Characterization of the Soluble and Insoluble Fractions of Upper

Energy & Fuels 1997, 11, 495-501

495

Characterization of the Soluble and Insoluble Fractions of Upper Freeport Coal in NMP/CS2 and Pyridine Using Small Angle Neutron Scattering George D. Cody* Geophysical Laboratory, Carnegie Institution of Washington, 5251 Broad Branch Road, Washington, D.C. 20015

Marcus Obeng Chemistry Division, Argonne National Laboratory, Argonne, Illinois 60439

Pappannan Thiyagarajan* Intense Pulsed Neutron Source, Argonne National Laboratory, Argonne, Illinois 60439 Received October 21, 1996. Revised Manuscript Received December 16, 1996X

Small angle neutron scattering (SANS) was used to characterize the solution structure of coal extract/solvent systems as well as swelling-induced structural changes in the nonextractable residues derived from the Upper Freeport (UF), medium-volatile bituminous, coal. Studies were performed on the pyridine soluble (PyS) fraction dispersed in perdeuterated pyridine (dPy); the PyS fraction dispersed in the mixed solvent [dm, perdeuterated N-methylpyrrolidinone (NMP) and carbon disulfide, CS2], and the mixed solvent soluble fraction (MS) in dM. No coherent scattering was observed for the UF-MS/dM sample in the accessible Q range, indicating that the solution contains no large particles or aggregates. Strong coherent scattering was observed for the UF-PyS/dPy and UF-PyS/dM solutions, indicating the presence of large particles, presumably molecular aggregates. Comparison of the SANS behavior of the three samples suggests that there exist molecular components intrinsic to the pyridine insoluble NMP/CS2 soluble fraction that may play a significant role in the enhanced solvation of the mixed soluble fraction of the Upper Freeport coal in NMP/CS2 solvent. SANS analysis of the dry and pyridine (d5)-swollen UF-PyI and UF-MI residues revealed significant structure evident by power-law scattering behavior in the low Q region of both sets of samples. In the case of the UF-MI samples, significant scattering intensity was also observed in the intermediate Q region. Significant structural changes were observed in the UF-MI fractions upon swelling in pyridine, whereas minimal structural changes were observed in the case of UF-PyI sample during solvent swelling. These results reveal that the physical consequences of mixed solvent extraction are significantly different from those due to pyridine extraction.

Introduction Many of the current theories regarding the macromolecular structure of coals (type III kerogens) are based on the fact that there exists a physically definable limit in the extractability of soluble organics from coal in certain basic organic solvents, such as pyridine.1 The nonextractable residue is generally considered to be a cross-linked macromolecular network.2 The extractsolvent “solutions”, although acknowledged to be far from ideal (thermodynamically), are presumed to be “molecular solutions” in the sense that random mixing of the constituents occurs within a single solution phase.3 Recently, a number of investigations probing various physical aspects of solvent-coal interactions have yielded results that conflict with the established model of coal Abstract published in Advance ACS Abstracts, February 1, 1997. (1) van Krevelan. Coal, 3rd ed.; Elsevier: Amsterdam, 1993; p 979. (2) Larsen, J. W.; Green, T. K.; Kovac, J. J. Org. Chem. 1985, 50, 4729. (3) Painter, P. C.; Park, Y.; Coleman, M. M. Energy Fuels 1988, 2, 693-702. X

S0887-0624(96)00182-X CCC: $14.00

structure. In particular, studies using mixed solvent systems [most notably, an N-methyl-2-pyrrolidinone (NMP) mixture with CS2] have demonstrated increases in extractabilities of more than twice that by pyridine in the case of some coals.4 These results suggest that either there exists a set of noncovalent interactions unperturbed by pyridine but disrupted by the mixed solvents or that the mixed solvents are degrading the network and enhancing the extraction yields through some, as yet unknown, chemical reaction. Additional evidence that complicates the prevailing theory of coal is that “solutions” of coal extracts in pyridine are not true molecular solutions as described above, but rather a dispersion of aggregates of coal macromolecules that exist even at low solute concentrations.5 The presence of aggregated structures is contradictory to the hypothesis that pyridine is an exceptionally good solvent for coal in a thermodynamic sense. (4) Iino, M.; Takanohashi, T.; Ohsuga, H.; Toda, K. Fuel 1988, 67, 1639-1647. (5) Cody, G. D.; Thiyagarajan, P.; Botto, R. E.; Hunt, J. E.; Winans, R. E. Energy Fuels 1994, 8, 1370-1378.

© 1997 American Chemical Society

496 Energy & Fuels, Vol. 11, No. 2, 1997

Cody et al.

Table 1. Samples and Their Physical Characteristics APCS sample rank 1 3 c

mv HvC

% E% Ca % H % O % S (NMP/CS2)b % E(Pyr)c 86 78

4.7 5.0

8 14

2.3 4.8

56.0 NA

25.5 29.0

a Dry ash free basis. b Percent extractable (wt) in NMP/CS . 2 Percent extractable (wt) in pyridine.

Although there is no debate that pyridine has a strong interaction with acidic functional groups in coal (∼7 kcal/mol6 ), in the case of coal extracts, pyridine does not appear to have exceptional solvating capabilities beyond these specific interactions.5 The characterization of solution structures of dilute mixtures of macromolecules in various solvents is readily attainable using small angle neutron scattering (SANS).5,7-17 The theory of small angle scattering is well established, and the method has been used by many researchers for the purpose of elucidating the configurational state of polymers in solution,7 the structure of asphaltenes in apolar solvents,8-13 and the study of proteins14 as well as dilute coal extract “solutions”.5,15,16 Neutron scattering is particularly useful in revealing the spatial characteristics of structures composed of low atomic number (Z) elements. This is due to the fact that neutrons, uncharged and with spin (I) ) 1/2, interact with a given element through their respective angular momenta. The magnitude of this interaction has no correlation with Z. X-rays, on the other hand, interact with the electronic shell surrounding the nuclei, and the magnitude of the coherent scattering amplitude varies as a strong function of Z. Although lighter elements are nearly “invisible” to X-rays, the unique nature of the interaction of the neutron with spin (I) > 0 nuclei results in the coherent scattering cross section of low Z elements being similar in magnitude to that of many of the higher Z elements;17 for example, deuterons scatter neutrons as efficiently as gold. In the following we report on SANS experiments designed to characterize the solution structure of mixed solvent extract “solutions” and pyridine extract “solutions” from the same coal. The goal of this study is to identify features or characteristics of the mixtures that may highlight the mechanism behind the exceptional solvating capabilities of the mixed solvents as compared with other polar solvents such as pyridine. For the present study we chose Upper Freeport coal which is ∼25% (wt) extractable in pyridine, but over 50% (wt) extractable in NMP/CS2 (see Table 1). In addition to characterizing the solution structure, SANS is also used to characterize the structure of the solvent insoluble fractions. Previous SANS analysis has (6) Arnett, E. M.; Mitchell, E. J.; Murty, T. S. R. J. Am. Chem. Soc. 1974, 96, 3875. (7) Bates, F. S. J. Appl. Crystallogr. 1988, 21, 681. (8) Herzog, P.; Tchoubar, D.; Espinat, D. Fuel 1988, 67, 245. (9) Ravey, J. C.; Ducoret, C.; Espinat, D. Fuel 1988, 67, 1560. (10) Overfield, R. E.; Sheu, E. Y.; Sinha, S. K.; Liang, K. S. Fuel Sci. Technol. Int. 1989, 7, 611. (11) Senglet, N.; Williams, C.; Faure, D.; Des Courieres, T.; Guilard, R. Fuel 1990, 69, 72. (12) Storm, D. A.; Sheu, E. Y.; DeTar, M. M. Fuel 1993, 72, 977. (13) Thiyagarajan, P.; Hunt, J. E.; Winans, R. E.; Anderson, K. B.; Miller, J. T. Energy Fuels 1995, 9, 829. (14) Stuhrmann, H. B.; Miller, A. J. Appl. Crystallogr. 1978, 11, 325. (15) Ho, B.; Briggs, D. E. Colloids Surf. 1982, 4, 285. (16) Triolo, R.; Child, H. R. Fuel 1984, 63, 274. (17) Bacon, G. E. In Neutron Diffraction; Clarendon Press: Oxford, U.K., 1975; pp 38-40.

shown that swelling in pyridine changed the intermediate range structure of a bituminous coal.18 In the present case we are interested in detecting any structural changes resulting from extensive extraction and swelling. The physical-structural consequences related to such extensive mass losses are not known; for example, does the extraction of such material lead to the formation of void space? In regard to the physicalstructural changes that occur with solvent swelling, it has not been determined whether the solvent-induced dilation is homogeneous and affine or, at the ultrafine scale, heterogeneous. Heterogeneous dilation would result in the formation of solvent-rich and coal-rich regions which may be detectable using SANS. Answers to these questions may provide information on network flexibility and topology of the insoluble fractions of the Upper Freeport coal. Experimental Section Solvent Extraction. Extraction of the Upper Freeport (UF) coal (APCS No. 1) with NMP/CS2 follows the procedure of Iino et al.4 with minor modifications. Approximately 5 g of 100 mesh APCS No. 1 was stirred in 150 mL of a 1:1 mixture (by volume) of N-methyl-2-pyrrolidinone (NMP) and carbon disulfide (CS2) (mixed solvent, MS) for approximately 6 h. The solvent extract was decanted and then centrifuged up to 14 000 rpm to sediment out any entrained solids. In the case of pyridine extraction, approximately 5 g of 100 mesh coal was refluxed with hot pyridine in a Soxhlet extraction apparatus for >24 h. In both cases the extract was separated from the solvent using roto-evaporation. The extractability of the UF coal in the mixed solvent is over twice that in pyridine as is presented in Table 1. In the case of the solvent insoluble fractions, hand-picked vitrain with extremely low mineral content (determined using reflected light microscopy) was selected to obtain physically homogeneous specimens for SANS analysis. Throughout this paper, the various soluble and insoluble fractions of the UF coal (APCS No. 1) are designated as follows: the various extracts in their respective organic solvent are UF-PyS/dPy, UF-PyS/dM, and UF-MS/dM, which correspond to pyridine solubles (PyS) from UF in pyridine-d, pyridine solubles (PyS) from UF in NMP/CS2-d, and mixed solvent (NMP/CS2) solubles (MS) from UF in NMP/CS2-d, respectively. The insoluble residues are designated UF-PyI/ dPy and UF-MI/dPy, corresponding to the pyridine insoluble fractions (PyI) of the UF coal swollen in pyridine-d (dPy) and mixed solvent insoluble fraction (MI) of the UF coal in pyridine-d, respectively. SANS analyses of the dry insoluble residues are likewise designated UF-PyI/dry and UF-MI/dry. SANS. Solutions of the extracts (5 wt %) were prepared by mixing dry extract with either of the perdeuterated solvents. This concentration was selected to yield sufficient scattering intensity without introducing scattering intensity due to interparticle interactions. Previous work on the SANS behavior of coal extract solutions in pyridine exhibited no dependence on scattering behavior with solvent concentration (at 1, 5, and 10 wt %) other than a linear increase in incoherent and coherent scattering intensity with increasing weight percent of extract.5 Solutions were contained in Suprasil cylindrical cells with a 2 mm path length (volume ) 0.7 mL) for the SANS analysis. SANS data were measured at the Intense Pulsed Neutron Source (IPNS) of Argonne National Laboratory, using the Small Angle Diffractometer (SAD). This instrument uses pulsed neutrons derived from spallation with wavelengths in the range of 0.5-14 Å and a fixed sample-to-detector distance (18) Winans, R. E.; Thiyagarajan, P. Energy Fuels 1988, 2, 356.

Characterization of Upper Freeport Coal

Energy & Fuels, Vol. 11, No. 2, 1997 497

of 1.54 m. The scattered neutrons are measured using a 64 × 64 array of position sensitive, gas filled, 20 × 20 cm2, proportional counters with the wavelengths measured by timeof-flight by binning the pulse to 67 constant ∆t/t ) 0.05 time channels. The size range in a SANS experiment is constrained by both the geometry of the instrument and the wavelength of the neutrons which determine the working range of momentum transfer Q

Q ) 4πλ-1 sin θ

(1)

where θ is half the Bragg scattering angle and λ is the wavelength of the neutrons. Given the characteristics of the SAD19 at the IPNS, useful SANS data in the Q range of 0.0060.25 Å-1 can be obtained in a single measurement. The reduced data for each sample are corrected for the backgrounds from the instrument, the Suprasil cell, and the solvent as well as for detector nonlinearity. Data are presented on an absolute scale by using the known scattering cross section of a silica gel sample. The absolute cross section for this sample has been measured with the SANS instrument at ORNL. Standard Guinier analysis in the region of QRg < 1.0 can be used to extract the radius of gyration, Rg, and I(0) values by using the equation

I(Q) ) I(0) exp(-Q2Rg2/3)

Figure 1. Absolute differential SANS cross section of APCS No. 1 pyridine extract 5% (wt) dispersed in pyridine-d5 (0), APCS No. 1 pyridine extract 5% (wt) in NMP/CS2 (b), and APCS No. 1 NMP/CS2 extract 5% (wt) in NMP-D9/CS2 (O).

(2)

Rg is the root-mean-squared distance of all of the atoms from the centroid of neutron scattering length density of the particle, and I(0) is the absolute scattering cross section at Q ) 0, which is defined as follows:

I(0) ) Np(F - Fs)2V2

(3)

Here Np corresponds to the number density of particles, V is the volume of the particle, and F and Fs are the scattering length densities of the particle and the solvent, respectively. The absolute scattering data of the silica gel standard give Rg ) 44.7 ( 0.2 Å and I(0) ) 70 cm-1. The magnitudes of Rg and I(0) for this sample are routinely measured in the same Q region and are used to determine scale factors to place the scattering data on an absolute scale in units of cm-1.

Results SANS of Soluble Fractions in Organic Solvents. The SANS data for UF-PyS/dPy, UF-MS/dPy, and UFMS/dM solutions on an absolute scale are presented as the intensity of the scattered neutrons I(Q) versus the magnitude of the momentum transfer (Q) in Figure 1. The solution of mixed solvent solubles in NMP/CS2 (UFMS/dM) exhibits virtually no coherent scattering intensity across the accessible Q range of the SAD, suggesting that there is no aggregation. It is important to note that the measured scattering intensity [I(Q)] from any SANS experiment is always the sum of both coherent and incoherent components.17,20 The flat scattering curve for UF-MS/dM, therefore, is mainly due to the strong incoherent scattering from extract-bound protons. The level of incoherent scattering observed for the various samples at the high Q region is useful to assess the concentration of hydrogen atoms within each of the samples. The incoherent scattering must be subtracted (19) Thiyagarajan, P.; Epperson, J. E.; Crawford, R. K.; Carpenter, J. M.; Klippert, P. E.; Wosniak, D. G. J. Appl. Crystallogr. 1997, in press. (20) Feigin, L. A.; Svergun, D. I. Structure Analysis by Small Angle X-ray and Neutron Scattering; Plenum Press: New York, 1987.

Figure 2. Guinier plot for APCS No. 1 pyridine extract 5% (wt) dispersed in pyridine-d5 (O) and APCS No. 1 pyridine extract 5% (wt) in NMP-D9/CS2 (b).

prior to data analysis, and such a subtraction has been done for all of the solution data that exhibited coherent scattering. The significant scattering intensity at low Q for both UF-PyS/dPy and UF-PyS/dM solutions (Figure 1) indicates the presence of relatively large particles. Similar scattering behavior was observed in a mixture of APCS No. 3 (Illinois No. 6) pyridine extract in pyridine-d (ILPyS/dPy).5 Figure 2 shows the Guinier plots for the UFPyS/dPy and UF-PyS/dM solutions which have particles whose Rg values are 37.1 ( 1.3 and 54.4 ( 3.0 Å, respectively. The particles in these solutions are aggregates as shown earlier for IL-PyS/dPy.5 The larger aggregates observed in the UF-PyS/dM solution may be due to poor solubility of the coal extracts in this solvent. The I(0) values determined from the Guinier analysis are 0.54 ( 0.01 and 0.43 ( 0.02 cm-1, respectively, for UF-PyS/dPy and UF-PyS/dM solutions. In principle, I(0) values should be larger for larger particles, when all other factors are equal (eq 2). However, the I(0) is smaller for the UF-PyS/dM solution even though it contains larger particles, as revealed by its Rg value. This is due to the lesser contrast provided by deuterated NMP/CS2 solution when compared to C5D5N for the neutron scattering of coal extracts. Our contrast variation study (data not shown) indicates that the scattering length density (F) of aggregates of coal extracts from APCS No. 3 is 0.428E10 cm-2. Thus, the contrast factor (F - Fs)2, included in eq 3, for the aggregates will be

498 Energy & Fuels, Vol. 11, No. 2, 1997

Figure 3. Absolute differential SANS cross section of APCS No. 1 pyridine extract 5% (wt) dispersed in pyridine-d5 (b) and APCS No. 3 pyridine extract 5% (wt) dispersed in pyridine-d5 (O).

Figure 4. Coherent SANS data of UF-MI/dry (b) and UFMI/dPy (O) samples. The power-law exponents are -3.1 ( 0.07 and -1.98 ( 0.09 for the UF-MI/dry and UF-MI/dPy samples, respectively.

smaller in 1:1 deuterated NMP/CS2 (Fs ) 3.787E10 cm-2) than in deuterated pyridine (Fs ) 5.71E10 cm-2). To assess the difference in the behavior of pyridineextracted species from UF and Illinois No. 6 coals, we plotted their absolute differential scattering cross sections in Figure 3. Remarkably, the scattering curves for these coals are virtually identical across the entire Q range. This strongly suggests that, in both the ILPyS/dPy and UF-PyS/dPy samples, the molecular aggregates are nearly identical in their chemical composition, size, shape, and number density. It must be noted that earlier experiments focusing on the pyridine extracts of lower rank coals exhibited radically different behaviors among the various samples.5 Thus, such similar SANS behavior among samples is hardly anticipated and should not be considered general behavior for pyridine extracts derived from different ranks. SANS of Insoluble Fractions in C5D5N. The coherent SANS data for UF-MI/dry and UF-MI/dPy solution are shown in Figure 4. The two data sets are remarkably different across the entire Q range. At low Q both exhibit power-law scattering behavior with slopes of -3.1 ( 0.07 and -1.98 ( 0.09 for the UF-MI/ dry and UF-MI/dPy, respectively. Power-law scattering indicates that there exists a power law in size distribution of the scattering entities. Such scattering behavior clearly precludes the use of the Guinier analysis (eq 2) to interpret the data. Recent theories regarding the physics of power-law scattering behavior exhibited by

Cody et al.

Figure 5. Coherent SANS data of UF-PyI/dry (b) and UFPyI/dPy (O) samples. The power-law exponents are -3.34 ( 0.15 and -3.10 ( 0.15 for UF-PyI/dry and UF-PyI/dPy, respectively.

fractal materials can be used to gain insight into the physical structure of these samples. For example, the larger power-law exponent of the dry powder suggests that the particles or voids responsible for the scattering are sufficiently large that scattering in the low Q range detects only their surface properties.21 The smaller exponent may be interpreted to result from an open structure exhibiting a mass fractal dependence on the density-density correlation.22-24 The coherent SANS data for the UF-PyI/dry and UFPyI/dPy samples are presented in Figure 5. The scattering behaviors of both samples are remarkably similar, indicating that in this case swelling has minimal effect on the coherent scattering. For both UF-PyI samples the scattering intensity is also characterized by a power-law relationship between I(Q) and Q at low Q with exponents of -3.34 ( 0.15 and -3.1 ( 0.15, for the dry and solvent-immersed powder samples, respectively. The mixed solvent insoluble fractions also exhibit scattering intensity in the intermediate Q range. To highlight the individual contributions from the powerlaw and intermediate Q scattering, one can subtract the power-law scattering, I(Q)P-L, from the total scattering intensity, I(Q)total, yielding a residual scattering intensity, I(Q)res. Such a separation is valid if the total scattering intensity is the result of two independent subsystems that are not spatially correlated. For example, a dilute solution of rod-shaped and spherical particles will exhibit coherent scattering intensity, which is the sum of the independent coherent scattering intensity derived from each ensemble of particles. Of course, in the present case, if the particles responsible for the power-law scattering are the same as those responsible for the scattering intensity in the intermediate Q region, then such a separation is not valid. For the moment, therefore, it will be assumed that I(Q)res results from scattering from a system of densely packed small clusters (particles) or voids that are independent of the particles that are responsible for the power-law scattering. (21) Guinier, A. X-ray Diffraction in Crystals, Imperfect Crystals, and Amorphous Bodies; W. H. Freeman: San Francisco, 1963; 378 pp. (22) Feder, J. Fractals; Plenum Press: New York, 1988. (23) Freltoft, T.; Kjems, J. K.; Sinha, S. K. Phys. Rev. B 1986, 33, 269. (24) McMahon, P.; Snook, I. J. Chem. Phys. 1996, 105, 2223.

Characterization of Upper Freeport Coal

I(Q)res ) I(Q)total - I(Q)P-L

Energy & Fuels, Vol. 11, No. 2, 1997 499

(4)

The large difference observed in the scattering curves of the UF-MI/dry and UF-MI/dPy samples (Figure 4) and subtle differences observed in the case of the UFPyI/dry and UF-PyI/dPy samples (Figure 5) are due to the structural changes in the insoluble residues accompanying solvent-swelling. The magnitudes of the differences between the SANS behavior of the respective dry and swollen residues presumably correlate with the degree of swelling, a conclusion supported by comparing the equilibrium swelling ratio of each residue in pyridine. The macroscopic equilibrium swelling ratio, R, of UF-MI swollen in pyridine is equal to 1.7, while UFPyI swollen in pyridine exhibits a ratio R ) 1.1 (where R ) Vs/Vdry; Vs is the swollen volume and Vdry is the dry volume, measured on coal powders in calibrated cylindrical tubes). Analysis of the residual scattering, I(Q)res, also reveals significant physical differences between the respective insoluble fractions. Figure 6 presents the smoothed residual scattering, I(Q)res, for UF-MI/dry and UF-MI/ dPy, while Figure 7 shows similar data for UF-PyI/dry and UF-PyI/dPy. Significant differences in the residual scattering are observed in the case of mixed solvent insolubles (UF-MI) in its dry and pyridine-immersed states (Figure 6). In the case of dry UF-MI, I(Q)res has the form of a relatively narrow peak near 0.05 Å-1; following immersion in pyridine, however, significant residual scattering intensity appears across the mid Q range, peaking near 0.1 Å-1. The intensity of I(Q)res is minimal for both UF-PyI samples (Figure 7) as would have been anticipated from observation of I(Q)total (Figure 5), which clearly shows that virtually all of the coherent scattering exhibits power-law behavior.

Figure 6. Residual scattering after subtraction of the powerlaw scattering, I(Q)res, for UF-MI/dPy (0) and UF-MI/dry (9). The lines represent smooth fits to the data by using KaleidaGraph for UF-MI/dPy (solid line) and UF-MI/dry (broken line).

Discussion

Figure 7. Residual scattering after subtraction of the powerlaw scattering, I(Q)res, for UF-PI/dPy (0) and UF-PI/dry (9). The lines represent smoothed fits through the data for UFPI/dPy (solid line) and UF-PI/dry (broken line).

The SANS studies of both soluble and insoluble fractions provide complementary insight into the nature of solvent-extract and solvent-residue interactions and their relationship to the macromolecular structure of coals. In the case of UF-MS/dM solution (Figure 1) the scattering intensity observed is totally incoherent, indicating that the majority of particles in the “solution” are relatively small. The lack of coherent scattering in the UF-MS/dM solution was not anticipated. In previous SANS studies of coal extract solutions, large particles were observed.5,16 Mass spectrometric analysis of the coal extracts, however, has revealed a numberaverage molecular weight in the 400-500 amu range,5 suggesting that the large particles detected via SANS were in fact aggregates. In the present case, laser desorption mass spectrometry of the UF-MS/dM has also shown that the number-average molecular weight of such extracts is below 500 amu.25 SANS analysis of the UF-MS/dM solution, therefore, reveals a welldispersed “molecular” solution of small particles. The fact that aggregation does not occur in UF-MS in NMP/ CS2 indicates that the mixed solvent solvates the extractable molecules much more effectively than pyridine, a result which is clearly consistent with the observation of overall enhanced extractability.

Given these conclusions, it is somewhat surprising that the pyridine solubles (UF-PyS), when resolubilized in NMP/CS2, exhibit a large coherent scattering at low Q (Figure 1). Indeed, the largest particles observed in the present study were found in the UF-PyS fraction dispersed in NMP/CS2 (Figure 2). This suggests that in the case of UF-PyS fraction, NMP/CS2 is a less effective solvent than pyridine. It has been previously established that the UF/PyS fraction constitutes a subset of the UF/MS fraction26 (where UF/PyS ∼ 0.5 UF-MS/MS, by weight). Thus, the question arises as to why large particles are observed in the UF/PyS solution but not in the UF/MS solution when solubilized in NMP/CS2 . A plausible explanation is that there exist molecular components in the total UF/MS fraction, but not in the UF/PyS fraction, which work in combination with NMP/CS2 to impart good solvation of the extract. Such an explanation is supported by the following observations. Iino and colleagues have observed that mixed solvent solubles (MS), when subsequently extracted with pyridine, leave a residue (PyI-MS) which is only partly soluble in NMP/CS2.26 They note, however, that when the PyS fraction is added back to the PyI-MS fraction, the total (MS fraction) retains its high

(25) Hunt, J. E.; Green, T. C. Unpublished results.

(26) Iino, M.; Takanohashi, T.; Obara, S.; Tsueta, H.; Sanokawa, Y. Fuel 1989, 68, 1588.

500 Energy & Fuels, Vol. 11, No. 2, 1997

Cody et al.

solubility in NMP/CS2. In the present study we observe aggregation within the PyS fraction. The cause of both the reduction in solubility of the PyI-MS fraction26 and the reduced solvation observed within the PyS fraction in NMP/CS2 using SANS is presumably related and indicative of a system where molecular components in MS fraction work in combination with NMP/CS2 to impart high solubility and good solvation. Partitioning of such molecular constituents within either the PyS or PyI-MS fraction evidently leads to a reduction in the solvating capabilities of NMP/CS2 on each fraction. Our earlier study5 showed that the pyridine solubles from coals of different rank behave differently. The nearly identical absolute differential scattering cross sections of the pyridine solubles from APCS No. 1 and APCS No. 3 are noteworthy (Figure 3). In our view this is a significant result as the system of molecules forming large aggregates in these two solutions should be chemically very similar as the size, shape, and number density of the particles in these two systems are nearly identical. It should be noted that this similarity is the exception; that is, such similarity between other coal extract solutions is not observed.5 The swelling and SANS behavior of the insoluble fractions is similarly illuminating (Figures 4 and 5). It is clear that significant structural changes are related to the substantial extraction of material with NMP/CS2. To fully appreciate these changes, we need to identify the structure responsible for the coherent scattering in these SANS experiments. In the data analysis described above, we chose to separate the total scattering intensity into contributions from I(Q)P-L and I(Q)res, with the assumption that the two components result from different physical entities and are not spatially correlated. Focusing first on I(Q)P-L, we note that power-law scattering is evident in each residue sample. The significance of the power-law exponent’s magnitude on the physical interpretation of the scattering behavior warrants some discussion. It has been previously suggested that the scattering behavior at low Q of lowrank coals, e.g. lignites, results from the presence of large particles (or voids) with topologically rough or selfsimilar (fractal) surfaces.27,28 An early theory derived by Bale and Schmidt27 to explain such scattering behavior predicts that the scattering intensity, I(Q), will exhibit a power-law relationship with Q given as

I(Q) ∼ Q-(6-D)

(5)

where D is the dimensionality of the surface; a slope of -4, therefore, corresponds to a smooth surface, and a slope of -3 corresponds to an essentially infinitely rough, fractal surface. Interested readers are directed to excellent papers by Wong and Bray29 and Sinha et al.,30 for discussions of the theory of surface scattering in considerable detail. In the case of the low-rank coal, Bale and Schmidt27 observed a power-law exponent of -3.44 for small angle X-ray scattering of the Buelah Zap lignite, which they interpreted to indicate a surface with a fractal dimen(27) Bale, H. D.; Schmidt, P. W. Phys. Rev. Lett. 1984, 53, 596. (28) Gethner, J. S. J. Appl. Phys. 1986, 59, 1068. (29) Wong, P-Z.; Bray, A. J. J. Appl. Crystallogr. 1988, 21, 786. (30) Sinha, S. K.; Sirota, E. B.; Garoff, S.; Stanley, H. B. Phys. Rev. B 1988, 38, 2297.

sion of 2.56. Indeed, an independent analysis of the topology of lower rank coal surfaces using scanning electron microscopy (SEM) provided qualitative support for their conclusions.31 If the scattering behavior of the UF-MI powder at low Q is due to surface roughness with self-similar characteristics, then a slope of -3.1 ( 0.07 would correspond to a fractal dimension of 2.9. In a recent scattering study, a Cu-catalyzed H2-treated sample of Loy Yang coal yielded an exponent of -3.14,24 corresponding to a surface fractal with d ) 2.86. In both cases, this interpretation indicates the presence of surfaces so topologically rough that they approach threedimensional objects. The scattering behavior of UF-MI swollen in C5D5N (Figure 4), on the other hand, exhibits a power-law exponent of -1.98 ( 0.09, which suggests a mass fractal structure.22-24 Mass fractal solids deviate from the relationship where mass is proportional to the cube of the particle radius, such that

M ∝ Rd

(6)

where d ) 3 for homogenous solids and d < 3 for mass fractal solids. An exponent of ∼2 indicates a very open structure; for example, silica smoke, a very low density material, exhibits mass fractal scattering characteristics with a magnitude of d ∼ 2.4.23 As stated above the power-law scattering exhibited by the UF-MI/dry (Figure 4) may result from large voids (r > 1000 Å) having extremely rough surfaces with surface fractal qualities in the SANS window. Given this interpretation, the following possible physical picture emerges for swelling-induced changes in the scattering behavior swelling of the UF-MI sample: swelling dilates the fractal surfaces, expanding their texturally complex surface into the void space, creating regions of low density, solvent dilated, mass fractal structure. In the case of the UF-PyI residues, immersion in pyridine results in only a minimal change in the void surface topology, although the direction of change is toward a rougher surface fractal (Figure 5). These differences in the magnitude of the swelling-induced changes between the two samples are consistent with the differences in their macroscopic swelling behavior. In addition to changes in the power-law scattering component with swelling, there were also changes in the residual scattering intensity of the UF-MI insoluble fraction. It is reasonable to conclude that the I(Q)res arises from correlations between regions of high scattering contrast derived initially from the extensive extraction. The shifts in I(Q)res with swelling reveal that the distribution in correlation lengths associated with these regions is modified by the presence of pyridine. The peaked nature of I(Q)res (Figure 6) is consistent with an interparticle scattering phenomenon, i.e., as opposed to scattering from individual particles. The regions of high contrast are presumably regions with a high density of microvoids. The size of the microvoids would necessarily have to be