Evaluation of Effect of Predrying on the PorousStructure of Water-Swollen Coal Based on the Freezing

Property of Pore Condensed Water

Koyo Norinaga,*,† Jun-ichiro Hayashi, Norihide Kudo, and Tadatoshi Chiba

Center for Advanced Research of Energy Technology (CARET), Hokkaido University N13,W8, Kita-ku, Sapporo 060-8628, Japan

Received February 16, 1999. Revised Manuscript Received May 17, 1999

The effect of the extent of predrying on the porous structure of water-swollen coal was examined.As-received Yallourn (YL), Beulah Zap (BZ), and Illinois #6 (IL) coals were used as the samples.They were predried at 303 K to different extents. Upon predrying, the coal samples releasedwater in the following order: free water identical to bulk water, bound water that froze at around226 K, and finally, nonfreezable water that never froze even at 123 K. Predried samples wereswollen in water at 303 K and subjected to 1H NMR measurements to characterize the freezingproperty of water retained in pores at a temperature range from 170 to 294 K. The total volumeof the pores filled with water (Vp) was defined as the amount of water that was not frozen at 260K. The removal of the nonfreezable water from YL coal by the predrying decreased the Vp of thewater-swollen coal, while removal of the other types of water had little effect on Vp. Completepredrying of the other coals also reduced Vp, but to a smaller extent than for YL coal. The freezingpoint distribution (FPD) for pore condensed water that froze at 213-260 K was determinedexperimentally by NMR and also simulated numerically using a Gaussian function. A modifiedGibbs-Thompson equation, which relates the freezing point depression to the pore dimensionsemploying a cylindrical-shaped pore model, was applied to convert FPD into pore size distribution(PSD). The PSD, expressed as pore radius, ranged from 1 to 3 nm, suggesting that the reductionof Vp for the YL coal was mainly due to the shrinkage or collapse of pores with radii around 2nm, which are abundant in water-swollen coal before predrying.

Introduction

When partially or completely dried brown coals orlignites are exposed to water, they swell, but often donot regain their original volumes.1,2 This irreversiblechange induced by drying has been partly attributed tocollapse of the colloidal gel1,3,4 accompanied by theformation of stronger and shorter hydrogen-bond bridgesbetween coal macromolecules. The gel collapse couldlimit the accessibility of organic solvents5 and masstransfer in aqueous media.4 The water sorption isothermon bed-moist brown coal6 shows strong hysteresisbetween the desorption and readsorption curves, andthe hysteresis persists at very low relative vapor pres-sures. At relative vapor pressures above 0.5, the normalcapillary condensation mechanism explains the hyster-esis with lowering of vapor pressure according to theKelvin equation.7 Although there is no generally ac-

cepted mechanism to explain the persistence of thehysteresis loop in the multilayer and monolayer waterregions of the isotherms, it is attributed to differencein the adsorption and desorption mechanisms, which areassociated with swelling and shrinkage effects such asthe irreversible shrinkage or collapse of capillaries withdrying. This study was undertaken to examine theirreversible nature of the colloidal gel structure of coalin the cycle of water removal and swelling, focusing onits porous structure.

Conventional techniques, such as gas adsorption/desorption and mercury porosimetry, are only utilizedto characterize dry materials, and can hardly be appliedto the pore structure analysis of water-containingmaterials. Drying induces irreversible pore collapse anda considerable reduction in the internal porosity. Hence,water itself is the only suitable probe molecule forinvestigating the porous structure of coal-sorbing water.In general, water sorbed in or on solid materials, suchas coal, has properties that differ from those of bulkwater in its normal thermodynamic states.8-14 Norinaga

* Author to whom correspondence should be addressed.† Present address: Institute for Chemical Reaction Science, Tohoku

University Katahira, Aoba-ku, Sendai, 980-8577, Japan. Fax: +81-22-217-5655. E-mail: norinaga@hisui.icrs.tohoku.ac.jp.

(1) Deevi, S. C.; Suuberg, E. M. Fuel 1987, 66, 454.(2) Woskoboenko, F.; Stacy, W. O.; Raisbeck, D. The Science of

Victorian Brown Coal; Durie, R. A., Ed.; Butterworth-HeinemannLtd.: Oxford, 1991; p 152.

(3) Evans, D. G. Fuel 1973, 52, 186.(4) Gorbaty, M. L. Fuel 1978, 57, 796.(5) Suuberg, E. M.; Otake, Y.; Yun, Y.; Deevi, S. C. Energy Fuels

1993, 7, 384.(6) Allardice, D. J.; Evans, D. G. Fuel 1971, 50, 236.

(7) Thompson, W. L. K. Philos. Mag. 1871, 42, 448.(8) Mraw, S. C.; Naas-O’Rourke, D. F. Science 1979, 205, 901.(9) Mraw, S. C.; Naas-O’Rourke, D. F. J. Colloid Interface Sci. 1982,

89, 268.(10) Lynch, L. J.; Webster, D. S. Fuel 1979, 58, 429.(11) Lynch, L. J.; Barton, W. A.; Webster, D. S. Proceedings of the

16th Biennial Low-Rank Fuels Symposium; Groenewold, G. H., Ed.;Energy and Environmental Research Center: Montana, 1991; p 187.

1058 Energy & Fuels 1999, 13, 1058-1066

10.1021/ef990024v CCC: $18.00 © 1999 American Chemical SocietyPublished on Web 07/10/1999

et al.14 classified water sorbed in various types of coalon the basis of its freezing properties, which wereevaluated using a combination of differential scanningcalorimetry (DSC) and proton magnetic resonance (1HNMR) techniques. They found two different types offreezable water; free water identical to bulk water andbound water that freezes at around 226 K. These twotypes of water account for only 35-78% of the totalwater content; the rest is nonfreezable water. Boundwater has a lower freezing point and congelation en-thalpy than bulk water. The difference in the freezingproperties of bulk and bound water are directly relatedto the size of the cluster of water molecules, that is, thesize of the space in which they are condensed. The lowermelting or freezing temperature of a solid or liquid inpores is frequently ascribed to the very small size of acluster in the pore15,16 and the large surface-to-volumeratio of material condensed in a capillary.16 Gibbs17

examined the thermodynamic stability of small crystalsformed from liquid confined within pores, and demon-strated that the melting transition temperature of porematerial depends on the surface curvature. Thompson7

derived a related theory in terms of the effect of thecurvature on the vapor pressure of liquid droplets. TheGibbs and Thompson equations can be combined torelate the reduction in the transition temperature to thecrystal dimensions. For a liquid confined within a porein which a crystal is forming, assuming that the contactangle between the liquid, solid, and pore wall is 180°,the temperature reduction, ∆T, is given by the Gibbs-Thompson equation as ∆T)(2γs,lT0/rF∆H), where γs,l isthe surface free energy at the liquid-solid interface, T0is the normal bulk melting point, r is the radius of thecrystal, ∆H is the heat of fusion, and F is the density ofthe solid formed by the phase transition. Using theGibbs-Thompson equation as a theoretical basis, it ispossible to predict the pore size distribution from thefreezing point distribution of water in pores. So far,inorganic porous materials have been characterized onthe basis of the freezing or melting properties of waterconfined in the pores that were evaluated by 1HNMR,18-25 DSC,26,27 or a combination of both.28 The 1HNMR technique is particularly suited to studying freez-

ing phenomena. It is easy to measure the signals dueto water remaining unfrozen in a sample, because ofthe large reduction in the transverse relaxation time ofwater when it turns into ice. We can see how much ofthe pore water is frozen at any particular temperatureby monitoring the signals while cooling the sample.

In this study, we present the pore size distribution ofrewetted coal samples prepared from non-, partially,and completely predried coals. Three kinds of as-received coals were used, ranging from brown to bitu-minous. The freezing properties of water in the coalsamples were evaluated in the temperature range from170 to 294 K using 1H NMR, and then the pore sizedistribution was estimated. The effect of the extent ofpredrying and coal type on the pore structure of coalafter rewetting is discussed.

Experimental Section

Coal Samples. Two Argonne Premium Coal Sample Pro-gram (PCSP) coals29 and a brown coal were studied. TheNippon Brown Coal Liquefaction Co. Ltd., Japan, and the CoalCorp. of Victoria, Australia, supplied the lumps of the browncoal. They were stored in gastight drums during transportationto prevent the evaporation of water. The elemental composi-tion, and ash and water contents of the coal samples are listedin Table 1. Yallourn (YL) coal was pulverized to grains finerthan 150 µm in a glovebag filled with nitrogen gas saturatedwith water vapor. Figure 1 outlines the procedure used toprepare the samples. Each coal sample was stored at 293 Kin a gastight vessel filled with nitrogen gas saturated withwater vapor for at least 2 weeks prior to use. The water contentof the coal samples was determined from the fractional massrelease by drying at 380 K under a nitrogen gas flow for 2 hin a thermobalance (TG-2000S, Mac Science Co. Ltd.). For eachcoal, samples with different residual water contents wereprepared at 303 K by varying the relative humidity in theclosed container from 5 to 97% using aqueous solutionssaturated with selected inorganic salts.1 The water content ofthe partially dried samples was determined in the samemanner as described above. Hereafter, the water content willbe referred to as wi for the raw coal samples or w for thepredried and rewetted samples, using grams of water per gram

(12) Barton, W. A.; Lynch, L. J. Proceedings of the 6th AustralianCoal Science Conference, Newcastle, 1994; p 65.

(13) Barton, W. A.; Cheng, J. Y. Proceedings of the 7th AustralianCoal Science Conference, Gippsland, 1996; p 521.

(14) Norinaga, K.; Kumagai, H.; Hayashi, J.-i.; Chiba, T. EnergyFuels 1998, 12, 574.

(15) Defay, R.; Prigogine, I.; Bellemans, A.; Everett, D. H. SurfaceTension and Adsorption; Wiley: New York, 1966.

(16) Jackson, K. A.; Chalmers, B. J. Appl. Phys. 1958, 90, 5420.(17) Gibbs, J. W. Collected Works; Longmans and Green: New York,

1928.(18) Strange, J. H.; Rahman, M.; Smith, E. G. Phys. Rev. Lett. 1993,

71, 3589.(19) Alnami, S. M.; Strange, J. H.; Smith, E. G. Magn. Reson.

Imaging 1994, 12, 257.(20) Overloop, K.; Van Gerven, L. J. Magn. Reson. Ser. A 1993, 101,

179.

(21) Hansen, E. W.; Schmidt, R.; Stocker, M.; Akporiaye, D. J. Phys.Chem. 1995, 99, 4148.

(22) Schmidt, R.; Hansen, E. W.; Stocker, M.; Akporiaye, D.;Ellestad, O. H. J. Am. Chem. Soc. 1995, 117, 4049.

(23) Schmidt, R.; Stocker, M.; Hansen, E. W.; Akporiaye, D.;Ellestad, O. H. Microporous Mater. 1995, 3, 443.

(24) Akporiaye, D.; Hansen, E. W.; Schmidt, R.; Stocker, M. J. Phys.Chem. 1993, 97, 7743.

(25) Hansen, E. W.; Schmidt, R.; Stocker, M. J. Phys. Chem. 1996,100, 11396.

(26) Ishikiriyama, K.; Todoki, M.; Motomura, K. J. Colloid InterfaceSci. 1995, 171, 92.

(27) Ishikiriyama, K.; Todoki, M. J. Colloid Interface Sci. 1995, 171,103.

(28) Rennie, G. K.; Clifford, J. J. Chem. Soc., Faraday Trans. 1 1977,73, 680.

(29) Vorres, K. S. User’s Handbook for the Argonne Premium CoalSample Program; Argonne National Laboratory: Argonne, IL, 1993.

Table 1. Properties of Coals Used

[wt % dafbcoal]coal (symbol) C H N S Oa

ash[wt % mfc coal]

moisture[wt % wet basis]

Yallourn (YL) 65.0 4.6 0.6 0.2 29.6 1.6 59.4Beulah Zap (BZ) 72.9 4.8 1.2 0.7 20.3 9.7 32.2Illinois #6 (IL) 77.7 5.0 1.4 2.4 13.5 15.5 10.0

a By difference. b Dry-ash-free. c Moisture-free.

Predrying Effect on Porous Structure of Water-Swollen Coal Energy & Fuels, Vol. 13, No. 5, 1999 1059

of moisture-free coal as the units. The completely driedsamples were prepared by drying under a pressure of 1 Pa at303 K for 48 h, which is long enough to attain a constantweight. The amounts of free, bound, and nonfreezable watercontained in the raw samples were determined by DSC asreported in a previous study.14

1H NMR. The coal sample was weighed and transferred toan NMR tube with a 10 mm o.d. The masses of the sampleson a moisture-free coal basis were 0.1 g for YL, 0.15 g for BZ,and 0.2 g for IL coal. A 5 mL aliquot of pure water was thenadded to the sample, which was stirred vigorously with a thinrod under a nitrogen atmosphere to remove gas trapped inthe slurry. The tube was sealed and then stored in an air bathkept at 303 K for at least one week, which was long enough toequilibrate the swelling. After centrifugation, the excess waterwas removed via a cannula so that the masses of the water-saturated samples were approximately 0.4 g for YL, 0.35 g forBZ, and 0.3 g for IL coal. The excess was removed to minimizethe inhomogeneity of the magnetic field influencing theobserved transverse relaxation. The water content of theswollen sample will be referred to as W and the units aregrams of water per gram of moisture-free coal. Since W of asample is much larger than wi, all the pores in the coal samplesare expected to be filled with water. The free induction decays(FID) after a single 90 degree pulse were measured attemperature intervals of 2-5 K while cooling from 294 to 170K, using a JEOL Mu-25 NMR spectrometer operating at aproton resonance frequency of 25 MHz. The signal at eachtemperature was recorded under temperature equilibration.Typical values for the pulse width, repetition time, and numberof scans were 2.0 µs, 6 s, and 16, respectively. Details of theNMR measurement are described elsewhere.14 A specificprocedure was adopted to measure the freezing of waterwithout significant supercooling. Preliminary observationsshowed that the temperature had to be lowered to about 258K before spontaneous freezing of bulk water occurred. There-fore, the sample was cooled to 253 K for 10 min aftermeasuring the FID at 274 K, and then warmed to 272 K. Thisprocedure ensured that all the interparticle and bulk waterwas frozen before starting to record the actual FID below 272K.

Swelling and Specific Gravity Measurements in Wa-ter. The effect of predrying on the volume of rewetted coalwas examined using a conventional solvent swelling tech-nique.30 Approximately 0.3 g of a coal sample with known wwas put in a cylindrical glass tube. Roughly 5 mL of waterwas added to the sample and the contents were stirredvigorously under a nitrogen atmosphere. The tube was sealedand stored at 303 K for at least one week. The contents werethen centrifuged and the height of the bed of water-swollencoal, Hw, was measured. Here, the height of the coal particlebed per unit mass of coal on a moisture free basis, Hw

n, isdefined as,

where, m is the mass of the coal samples. Assuming that thevolumetric fraction of the swollen coal particles in the bed isconstant at any w, the volume of the swollen coal at w relativeto that at wi is given by Hw

n/Hwin.

The apparent densities of the nondried and completely driedcoal samples were also measured by pycnometry. Approxi-mately 0.25 g of the sample and about 4 mL of water wereintroduced into a 5 cm3 Wadon specific gravity bottle, and thecontents were stirred vigorously with a thin rod. The bottlewas placed in a vacuum desiccator and the pressure wasgradually reduced to remove gas trapped in the slurry. Thebottle was placed in an air bath kept at 303 K for at least oneweek, before leveling the water and weighing. The reproduc-ibility of the determination of Hw

n/Hwin and the density were

within (2% and (0.5%, respectively.

Results and Discussion

Volume of Water Condensed in Coal Pores.Figure 2 illustrates the FID of YL with wi ) 1.46 atdifferent temperatures. Each FID consists of two dif-ferent decays: a rapid Gaussian decay arising from“rigid” or “immobile” protons and a slow exponentialdecay arising from “mobile” protons. The mobile protonsinclude the water protons and a portion of the coalprotons. The latter are nearly identical to hydroxylic

(30) Green, T. K.; Kovac, J.; Larsen, J. W. Fuel 1984, 63, 935.

Figure 1. Procedure used to prepare samples.

Figure 2. Temperature-dependent changes in the 1H NMRtransverse relaxation signals for YL coals. (wi ) 1.46, W )2.89)

Hwn )

Hw(1 + w)m

(1)

1060 Energy & Fuels, Vol. 13, No. 5, 1999 Norinaga et al.

protons, which interact with water as a plasticizingagent via hydrogen bonds.31 The rigid protons are iceprotons at temperatures below the freezing point of bulkwater, and rigid coal protons that constitute the majorportion of the coal protons. The spin-spin relaxationtimes of ice and rigid coal protons are 6 µs20 and lessthan 30 µs,32 respectively. Therefore, it is reasonableto assume that the transverse magnetization at 40 µsat a given temperature, T, defined as IT, correspondsquantitatively to the amount of the mobile protonscontained in the sample. The validity of this assumptionwill be discussed later. I256, I221, and I190 are definitelysmaller than I294, and this is primarily due to thetransition of water into ice. The IT of the other samplesalso changed with temperature in the same manner asYL with wi ) 1.46. At 294 K where no ice is formed,I294 represents the sum of the numbers of water protonsand mobile coal protons. Since W ) 2.89 for YL with wi) 1.46, the coal contains 0.32 mol of water protons and0.015 mol of mobile coal protons31 when exposed toexcess water at 294 K. Thus, the latter represents 4.5%of the total mobile protons. The percentages for BZ andIL coals were 5.7% and 4.8%, respectively. Although themobile coal protons can never be distinguished from thewater protons, their fraction to the total mobile protonsis small enough to consider I294 represents the amountof water protons for expedience. On this basis, theamount of water per unit mass of moisture-free coal atT is here defined as W(T), which can be expressed asgrams of water per gram of moisture-free coal by

W(T) can be converted into the volume of water, V(T),in cubic centimeters of water per gram of moisture-freecoal by assuming the density of water is 1 g/cm3.

Figure 3 shows V(T) as a function of temperature forthe coal samples that were not predried. In the courseof cooling, the volumes of all the coals decrease sharplyin the range from 273 to 260 K, indicated as Region I.This reduction is attributed to the freezing of bulkwater. The Gibbs-Thompson equation predicts thatwater freezes at 260 K when it is confined in a pore witha radius of 5 nm. Thus, a portion of the bulk water maybe confined in pores with a radius larger than 5 nm.The changes in V(T) in this region are continuous, andtherefore it is difficult to distinguish the freezing ofwater condensed in relatively large pores from that ofextra-pore water. The presence of metallic cationsdissolved in water or some colloidal substances insidecoal pores would affect the freezing properties of thepore water, and these effects may contribute thiscontinuous change in V(T). The identification of sucheffects requires the reference coal samples, from whichthe above-described materials have been removedthrough some pretreatments. The treatments would

involve washing, filtration, and sometimes drying, andmay alter the porous structure of the sample. In thispaper, we intend to evaluate the drying-induced ir-reversible changes in the porous structure by using as-received and nondried coals as the starting materials.Therefore it is presently difficult to identify the aboveeffects on the freezing characteristics of the pore water.In Region II, which ranges from 260 to 213 K, V(T)decreases due to the freezing of “bound water” con-densed in pores with radii smaller than 5 nm. Thefreezing of bound water is a phase transition of thewater into ice, and is detected by DSC as an exothermicprocess.14 The decrease in V(T) in this region is causedby a reduction in the mobility of “nonfreezable water”,which should be recognized as an apparent phasetransition as discussed extensively by Overloop et al.20

This allows us to determine the volumes of nonfreezablepore water (Vnf), freezable pore water (Vf), and bulkwater (Vb), respectively as V(213), decrement of V(T) at260-213 K, and that at 273-260 K. Vnf, Vf, and Vb areall indicated in cubic centimeters of water per gram ofmoisture-free coal and hence the sum of these equalsV(294).

Vnf and Vf for the coals that were not predried aresummarized in Table 2 and compared with the amountsof nonfreezable (Wnf-dsc) and freezable bound (Wf-dsc)water determined previously,14 respectively. For YL

(31) Norinaga, K.; Kumagai, H.; Hayashi, J.-i.; Chiba, T. EnergyFuels 1998, 12, 1013.

(32) Yang, X.; Garcia, A. R.; Larsen, J. W.; Silbernagel, B. G. EnergyFuels 1992, 6, 651.

W(T) ) W(294)IT

I294(2)

V(T) ) W(T) (3)

Figure 3. Change in V with temperature for the samplesprepared from raw coals. Region I (273-260 K), the freezingof bulk water; Region II (260-213 K), the freezing of porecondensed water; Region III (less than 213 K), the reducedmobility of nonfreezable pore water.

Table 2. Comparison of Vf, Vp, and Vnf with the Amountof Bound, Nonfreezable Water, and Sum of These Two

Types Water1H NMR DSCa

[cm3/g-mf coal] [cm3/g-mf coal]c

coalb Vf Vnf Vp Wf-dsc Wnf-dsc sum

YL 0.32 0.32 0.64 0.35 0.30 0.65BZ 0.24 0.27 0.51 0.17 0.31 0.48IL 0.13 0.06 0.19 0.05 0.05 0.10

a Ref 14. b Raw samples. c Represented by volumetric unit byassuming the density of water as 1 g/cm3.

Predrying Effect on Porous Structure of Water-Swollen Coal Energy & Fuels, Vol. 13, No. 5, 1999 1061

coal, Vnf and Vf both agree well with the correspondingamounts and the agreement confirms the validity of theabove-described analysis of FID. Unlike YL, Vf issomewhat larger than Wf-dsc for the other coals. Thismay result from a much larger W in the present NMRmeasurements (1.33 for BZ and 0.56 for IL) than in theDSC measurements,14 where the water content of thesamples was identical to the wi (0.48 for BZ and 0.11for IL). Considering that DSC did not detect any bulkwater in the raw BZ and IL coals,14 the amount ofinherent water would not be enough to fill or fullyexpand the pores in these coals and therefore thesignificantly larger W could result in a greater Vf of thecoals.

The raw YL coal with wi ) 1.46 was partially orcompletely dried, as described in Experimental section,so that the coal had values of w ranging from 0 to 0.90.The dried samples were subjected to the DSC analysisand the amounts of bulk, freezable-bound, and non-freezable water remaining in the samples were deter-mined. The analysis revealed that drying removed thewater sorbed in the raw YL coal in the order bulk,bound, and nonfreezable water as reported elsewhere.31

The samples with a different w were further analyzedby NMR after they were swollen in water, and then Vb,Vf, and Vnf were determined. In Figure 4 are plotted thetotal volumes of freezable and nonfreezable pore waters,Vp, Vf, and Vnf of the samples, against w. When w is0.33 or larger at which bound and/or free water has beenremoved but nonfreezable water never done, Vp, Vf, andVnf are all seen to remain unchanged. The volumesindependent of w indicate that the volume of the poresfilled with water is not affected by drying when w isgreater than 0.33. This result also suggests that theporous structure of YL coal changes reversibly in thedrying-swelling cycle, unless the nonfreezable water isremoved. When the nonfreezable water is removed bydrying beyond w of 0.33, Vp decreases from 0.64 to 0.53.The change in the volume of water-retaining pores thusbecomes irreversible with extensive drying. The reduc-tion in Vp also seems to be caused by the decrease inVf. Table 3 compares Vp, Vf, and Vnf for the raw BZ (wi) 0.48) and IL (wi ) 0.11) with those for completelydried samples at w ) 0. The complete drying of IL

reduces Vf, thereby decreasing Vp as observed in thedrying of YL coal. As expected, the drying of BZdecreases Vf, but a slight decrease in Vnf also seems tooccur. The results shown in Figure 4 and Table 3 provethat the irreversible volumetric shrinkage of the coalsresults from the decrease in the volume of the pores,which is described quantitatively by the reduction of thevolume of pore condensed water. It is also demonstratedthat the removal of inherent nonfreezable water isresponsible for the irreversible change.

The irreversible volumetric shrinkage of coals wasalso examined using the conventional solvent-swellingtechnique.30 Table 4 presents Vp, the volumetric recov-ery, Hw

n/Hwin, and the density measured in water, Fc,

for the raw and the completely predried coals that weresubsequently immersed in water. For all the predriedcoals, Hw

n/Hwin is smaller than unity, which implies an

incomplete recovery of the volume in the swollen stateor an irreversible volumetric shrinkage induced bydrying. Hw

n/Hwin becomes smaller as the coal rank

decreases. On the other hand, drying had little effecton the density as measured by water pycnometry. Thismeans that drying does not irreversibly alter the volumeof the dense part of the coal samples, i.e., the excludedvolume of coal macromolecules. The coal volume swollenby water, Vc, which includes the volumes of the denseand porous parts, can be expressed as

Vc of the swollen coal at w ) 0 relative to that at theraw state is here defined as fV and is shown in Table 4.As the coal rank decreases, fV becomes smaller, alongwith Hw

n/Hwin. Thus, the irreversible decrease in Hw

n/Hwi

n upon drying can be partially attributed to theirreversible change in Vp, which leads to a loss inmoisture-holding capacity. The collapse or shrinkage ofpores induced by drying contributes to the irreversiblereduction in pore volume. When fV is larger than Hw

n/Hwi

n, it indicates that the irreversible change in thelatter is due not only to the decrease in Vp, but also toa collapse of macropores that are not included in Vp oran increase in the packing efficiency of the particles dueto changes in their shapes.

Freezing Point Distribution of Pore CondensedWater. To estimate the predominant size of the pores

Figure 4. Changes in Vf, Vnf, and Vp with w for YL coal.

Table 3. Comparison of Vp, Vf, and Vnf for Raw BZ and ILwith the Values for Completely Predried Samples

[cm3/g-mf coal]coal

w[g/g-mf coal] Vf Vnf Vp

BZ 0.48 0.24 0.27 0.51BZ 0 0.17 0.24 0.41IL 0.10 0.13 0.06 0.19IL 0 0.10 0.06 0.16

Table 4. Vp, Hwn/Hwi

n, Gc, and fV of Raw and CompletelyPredried Coals Subsequently Immersed in Water

coalw

[g/g-mf coal]Vp

[cm3/g-mf coal]Hw

n/Hwin

[-]Fc

[g/cm3-mf coal]fV[-]

YL 1.46 0.64 1.00 1.49 1.00YL 0 0.53 0.75 1.51 0.91BZ 0.48 0.51 1.00 1.53 1.00BZ 0 0.41 0.80 1.52 0.92IL 0.11 0.19 1.00 1.44 1.00IL 0 0.16 0.93 1.46 0.96

Vc ) 1/Fc + Vp (4)

1062 Energy & Fuels, Vol. 13, No. 5, 1999 Norinaga et al.

that are involved in the change in Vp, the freezingproperties of the freezable pore water were analyzedfurther. It is well-known that the freezing point tem-perature of pore condensed water is a function of thepore size. However, there are always some interactionsbetween coal functional groups with water molecules toform local hydrogen-bonding structures or networks,which can alter the freezing property. We studiedpreviously the changes in the hydrogen mobility in sixdifferent coals induced by drying at 303 K by means ofthe pulsed 1H NMR technique.31 It was found that aportion of the coal hydrogen appears to be mobile in theNMR sense and the amount of the mobile coal hydrogenvaried inversely with the amount of the nonfreezablewater, while the release of the free and the bound waterhad little effect on this amount. The chemical nature ofthe mobile hydrogen was identified as the hydroxylichydrogen by using a hydrogen-deuterium exchangetechnique. Accordingly, the nonfreezable water interactsdirectly with coal hydroxyls to form local hydrogenbonding structures, while bound water does not seemto participate in the structure. Therefore, the effect ofhydrogen-bonding formation with coal functional groupsonly makes a minor contribution to the freezing propertyof free and bound water. In this paper, we estimate thepore size distribution based on the freezing property offreezable pore condensed water that should be affectedmainly by their cluster size, i.e., size of space (poredimension) in which they are condensed. The distribu-tion of the size of pores retaining freezable water canbe derived from the range in the freezing point of thewater. Figure 5 shows the volume of freezable waterthat has not yet been converted into ice, V, as a functionof inversed temperature X ()1000/T) for the samplesprepared from raw YL, BZ, and IL coals. V(X) at upperlimit of X, 3.85 (T ) 260 K), and the lower limit, 4.69(T ) 213 K), are identical to Vp and Vnf, respectively.Each solid line in the figure represents V(X) calculatedusing the integrated Gaussian distribution function ofSchmidt et al., who analyzed the freezing properties of

water confined in mesoporous silica materials.22

where Xc, ∆, and k are X at the center of the distribution,the width of the distribution, and an integral constant,respectively. These parameters were determined byfitting the calculated V(X) to the observed value, usinga nonlinear least-squares method. Table 5 summarizesthe combinations of Xc and ∆ giving the best fit to V(X)observed for the individual samples. The temperatureat the center of distribution of the samples, Tc ()1000/Xc), is between 222 and 233 K, and the range agreeswell with the exothermic peaks arising from the conge-lation of freezable bound water around 227 K, whichare detected by DSC.14 By differentiating eq 4 withrespect to X, we obtain

Figure 5 presents dV/dX for the individual coals withdifferent values of w as a function of X. The removal ofnonfreezable water prior to the swelling of YL coal,resulting in w less than 0.33, is seen to shift the freezingpoint distribution (FPD) toward a larger X correspond-ing to a lower T. The shift due to the decrease in w from1.46 to 0.33 is smaller than this. For the other coals,the FPD also shifts toward a larger X with completedrying, although the changes are less significant thanthose for YL coal. According to the Gibbs-Thompsonequation, the freezing of pore water at a larger Xindicates smaller pores. Therefore, the results suggestthat irreversible shrinkage of the pores is induced bydrying.

Estimation of the Size Distribution of PoresRetaining Freezable and Nonfreezable Water. TheGibbs-Thompson equation provides a theoretical basisfor the analysis of the FPD. It relates the freezing pointdepression (∆T) of a liquid filling a pore to the size ofthe ice crystal formed by the freezing, expressed as itsradius, r. Assuming all the parameters in the equationare independent of the temperature and pore dimension,this equation can be simplified to

Figure 5. V versus inverse temperature X ()1000/T) forsamples prepared from raw YL, BZ, and IL coals. The solidlines are nonlinear least-squares fits to eq 6.

Table 5. Values for the Analytical Fit of the Intensityversus Inverse Temperature Curves

coalw

[g/g-mf coal]Xc (Tc)

[K-1] ([K]) ∆

YL 1.46 4.30 (233) 0.17YL 0.33 4.31 (232) 0.22YL 0.25 4.35 (230) 0.16YL 0.13 4.48 (223) 0.26YL 0 4.50 (222) 0.28BZ 0.48 4.36 (229) 0.27BZ 0 4.38 (228) 0.25IL 0.11 4.25 (235) 0.21IL 0 4.30 (233) 0.13

V(X) )Vp

xπ∫

0

X - Xc

x2∆ exp(-u2)du + k )

Vp

2erf(X - Xc

x2∆ ) + k (5)

dVdX

)Vp

x2π∆exp(-

X - Xc

x2∆ )2

(6)

∆T ) Rr

(7)

Predrying Effect on Porous Structure of Water-Swollen Coal Energy & Fuels, Vol. 13, No. 5, 1999 1063

The coefficient R is a characteristic of the condensedliquid, and also depends on the interaction between thecondensates and the host materials. Presented in theform of the Gibbs-Thompson equation, it is assumedthat the parameters can be represented as those of abulk liquid. However, it is well-known that the molarheats of fusion of a number of organic liquids confinedin the pores of different types of glass decrease with adecreasing pore radius smaller than 10 nm.33 The heatof fusion of freezable bound water in coals has also beenreported to be lower than that of bulk water.8,9,14

Furthermore, many different values for the ice-waterinterfacial tension have been reported or assumedtacitly in the literature.34-39 Accordingly, the inverseproportional coefficient R must be calculated using amodel for the structure of the pores in coal, as describedbelow. In addition, eq 7 assumes that r is equivalent to

the dimension of a crystalline solid formed by phasetransition. A portion of the pore water in the systemsexamined is nonfreezable, as described above. This typeof water may not be isolated from freezable water andwould act as an interface shield between the solid coalmatrix and the inner ice core, as some workers havereported.26,27 This concept allows us to use the actualpore radius, Rp, which is the sum of r and the thicknessof the layer of nonfreezable water, â, which depends onthe surface chemistry of the pore wall. Equation 7 isthen modified as

The assumptions made to develop this model of theporous structure of water-swollen coal are (i) the poresare isolated from each other, (ii) all the pores arecylindrical in shape as shown schematically in Figure7, (iii) nonfreezable water exists exclusively at theinterface between the core of ice formed from thefreezable bound water and the pore wall, and (iv) â isindependent of the pore radius, temperature, and w, andis assumed to be 0.6 nm for YL and BZ coals and 0.3nm for IL coal.

Although there is presently little information avail-able for determining the â values for coal, the followingdiscussion supports assumption iv. Since water becomesnonfreezable when the molecular cluster is too small,nonfreezable water is likely to be dispersed on amolecular scale. Nonfreezable water is bound to specificsites via specified interactions such as hydrogen bondswith hydrophilic sites in the coal matrix. The numbersof hydroxyl groups, determined by a hydrogen-deute-rium exchange technique,31 are 8.1, 8.4, and 4.2 mmol/g-mf coal for YL, BZ, and IL coal, respectively. There-fore, the number of nonfreezable water molecules permole of hydroxyl groups is 2.0, 2.0, and 0.7 for YL, BZ,

(33) Jackson, C. L.; McKenna, G. B. J. Chem. Phys. 1990, 93, 9002.(34) Skapski, A.; Billups, R.; Rooney, A. J. Chem. Phys. 1957, 26,

2754.(35) Ketcham, W. M.; Hobbs, P. V. Philos. Mag. 1969, 19, 1161.(36) Hardy, S. C. Philos. Mag. 1977, 35, 471.(37) Brun, M.; Lallemand, A.; Quinson, J. F.; Eyraud, C. Thermo-

chim. Acta 1977, 21, 59.(38) Deodhar, S.; Lunner, P. In Water in Polymer, ACS Symp. Ser.

No. 127; Rowland, S. P., Ed.; American Chemical Society: Washington,DC, 1980.

(39) Homshow, L. G. J. Colloid Interface Sci. 1981, 84, 141.

Figure 6. Freezing point distribution of freezable pore waterfor water-swollen YL (a), BZ (b), and IL (c) coals.

Figure 7. Schematic representation of a cylindrical pore.

Rp ) R∆T

+ â (8)

1064 Energy & Fuels, Vol. 13, No. 5, 1999 Norinaga et al.

and IL coal, respectively. Assuming that the walls ofpores where water is condensed are completely coveredwith hydroxyl groups as the predominant hydrophilicsites, the number of monolayers of nonfreezable porewater is approximately two for YL and BZ coals, andone for IL coal. Since the van der Waals radius of awater molecule is about 0.3 nm, the â values are then0.6 nm for YL and BZ coals and 0.3 nm for IL coal. Thesevalues of â are within the range of that of porous silica,in which the number of monolayers has been deter-mined to be one to three theoretically and experimen-tally.26, 40-46

For the cylindrical-shaped pore model, the ratio of thetotal pore volume, Vp, to the volume of freezable porewater, Vf, can be expressed by

where Rp(ave) and r(ave) are the volume-averaged radiiof the actual pore and the ice core, respectively. r(ave)can be given by

Then, R can be determined by combining eqs 9 and10

As listed in Table 6, R ranges from 45 to 57 (nm K)depending on the samples, and this agrees well withthe values reported by Schmidt et al.22 They estimatedR to be about 50 for water confined in regular mesopo-rous MCM-41 materials with pore radii ranging from 1to 3 nm, by combining 1H NMR and N2 adsorption data.

It is possible to derive a function describing the poresize distribution (PSD), dVp/dRp, from FPD and the

determined values of R and â by noting that

and

The following formula is derived by combining eqs 6,12, and 13:

All the symbols have already been defined. The PSDcurves, dVp/dRp vs Rp, are shown in Figure 8. Rp isdistributed over a range from 1 to 3 nm. For YL coal,the distribution shifts toward a smaller Rp when wdecreases from 0.33 to 0, at which point the inherentnonfreezable water is removed. The change in thedistribution is much less significant at greater values

(40) Antoniu, A. A. J. Phys. Chem. 1964, 68, 2745.(41) Litvan, G. G. Can. J. Chem. 1966, 44, 2617.(42) Drost-Hansen, W. Ind. Eng. Chem. 1969, 61, 10.(43) Pearson, R. T.; Derbyshire, W. J. Colloid Interface Sci. 1974,

46, 232.(44) Bruggeller, P. J. Colloid Interface Sci. 1983, 94, 524.(45) Handa, Y. P.; Zakrzewski, M.; Fairbridge, C. J. Phys. Chem.

1992, 96, 8594.(46) Overloop, K.; Van Gerven, L. J. Magn. Reson. Ser. A 1993, 101,

147.

Table 6. r, â, Rp(ave), and SH2O Values and the ReportedCO2 Surface Areas

coalw

[g/g of mf coal]R

[nm K]â (fixed)

[nm]Rp(ave)

[nm]SH2O[m2/g]

SCO2

[m2/g]

YL 1.46 57 0.6 2.0 633YL 0.33 58 0.6 2.0 617YL 0.25 45 0.6 1.6 730YL 0.13 46 0.6 1.5 711YL 0 52 0.6 1.6 664 270a

BZ 0.48 58 0.6 1.9 537BZ 0 48 0.6 1.7 494 274b

IL 0.11 53 0.3 1.7 224IL 0 50 0.3 1.5 213 132b

a Ref 2. b Ref 47.

Vp

Vf)

Vf + Vnf

Vf) (Rp(ave)

r(ave) )2

) (1 + âr(ave))2

(9)

r(ave) ) RT0 - Tc

) RT0 - 1000/Xc

(10)

a )â(T0 - Tc)

xVp/Vf - 1(11)

Figure 8. Pore size distribution for water-swollen YL (a), BZ(b), and IL (c) coals.

dVdRp

) dVdX

dXdRp

(12)

X )103(Rp - b)

(Rp - â)T0 - a(13)

dVdRp

) 103R{(Rp - â)T0 - R}2

Vp

x2π∆exp(-

X - Xc

x2∆ )2

(14)

Predrying Effect on Porous Structure of Water-Swollen Coal Energy & Fuels, Vol. 13, No. 5, 1999 1065

of w, where the free and bound water are removed. Themean pore radius, Rp(ave) ()r(ave) + â), was calculatedand is listed in Table 6. Rp(ave) decreases from 2.0 nmat wi ) 1.46 to 1.6 nm at w ) 0. Therefore, the changein the PSD clearly demonstrates that the irreversibledecrease in Vp can be explained by an irreversibleshrinkage of pores with radii of approximately 2 nm,which are abundant in the raw coal. For BZ coal, thePSD seems to shift slightly toward a smaller Rp withwater removal while no meaningful changes wereobserved for IL coal, suggesting that the effect of dryingon the irreversible change in the porous structurebecome less significant with increasing coal rank.

Finally, we address the effect of drying on the surfacearea of the coal. For cylindrical pores, the specificsurface area of the water-swollen coal, SH2O, can be givenby

In Table 6, SH2O is listed with the reported surfacearea, SCO2, of each dried coal measured using a CO2 gasadsorption technique.2,47 For the individual coals, SH2Oestimated by the present method is 1.6-2.7 times largerthan SCO2. The SH2O/SCO2 ratio seems to be moresubstantial with decreasing coal rank. This is due tothe swellable colloidal nature of coal. The lower rankcoals are swollen by water to a greater extent becauseof their greater hydrophilicity. In the expanded coal, theincrease in the free volume of coal macromolecules givesthe coal a larger surface area.

Conclusions

The effect of predrying on the pore structure of water-swollen coal was evaluated on the basis of the freezing

properties of water condensed in pores. Within the limitsof the present experimental conditions, the followingconclusions can be drawn:

(1) In the course of predrying and subsequent swellingin water, an irreversible decrease in Vp and Vf takesplace when the predrying removes nonfreezable water.In contrast, Vnf is almost independent of the extent ofthe predrying. The reduction in Vp is therefore explainedby the reduction in Vf.

(2) The Gibbs-Thompson equation was used as atheoretical basis for converting FPD into PSD employinga cylindrical pore model. The change in the PSD of YLcoal clearly demonstrates that the irreversible decreasein Vp can be explained by the irreversible shrinkage ofpores with radii of approximately 2 nm, which areabundant in the raw coal before predrying.

(3) The specific surface areas of the water-swollencoal, SH2O, are 1.6-2.7 times larger than those of thedried coals determined using a CO2 gas adsorptiontechnique, SCO2. This provides evidence that dryingconsiderably reduces the internal porosity of coal.

Acknowledgment. The authors are grateful to Drs.Tadashi Yoshida and Masahide Sasaki of the HokkaidoNational Industrial Research Institute for their usefuladvice on the NMR measurements. Messrs. Ryo Moriya-ma and Nao Kashimura in CARET are acknowledgedfor their help on numerical analyses. This work wassupported in part by a “Research for the Future Project”(Coordinator: Prof. M. Iino, Tohoku University) grantfrom the Japan Society for the Promotion of Science(JSPS), through the 148th Committee on Coal Utiliza-tion Technology.

EF990024V(47) Larsen, J. W.; Hall, P.; Wernett, P. C. Energy Fuels 1995, 9,

324.

SH2O )2Vp

Rp(ave)(15)

1066 Energy & Fuels, Vol. 13, No. 5, 1999 Norinaga et al.