( 240) 910(130)

O Standard deviations are given in parentheses.

Table 5;. Spin-Spin Relaxation Times, T2, of the Narrow Inertinite EPR Lines of the Whole Argonne Premium Coals

and Their Residues.

coal whole coal* residue Wyodak Anderson 490(10) Illinois No. 6 370(40) 297(3) Blind Canyon 290(40) Pocahontas No. 3 830(100) 626(5) Pittsburgh No. 8 358(1) Lewiston Stockton 370(50) 322(5) Upper Freeport 660(50) 420(20)

ONo appreciable narrow-line signal was found in the extracts. Standard deviations are given in parentheses. * From ref 13.

correspondence is observed13 for the whole coals. As will be discussed in the Discussion section, there 2'2 values may have been influenced by the presence of mineral matter in the whole coals and residues.

The EPR 2'2 times for the narrow, inertinite lines of the residue are compared with the whole coal values in Table 5. Based upon the whole coal data available from our previous study,13 there appears to be a uniform increase in residue intertinite 2'2 with whole coal inertinite 2'2. The residue inertinite 2'2 are all significantly reduced from the values in the whole coals, with the possible exception of the significance of the LS difference. NMR Experiments. The proton NMR SLR 2'1 time

measurements of the Argonne Premium Coals and their extracts and residues are shown in Figure 9a. The results are superficially similar to the EPR 2'1 results shown in Figure 8. The residue 2'1 values rise fairly uniformly with the rank of the whole coal except for the high BC residue 2'1 value. The whole coal and extract 2'1 values have overall rising trends with whole coal rank but with less uniformity. Like the EPR results, the proton NMR whole coal T<s are mostly greater than or comparable with the 2'1 of the extracts and residues. Unlike the EPR results, the extract proton NMR 2'1 are often longer than the residue 2'1. The data are plotted against the individual C contents12 of whole coal, extract, and residue in Figure 9b. It can be seen from the uniformity of the 2'1 dependence on C content that much of the variation between the 2'1 of a given whole coal and its extract and residue is simply a function of C content. The BC and UF extract 2'1 appear to be lower than the overall trend. (Interestingly, a presentation of the EPR 2'1 data versus individual C content similar to Figure 9b does not eliminate the disparities between the 2'1 of a given whole coal and its extract and residue observed in Figure 8.) We do not find the good correlation with the elemental oxygen content of the extracts and residues that has been observed by others% for whole coals. We also do not see a correlation with the

2

1.5 n

w F : 1 W

0.5

0

Energy & Fuels, Vol. 8, No. 4,1994

Spin-Lattice Relaxation

UF

BZ WA a

74.05 76.04 80.73 81.32 84.95 85.47 88.08 91.81 Percent Carbon

Extract m Residue Whole Coal

NMR T I vs Percent Carbon 2 I----

1.5 1 UF A

A

* A

Dl5

0 ' t *AA,B*:L--.-----

65 70 75 80 85 90 95 100 Percent Carbon

Extract e Residue A Whole Figure 9. Proton NMR spin-lattice relaxation times, 2'1, of the Argonne Premium Coals and their extracts and residues. Data at the top are shown versus C content (rank) of the whole coal. Data at the bottom are displayed, with only a few outlying points labeled, as a function of the individual C content of the whole coal or fraction. The largest uncertainty in the values presented here is 11 %.

numbers of electron spins, in agreement with previous findings.9

A sample of one of the whole coals (UF) was prepared by direct transfer from the glovebox in a sealed tube to the vacuum line with subsequent removal of water by heating and pumping procedures that are equivalent to the normal procedure before flame-sealing. This procedure eliminates any prolonged exposure to low levels of molecular oxygen that might have been involved with the transfer from the glovebox (after extraction in the cases of extract and residue) to the "vacuum oven", the vacuum drying in the "oven", the transfer back to the glovebox, and the transfer to the vacuum line for evacuation and sealing. No appreciable difference in the 2'1 of the sample was observed.

Discussion

The extraction yields generally conform to the classic shape of extraction yield versus C content that has been observed for a wide sampling of many coals by Iino e t aZ.,18 in which extraction yields increase to a maximum at 85-87 % C and drop rapidly above this rank. It is a t this range of C content that the intensity of the aromatic proton NMR region of the extract maximizes and begins to drop off, presumably as a result of the increasing polycondensation of aromatic rings and loss of H. Heteroatom content also tends to drop off more rapidly with rank above 87 % C.12 As the C content of the whole coal decreases below 85% C, we observe methyl and methylene CH stretching intensity ratios in the IR of the extracts that are indicative

916 Energy & Fuels, Vol. 8, No. 4,1994 Doetschman et al.

rrprr M~Ll#nkAumSC b ” I C - I t r r m M m t m C O l h l Hd.m.bmpool

w--- Lm P w m i c . P ~ + ” c W W P a S m n a m- Ww-0 F m Y h P t S p k l a o FMIh-spdOS FW-SPQI

KBoldc&s%r*ad n e a 8 “ M * a C S p * * P q m m M d cu&r8ycvcmulned

L a - --On S i n m r f i r t l c C t U r a n d A l p h ( c a m L W 8 .

Figure 10. Schematic representation of the variation of the whole coal chemical characteristics that are consistent with the results of this study. The long saw-tooth lines represent covalently or H-bond cross-linked polymer. The short saw-tooth lines represent aliphatic hydrocarbons.

of increasingly longer aliphatic chain structures, which is, of course, being accompanied by an increasing heteroatom content.12 There is a well-known propensity of coals with decreasing rank to retain increasing amounts of water or pyridine solvent, like the similar trend demonstrated here for NMP, probably through acid-base interactions or H-bonding.16 Given the natural, biological origin of coals, together with these facts, it seems likely that, from 85% C downward, the coals become increasingly polymeric, resembling the biopolymers from which they derive in regard to H-bond cross-linkage. Figure 10 shows, sche- matically, the variations in coal structure and properties from low to high rank.

Thus, the amount of extractable molecular species increases with rank to 85-87 % as the polymeric, H-bonded macromolecular network breaks up. Above 85-87 % C the extractable, largely aromatic molecular species have been increasingly transformed with increasing rank into an increasingly polycondensed, insoluble, macromolecular matrix. This picture is also consistent with the elemental analyses of the extracts and residues from the Argonne coals.12 Except for the BZ lignite, the C content of the soluble, more aromatic extract is greater than the C content of the more aliphatic, insoluble, polymeric residue below 87 5% C.12 The situation reverses for the PO coal,12 in which the insoluble, polycondensed, aromatic residue is richer in C and the soluble, more aliphatic extract is poorer in C. Likewise, except for the BZ lignite, the heteroatom content of the soluble, more aromatic extract is less than the heteroatom content of the more aliphatic, Lewis and Bronsted acid-rich, polymeric residue below 87 5% C.12 The situation reverses for the PO coal, in which the insoluble, polycondensed, aromatic residue is poorer in heteroatoms and the soluble, more aliphatic and acid-functionality- rich extract is richer in heteroatoms.12 The BZ lignite is probably an exception because of the small degree of coalification of the organic matter. What very little material that can be extracted from BZ, because it is almost entirely polymeric, has a composition almost identical to the residue.12 See Figures 10 and 11 which summarize the nature of the extraction process for low-to-medium rank bituminous coals.

The increases in total coal weight and the increases in elemental N upon extraction indicate that solvent is being retained in spite of the rigorous vacuum heating to constant

weight under strictly oxygen-free conditions. The extract IR spectra, the extract solution proton NMR spectra, the proton NMR spectra of the acetoned6 washings, and the proton NMR spectra of the solid extract and residue all show retention of NMP. The absence of any detectable CS2 in the extract or in the acetone wash 13C NMR spectra and the well-behaved elemental S content, together with the rigor of the “vacuum oven” conditions of solvent removal relative to the high CS2 vapor pressure and low boiling point, all indicate that it is only NMP that is being retained.

The total amount by weight of NMP that is retained in the extraction shown in Figure 2 appears to rise generally with heteroatom content in the whole coal, perhaps with LS and BC coals dropping below the trend. (We have chosen to associate heteroatom content with the combined 0 and S content because, while the actual coal N content is generally small and not highly variable, the measured elemental N contents of the extracts and residues are severely complicated by NMP retention.) The amount of NMP that cannot be washed from the residues increases roughly with heteroatom content of the residue, suggesting not only that NMP retention increases with heteroatom content but also that it is more tightly bound. The fact that the amount of NMP retained in the extract so very closely follows the amount that can be washed from the residue (except for the PO fractions), as shown in Figure 3, suggests that the entities to which NMP binds tightly are largely absent in the extract. More specific evidence about the tightness of NMP binding comes from the solid proton NMR studies in which it was found that only a fraction of the total NMP retained had rotational correlation times short enough (1.35 X 10-6 s) to observe the NMR peak of NMP. This fraction drops precipitously with heteroatom content, or in other words, the fraction of highly immobilized NMP increases dramatically with heteroatom content. These results appear to be consistent with the H-bonding or acid-base immobilization mechanism that has been proposed23 for pyridine-coal interactions. The unusually large amount of solid NMR- observable NMP in the PO residue seen in Figure 5 may have to do with a substantial depletion of the heteroatom- containing sites to which NMP could tightly bind and which would result in the disappearance of NMP from the NMR line. Even the PO extract, which has almost twice the combined 0 and S content12 found in the PO residue, has enough tight binding sites to suppress the observable NMP NMR considerably.

It is evident from the data presented in Figure 6 that there is a substantial loss of free radicals in the extraction process, carried out in the strict absence of oxygen. The survival of both extract and residue free radicals decreases with increasing rank. This suggests that it may be the more aromatic free radical molecules that fail to survive. It is well established and confirmed here, to some extent by our IR and NMR characterization, that the aromaticity of the coals, extracts, and residues increases with coal rank. For example, see the fraction of aromatic carbons given in ref 29. Also, generally more free radicals are extracted than remain in the residue, as can be seen in Figure 6, except for the IL and LS coals. In many coals, the ratio of extract radicals to residue radicals exceeds the ratio (by weight) of extract to residue. If the initially extracted radicals, like extract material in general, tend to be more aromatic than the residue material (with the possible exception of the PO extract), one of two outcomes could

A Study of Coal Extraction Energy & Fuels, Vol. 8, No. 4,1994 917

numbers of spins (see Figure 11). The downturn in g change for the PO extraction products may be related to decreasing numbers of heteroatom-containing radicals present in the coal. The negative BZ extract g change may represent a preference for the extraction of more aromatic, heteroatom-poor free radicals from an otherwise more tightly H-bond cross-linked, heteroatom-rich, macromolecular matrix. The anomalous behavior in the IL coal extraction that leaves behind mostly heteroatom-poor radicals and extracts relatively few, but more representative heteroatom-rich, free radicals is puzzling. It is possible that the high S content of the IL coal is related to this anomaly. Elemental analysis of the IL residue shows a decrease in S content relative to the whole coal S content, whereas all other residues that we spot-checked showed a relative increase.

The EPR SLR rates (2'1-9, shown as 2'1 in Figure 8, do not correlate well with either electron spin concentration or the relatively small variations in H content, as is required respectively by electron or nuclear intermolecular dipolar interactions. Therefore, one concludes that modulation of intramolecular interactions, such as the Zeeman or molecular electron-proton dipolar interactions by flexing of the radicals in their sites in the coal matrix, is the operative SLR mechanism. Since the g anisotropies in aromatic hydrocarbon radicals are small, one would expect that only the heteroatom-containing free radicals would be influenced by this mechanism. However, there appears to be no strong correlation of SLR rate (Tl-l), shown as 2'1 in Figure 8, with heteroatom content (e.g., the heteroatom contents of BZ and its fractions are ca. 20 times greater than those of UF but the 2'1-l are only 2-4 times greater). The only remaining and often considerably stronger mechanism at the fields employed in these experiments is the modulation of molecular electron- proton dipolar interactions,13 and we will attempt to show that this is reasonable in terms of the observed SLR rates in Figure 8.

SLR occurs as a consequence of components of time varying local magnetic fields at the Larmor frequency of the electron spins that induce transitions, much as does the applied microwave magnetic field.% Generally two distinct, but not easily experimentally distinguishable, factors contribute to the amplitude of the local magnetic field induced on the electron spin by the varying electron- proton dipolar interaction of the flexing of the free radical in ita matrix. One factor is the actual angular amplitude of the motion of the free radical at the Larmor frequency in ita site in the solid matrix. The other factor is the magnitude of the anisotropy of the electron-proton dipole interaction in the free radical molecule. The latter is a molecular property of the free radical and scales as the spin density mainly at the CH groups in odd-alternate, polycondensed ring systems. We were not successful in separating these two factors by employing observed line width as a measure of the dipolar anisotropy, perhaps because there remains a distribution of g values that also contributes to the line width.

In order to learn whether the variation in spin density or the flexing of the free radical in the matrix is the predominant cause of 2'1 variation, we attempt to estimate the degree of variation over the range of Argonne whole coals. Argonne coal 13C solid-state NMR experiments by Solum et aZ.= yield the numbers of C atoms per poly-

Molecular Extraction 15-81%C

t I

I EmctMolecules 1 1 Residue Molecules 1 Aromatic Rich

Aliphatic & Hetero- Aromatic Poor

Aliphatic & Hetero-

Preferential Free Radical Reactions of

Extended Aromatics

E m c t Free Radicals

Aliphatic & Hetero-

Residue Free Radicals

Often Aliphatic & atom Enriched Heteroatom Enriched

Figure 11. Schematic representation of the chemical characteristics of the coal fractions and the free radicals in them that are obtained from the extraction of medium rank (bituminous) coals.

result. One is that these extracted radicals, being more aromatic, have a lower survival rate, and contrary to observations, the resulting, stable residue radicals predominate. Clearly then, in spite of the disadvantage of low survival, the free radicals are overwhelmingly in the aromatic material (see Figure 11). This trend is dramatically displayed in the well-known increase in spin concentration with rank, also shown for the Argonne coals in Figure 6. The evidence for the chemical lability of the aromatic free radicals appears to run counter to arguments in the literature for their exceptional stability, mentioned in the Introduction. However, there is growing evidence from published% and unpublished37 work of Larsen and co-workers that the coal materials with the most highly extended interacting networks of aromatic systems are, in a number of ways, the most chemically reactive. For example, it was shown that the free radical polymerization initiation by coals results in the preferential loss of the inertinite free radicals.%

The EPR line widths are in agreement with increasingly more highly polycondensed odd-alternate aromatic free radicals with increasing coal rank. The line widths are consistent with extract free radicals being more highly aromatic than residue free radicals at lower ranks crossing over to less highly aromatic extract free radicals than residue free radicals at higher rank.

There is an overall tendency for theg values of both the coal extracts and residues to exhibit increasingly positive changes from the whole coal values with increasing whole coal rank. The well-established correlation of positive g shifts with coal heteroatom content16J7 suggests that the products of extraction contain increasingly more heteroatom-rich free radicals than does the whole coal. This can also be explained by the low survival rate of the more aromatic, heteroatom-poor free radicals during extraction that seems to be called for to explain the trends in the

~~

(36) Flowers, 11, R. A.; Gebhard, L.; h e n , J. W.; Silbemagel, B. G.

(37) Proceedingsofthe1992and 1993D.O.E.UnivmityCoalReaezvch Energy Fueb 1992,6,455.

Contractors Meetings, Pittsburgh. (381 Solum, 'M. S.; Pugmire, R. J.; Grant, D. M. Energy Fuekr 1989,

3, 187.

918 Energy &Fuels, Vol. 8, No. 4, 1994

condensed aromatic ring cluster. The numbers range with increasing rank from about 9 for the BZ coal to a more- or-less level 14-16 C per cluster for the WA, BC, IL, PI, and LS coals and rise to 18 C at UF and to 20 C at the PO coal. One may roughly assume that the average aromatic, peripheral C spin density scales inversely with the number of C atoms in an odd-alternate free radical cluster. This would clearly be the case in the limit of large clusters. Thus, a drop in the average peripheral spin density of little more than a factor of 2 would be predicted across the range of coal ranks. There is presently no way to estimate the particular average peripheral spin densities in the coal extracts and residues or in the inertinite macerals. However, the small variations in the whole coal spin densities, which we can estimate, suggest that the free radical molecule flexing amplitude in the solid matrix is the main determinant of the TI values and their wide variations in Figure 8.

The non-inertinite and inertinite whole coal and residue free radicals display a generally increasing TI with coal rank (from 25 to 30 ps at 75 % C to as much as 150 ps above 90% C for the non-inertinites). This is consistent with both the decreasing CH spin densities and the decreasing flexing amplitude of increasingly large polycondensed ring systems. The 7'1 values of the more rigid residues are generally greater than or in some cases comparable with the T1 of the generally more flexible extract. This would be consistent with a generally more rigid macromolecular matrix free radical environment than the molecular environment in the extract. The relatively flat extract TI trend with rank is somewhat puzzling. I t would be consistent with the opposing trends with decreasing rank toward more rigid H-bonding or dipolar intermolecular binding forces but greater CH spin densities in the smaller molecular polycondensed free radical systems. Conversely, it would be consistent with the opposing trends with increasing rank toward less rigid van der Waals binding forces but smaller CH spin densities in larger molecular polycondensed free radical systems.

The NMR SLR results, shown in the two TI presenta- tions in Figure 9, exhibit a markedly monotonic increasing trend with C content of the individual coal or coal fraction. Since there is no obvious direct connection between TI and C content, we look for other potential causative factors that correlate strongly with C content, such as electron spin concentration, H content, or presence of molecular 02. However, TI fails to correlate very well with the first two, as previously obse r~ed .~ The failure to correlate with spin concentration can be seen, for example, as a failure of the T1 to follow the loss of electron spins that occurs upon extraction, shown in Figure 6. A correlation between NMR TI and coal 0 content in a study of a series of coals was the basis for an earlier conclusion29 that molecular 0 2

was a major source of proton NMR SLR. We have taken extraordinary pains to exclude 0 2 in the present work and have failed to find a good correlation between NMR TI and 0 content of the coals, extracts, and residues. We speculate that it is either the progressive loss of flexible or rotating, fast aliphatic chain protons with rank or the increase in rigidity of the increasingly polycondensed molecules whose aromatic protons begin to predominate with increasing rank. These questions could be answered more specifically with TI measurements of the separate functional groups of coals, extracts, and residues in magic- angle spinning solid-state NMR experiments. We believe that the rigidity of the matrix must also be playing some

Doetschman et al.

role in determining NMR SLR rates. The whole coal T1 would otherwise tend to be an average of extract and residue values, whereas they are observed to be greater than the average in all but one case.

It has been noted by Silbernagel et aLn, that CW EPR measurements of the numbers of spins per gram of organic matter in whole coals are, with few exceptions, increased by as much as a factor of 3 by demineralization of the whole coal. No similar systematic study of pulsed EPR measurements of numbers of spins in coalla has been done. If, as suggested by Silbernagel et ~ 1 . : ~ different relaxation times in the presence of mineral matter are the problem, the ability of pulsed EPR to measure these relaxation times should be employed. In order to check that present results are not being complicated by the presence of mineral matter, we have examined the effects of the HF/HC1 demineralization procedure of Silbernagel et ~ 1 . ~ 9 on several relevant samples. We examined whole, dried Pittsburgh No. 8 and Pocahontas No. 3 Argonne coals, which Silbernagel et ~ 1 . ~ ~ found to have the greatest increase in numbers of spins, and dried Ashland pitch and charcoal samples, which contain no mineral matter. The samples exhibited decreases of 5 1 4 % in the TI of the narrow line, the largest of which was observed in the mineral matter free charcoal. The samples showed T2 increases for the narrow line of 4-44 % , the largest of which was observed in the mineral matter free charcoal narrow line. Decreases of 27-53 5% were observed in the 2'1 of the broad line, the largest of which were in the coals. While we observed a 10% increase in the spin coherence decay time, TM, for the broad pitch line, 3449% decreases were observed in the coals.

Several conclusions may be drawn from these studies of the effects of the demineralization process. One is that the process used for demineralization does not have a negligible effect on the EPR relaxation times of free radicals in carbonaceous materials in the absence of mineral matter. The second is that the conclusion of Silbernagel et dS9 that the demineralization process causes changes in the spin relaxation rates is confirmed by these results, especially for the broad EPR line in coals. However, the broad-line coal TM are determined in large measure by instantaneous d i f f u s i ~ n , ~ ~ ? a while the low spin concentration pitch TM are Instantaneous diffusion depends on the spin concentration of the resonant ~pins.~2 Clearly more work needs to be done on this subject. In view of these findings, we have refrained from drawing conclusions about EPR TI and spin concentration differences of less than a factor of 2-3.

Conclusions The extraction study of the Argonne Premium Coals

with NMP-CS2 mixed solvent with proton NMR and EPR spectroscopy and relaxation leads to a self-consistent characterization of the molecules and free radicals extracted and the macromolecular material and free radicals left behind in the residue. An important part of this characterization is the NMP solvent shown to be retained in both extract and residue.

(39) Silbernagel, B. G.; Gerhard, L.; Flowere 11, R. A.; Larsen, J. W. Energy h e l s 1991,5,661.

(40) Thomann, H.; Silbernagel, G. B.; Jin, H.; Gebhard, L. A.;Tindall, P. Energy Fuels 1988,2, 339.

(41) Doetschman, D. C.; Maetaf,D.; Singer,L. L. J. Phys. Chem. 1988, 92, 3663.

(42) Salikhov, K. M.; Tsvetkov, Yu. D. In Time Domuin Electron Spin Resomame; Kevan, L., Schwartz, R. N., Ede.; Wiley: New York, 1979; Chapter 7.


Largely aromatic and heteroatom-poor molecular material is extracted from low-to-medium rank bituminous coals, which have a more aliphatic heteroatom-rich, H-bond cross-linked macromolecular matrix. As the aromaticity increases in both the coal molecules (extract) and macromolecules (residue) with rank and the coal materials become more heteroatom-poor, the proportion of molecular material increases, as evidenced by the increasing extraction yield. The high rank bituminous coal evidently begins to form a macromolecular network of interconnected polycondensed ring clusters, where the molecular species (extract) become less aromatic and more rich in heteroatoms than the macromolecular matrix (residue). This increasing polycondensation leads to a drop in the proportion of molecular species (extraction yield). The lignite behavior can be understood in terms of very low rank material that is almost entirely macromolecular and that has undergone minimal polycondensation.

The extraction of free radicals follows basically the same trend as the extraction of molecular species except for the complication of free radical reactions that are solvent- facilitated by the mobilization of molecular free radicals and the exposure of macromolecular free radicals. Results show that the free radicals in the more highly aromatic polycondensed material have a lower survival rate than other radicals. Thus, one finds that free radical survival drops with increasing rank and it is least in the inertinites. There is a consistently lower survival of the more aromatic, heteroatom-poor free radicals of the extract, with the possible exception of the high-sulfur, Illinois No. 6 coal. The character of residue free radical survival is more mixed but is more often like the pattern of extract survival than not.

The total NMP retention is found generally to increase with whole coal heteroatom content. Of this NMP, there appears to be two kinds: One weakly bound kind is found in the extracts and is found to be acetone-washable from the residues of the low-to-medium rank bituminous coals in amounts proportional to the extract. The second, more tightly bound kind of NMP cannot be acetone washed from the residue and is retained in amounts that increase with residue heteroatom content. Additional spectroscopic evidence for these two classes of NMP in the residue was found in the presence of a relatively narrow solid-state proton NMR peak of a relatively mobile NMP molecule whose intensity appears to correlate with the loosely bound NMP. The more tightly bound NMP was not observable.

The EPR SLR TI times are shown primarily to be a measure of the flexing amplitude of the free radical in its matrix at the Larmor frequency (9 GHz). There is a tendency for the rigidity of the residues and the whole coals to increase with the rank of the whole coal, although there are a number of individual exceptions. The vitrain material of the whole coals is usually more rigid than that of the residue, which in turn is usually more rigid than the extract. This seems to be consistent with a most highly cross-linked matrix in the whole coal. The extraction process tends to disrupt the cross-linking of the macromolecular matrix, perhaps by removal of some participat- ing molecular material, leaving a less rigid residue. The molecular, and usually more aromatic, heteroatom-poor, extract material has still less cross-linking ability and is most flexible.

The inertinite spin-lattice relaxation times of a few hundred nanoseconds are in the exchange-narrowed

Energy & Fuels, Vol. 8, No. 4, 1994 919

regime, where often T I - Tz. Similar situations exist in charcoal and graphitic solids, which have very highly extended polycondensed aromatic systems. The fact that the T I are longer in the residues than in the whole coals suggests that some of the less extended paramagnetic systems are effectively removed from the material. Whether this is from actual extraction or from selective chemical reaction is not clear from these results.

There are number of questions raised by the present work that require further research. First, the present research gives no direct indication of what kind(s) of free radical chemistry is (are) taking place during the extraction that reduces the total numbers of spins. Our conditions and results indicate that the chemistry in not oxidative and appears to favor more highly polycondensed aromatic systems. Much more work exploring the free radical reactivity of coal materials should be done, such as the work of Larsen and co-workers on free radical initiation of olefin p o l y m e r i ~ a t i o n . ~ ~ ~ ~ ~ Their work and our own may involve conditions in the original coals in which the radicals are kinetically stabilized. Presentation of solvent to the coal then reduces the stability of the radicals, which may subsequently be free to combine with one another, to disproportionate or to initiate olefin polymerization. The presence of these kinds of reactions appears to call the established view into question that these radicals are of the thermodynamically stable odd-alternate type. The possible role of the variation of molecular spin density distribution in determining EPR TI deserves further confirmation, especially from the 13C solid-state NMR determination of the average numbers of C atoms per polycondensed aromatic ring system in the extracts, residues, and macerals of a range of coals. This information would also be very useful in providing hard evidence on the trends that our results indicate for the degree of aromaticity and extent of aromatic polycondensation. We have found little EPR evidence for the obviously multiexponential relaxation phenomena that prevail in NMR studies of coal materials. While this probably means that molecular and macromolecular free radicals in the whole coal flex in consort with one another, it would be desirable to have a clearer molecular level picture of why there are classes of protons in coal materials undergoing nuclear relaxation a t distinguishable rates. The questions of whether these classes have to do with molecular and macromolecular constituents or whether they have to do with different functional groups urgently need a definitive answer that perhaps could be provided by further relaxation studies of resolved lines in magic angle spinning, solid-state NMR studies, such as those of Solum et a1.% Finally, more work is required on the mechanism of NMP retention. We have similar studies of swelled whole coals and reswelled residue under way. A quantitative magic angle spinning solid-state NMR study of the extracts and residues should be done in order to measure the amounts of loosely- and tightly-bound NMP alike.

Acknowledgment. We thank the U.S. Department of Energy for its financial support through Contract DE- FG-22-91PC91299and the NSF for its support of the ESR Facility Users Group at Columbia University where the pulsed EPR measurements were made. We thank Brett Pleune for his assistance with the dataanalysis. We thank Jing Wei for her prepration of the demineralized samples.

Documents

A Study of Coal Extraction with Electron Paramagnetic Resonance and Proton Nuclear Magnetic Resonance Relaxation Techniques