Mapping internal structure of coal by confocal micro-Raman spectroscopy and scanning microwave microscopy

Fuel 126 (2014) 32–37

Contents lists available at ScienceDirect

Fuel

journal homepage: www.elsevier .com/locate / fuel

Mapping internal structure of coal by confocal micro-Ramanspectroscopy and scanning microwave microscopy

http://dx.doi.org/10.1016/j.fuel.2014.02.0290016-2361/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +1 865 574 4684.E-mail address: [email protected] (A. Tselev).

Alexander Tselev a,⇑, Ilia N. Ivanov a, Nickolay V. Lavrik a, Alexei Belianinov a, Stephen Jesse a,Jonathan P. Mathews b, Gareth D. Mitchell c, Sergei V. Kalinin a

a The Center for Nanophase Materials Sciences, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USAb Leone Family Department of Energy and Mineral Engineering, EMS Energy Institute, The Pennsylvania State University, PA 16802, USAc EMS Energy Institute, The Pennsylvania State University, PA 16802, USA

h i g h l i g h t s

� Complexity of coal demands increasing spatial resolution in characterization.� Scanning probes offer various spatially-resolved characterization functionalities.� Coal samples were imaged with micro-Raman and Scanning Microwave Microscopy.� This combination allowed resolving coal structure below a 100-nm spatial resolution.

a r t i c l e i n f o

Article history:Received 12 October 2013Received in revised form 12 February 2014Accepted 15 February 2014Available online 28 February 2014

Keywords:Macerals mappingSub-100-nm spatial resolutionConfocal micro-Raman spectroscopyBayesian analysisScanning near-field microwave microscopy

a b s t r a c t

Structural complexity and variability of the chemical properties define technological applicability of coaland demand increasing accuracy and spatial resolution from the techniques used for coal characterizationfor development of new, clean, and efficient technologies of coal utilization. Here, we combined spatially-resolved reflectometry, fluorescence, and confocal micro-Raman spectroscopy with high-resolutionscanning probe microwave imaging to achieve a nondestructive sub-100-nm spatial resolution mappingof coal structure. It was found that this approach allows for high spatial resolution identification ofindividual elements in coal architecture, thus potentially generating valuable input for knowledge-drivenoptimization and design of coal utilization processes.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Development of the energy strategies based on direct andindirect coal liquefaction, coal gasification, direct carbon fuel cells,or solar-to-fuel cycles necessitates detailed understanding of thestructure and functional properties of coals to predict and designthe processing pathways. Advancement in this area began in1950s and 1960s with progress in characterization techniques suchas powder X-ray diffraction, infrared (IR) spectroscopy, adsorptionmeasurements, nuclear magnetic resonance (NMR), and electronparamagnetic resonance (EPR). These methods allowed bulk toatomic level insights into complex physical and chemical processes

associated with coal utilization [1]. However, fundamentally coal isa highly heterogeneous system on multiple length scales withmacro- to nanoscale porosity, mineral inclusions and associations,and a strong variability of local chemical compositions, as well asdegree of structural disorder [2]. These heterogeneities includedomains of inorganic and organic constituents. Among theinorganic mineral constituents of coal are silicates (quartz, mica,feldspars, zircon, clay minerals such as kaolinite and illinite, asexamples), sulfides (pyrite, sphalerite, galena, marcasite), carbon-ates (calcite, dolomite, siderite), sulfates (gypsum, barite), oxidesand hydroxides of iron, aluminum, magnesium, calcium, and others[2–5]. Organic components of coal—macerals—can be recognizedby optical microscopy. Originated from different dehydrogenatedparts of plants (roots, bark, leaves, pollen), depositional environ-ments, and degradation products, they are different in theirchemical composition and optical properties and may have more

A. Tselev et al. / Fuel 126 (2014) 32–37 33

oxygen-rich components (vitrinite), more hydrogen-rich mainlyaliphatic components (liptinite), and more carbon-rich highlyaromatic components (inertinite) [2,5].

The variability of the chemical properties of macerals definestechnological applicability of coal. Improvements in technologyof coal require increasing accuracy of characterization withincreasing demand on non-destructive real-time measurementsemploying advanced characterization tools [2]. The multiscale coalstructure naturally leads to increasing spatial resolution of thecharacterization tools up to the length scales of a few nanometers,the length scale typically cited as a length of structural coherencein coal macerals [2]. Evolution of coal analysis techniques—pro-gressively more sophisticated with access to smaller lengthscales—is needed in chemical and structural identification of differ-ent coal constituents in order to understand the chemical behaviorof coal during processing and utilization.

Among the old, well-established techniques, optical techniquessuch as measurements of reflectance and fluorescence candeliver general information about coal structure based on knownlarge-scale properties of coal constituents. However, the spatialresolution of these methods is limited by few micrometers. Awidely used high spatial resolution technique—transmissionelectron microscopy (TEM)—requires special sample preparationand is destructive. Spatial resolution of scanning electron micros-copy (SEM) can be below 10 nm, however, low atomic weight ofcarbon and oxygen do not allow reliable identification of maceralsat small-length scales. In turn, micro-Raman spectroscopy with aspatial resolution of about 1 lm was recently employed in studiesof inorganic mineral inclusions [4], amorphous and ordered carboncontent in coal [2,6–8], coal macerals (collotelinite, fusinite, andmacrinite) [9], chars [10,11], coke [12], and fly ash [13]. Complexityand high heterogeneity of the coal structure necessitate applicationof several different complimentary analytical techniques to gaininformation about local coal structural and chemical composition.A benefit of structural and functional imaging of the same areaof coal using multiple nondestructive, spatially resolved tech-niques including optical microscopy, micro-Raman spectroscopy,fourier-transform infrared spectroscopy (FTIR), Scanning electronmicroscopy (SEM) can be useful for further understanding ofstructure-properties relationship in coals [4,8,14–18]. However,coal heterogeneity expands below micron scale, making struc-ture–functional characterization with submicron resolutionhighly desirable. Towards this end, the current fleet of scanningprobe microscopy (SPM) techniques combining a variety of non-destructive probing functionalities offers an approach for spatiallyresolved coal characterization at the length scales down to 5 nm.

Here, further advancing the spatial resolution in non-destruc-tive coal characterization, we use combined confocal micro-Ramanand Scanning Microwave Microscopy (SMM) imaging [19,20] tonon-destructively resolve the structure of coal at the mesoscopic,sub-micron, lengths scale in situ. We combine spatially-resolvedreflectometry, fluorescence, and micro-Raman spectroscopy withhigh-resolution microwave imaging to achieve a sub-100 nm spa-tial resolution mapping of coal structure at the lithotype, maceral,and sub-maceral levels.

Fig. 1. Layout of the probe and operational principle schematic of a scanningmicrowave microscope.

2. Methods

2.1. Sample preparation

The initial candidate for exploration was a block sample ofsubbituminous/bituminous coal collected from an undergroundmine in the Uinta basin, Colorado, USA. The calorific value of thiscoal on a dry basis was 27.9 MJ/kg (ASTM D388, 2005). The fixedcarbon, volatile material and ash yield on the dry basis were

determined as 56.99%, 38.31%, and 4.70% (ASTM D7582, 2010),indicating bituminous rank.

2.2. Raman spectroscopy

Raman spectra were collected in backscattering geometry usinga Renishaw inVia Raman Microscope with a 50� objective (Leica,NA = 0.75) using 633 nm (HeNe laser) probing light. Raman mapswith 0.4-lm spatial resolution were collected using 10-s acquisi-tion time and 150-lW laser power on the sample surface. Reflec-tance spectra were acquired using Nikon Eclipse microscope witha Filmetrics F40 spectrometer attachment.

Preliminary Raman mapping of the coal sample showed no min-eral inclusions in the area of interest, justifying narrow spectralwindow for Raman mapping between 900 and 3200 cm�1. Thepoint-by-point mapping was done in confocal mode collecting1115 spectra over an area of 35 � 32 lm2 with a 1 lm step size.The Raman spectra were fit with a Gaussian–Lorentzian function,and the G and D peak intensity parameters were used to calculatethe map of G/D band intensity ratio. To attain insight into the spa-tial variability of Raman signal, this 3D spectral imaging data setwas analyzed using Bayesian deconvolution [21]. This approachrepresents data as Y = MA + N, where the collection of (noisy)observations Y is a linear combination of position-independentmatrices M (called endmembers) with respective relative abun-dances A corrupted by additive Gaussian noise N. The endmembersM are non-negative, and the respective abundances add up tounity. Hence, the spectrum at each location is decomposed into alinear combination of spectra of individual components in corre-sponding proportions. The unique aspect of this analysis is thatthe endmember spectra and abundances are estimated jointly ina single step, unlike multiple least square regression methodswhere initial spectra should be known.

2.3. Near-field microwave microscopy

Scanning microwave microscopy, Fig. 1, is a near-field scanningprobe technique, where electromagnetic field is coupled to asample through a sharp probe tip connected to a microwave trans-mission line. Microwaves are sent through a coaxial cable, reflectedoff the cantilever with corresponding amplitude and phase beingmeasured and recorded. The wave reflection coefficient dependson the impedance of the probe-sample system, and therefore, thereflected waves carry information about the sample’s complexdielectric permittivity, namely dielectric constant and electric

34 A. Tselev et al. / Fuel 126 (2014) 32–37

conductivity. Due to the field focusing effect of the sharp tip, localdielectric properties of the sample can be mapped with a lateralresolution close to the size of the tip apex. The tip simultaneouslyserves as a standard atomic-force microscope probe; it maintains agentle, but stable mechanical contact with a sample, necessary fora sufficiently strong tip-sample electrical coupling, and this allowssimultaneous acquisition of topographic and microwave images.The SMM system used in the experiments was designed by AgilentTechnologies and Asylum Research and is realized on an AsylumResearch MFP-3D platform using an Agilent E8364A PNA vectornetwork analyzer (VNA) as a reflectometer. The tip is electricallyinterfaced with a central conductor of a coaxial cable connectedto a port of the VNA through a matching circuit. The VNA measuresamplitude and phase of the waves reflected from the probe-samplesystem at a constant microwave frequency, which are plotted asimages together with topographic maps of the sample surface.The frequency used for microwave imaging was 5.2 GHz.

3. Results and discussion

3.1. Confocal micro-Raman spectroscopy

A fluorescence image of the coal surface in Fig. 2a obtained witha triple-band excitation reveals a variety of maceral types includedalong the bedding planes in a matrix of vitrinite maceral. Thisregion can also be visualized by micro-Raman spectroscopy(Fig. 2b). Fig. 2c shows a topographic AFM image of the sameregion. Here, the large ‘‘inclusion’’ corresponds to the dark region

Fig. 2. (a) A 100 � 100 lm2 fluorescence image of the coal surface obtained with triple bathe dark feature (inertodetrinite inclusion) in the center of the image in panel (a); the colsame region. (d) Three Bayesian endmembers of the Raman spectra of the map in panel (interpretation of the references to colour in this figure legend, the reader is referred to

in micro-Raman spectrum, Fig. 2b, and is identified as a particleof inertodetrinite (a highly reflecting particle in the correspondingmicrophotograph, Fig. S1� in Supplementary Material). The particlecontained in a matrix of vitrinite is a product of humification ofwoody plant tissue and corresponds to the highlighted area inmicro-Raman spectrum in Fig. 2b. The red areas in Fig. 2b corre-spond to the area of high fluorescence below the inertodetrinitefragment, which represent liptinite components or a combinationof liptinite and vitrinite depending upon sampling volume.

Results of Bayesian analysis of Raman data are illustrated inFig. 2d and e. It is immediately apparent that the inertodetriniteinclusion shows a clear spectral characteristic of being morecarbon-rich, whereas vitrinite matrix fluoresces strongly, asexpected. The increase in the number of Bayesian componentsfurther separates fluorescent spectra, but all the localized peaksremain within one component.

The spectrum of the first endmember (blue line in Fig. 2d) showstwo broad fluorescent peaks, with maxima at 2000 cm�1 (725 nm)and >3000 cm�1 (>781 nm). The Raman spectrum of this endmem-ber is dominated by G- and D-band features (peak frequencykG-band = 1602 cm�1, full width at half maximum FWHMG-band =25.1 cm�1 and kD-band = 1320 cm�1, FWHMD-band = 54.0 cm�1). The725-nm peak is also present in the second endmember (red linein Fig. 2d) and is mixed with the G and D bands associated withthe conjugated p system of double bonds. The spectra of the thirdcomponent (green line in Fig. 2d) shows residual G- and D-bandfeatures, however, their contribution is very small, and thespectrum is dominated by the background photoluminescence.

nd excitation. (b) A 50 � 50 lm2 map of the Raman signal intensity at 2996 cm�1 ator scale bar is in arbitrary units. (c) A 50 � 50 lm2 3D-AFM topography image of theb) color-coded by weight with corresponding loading maps shown in panel (e). (Forthe web version of this article.)

A. Tselev et al. / Fuel 126 (2014) 32–37 35

Considering that macerals have different reflectance, whichincreases for coals with higher maturity index, we measuredreflectance of each domain independently (Fig. S1� in Supplemen-tary Material). The intensity of white light reflectance in the400–700-nm wavelength range is stronger for the large inertodetr-inite region compared to the vitrinite matrix. Interestingly, adetailed analysis of the spectral maps illustrates the presence ofan interfacial layer between inertodetrinite and vitrinite inclusions,properties of which cannot be represented as a linear combinationof pure inertinite and vitrinite spectra. This finding is corroboratedby the three-wave fluorescent imaging.

To discern the internal structure of the inertodetrinite inclusion,we focused our analysis on the G- and D-band region. Spatiallyresolved Raman maps for the inertodetrinite inclusion acquiredin these spectral regions are shown in Fig. 3. The original Ramanspectra between 950 and 3200 cm�1 were deconvoluted using amixed Gaussian–Lorentzian fit with both peak and relative peakparameters analyzed. Fig. S2(a)–(f)� in Supplementary Materialshow color-coded spatial maps of the deconvoluted G- and D-peakparameters (intensity, position and FWHM) of the area of theinclusion.

Taking into account the non-graphitic nature of the inertodetr-inite inclusion, the D-band originates from the aromatic-quadrantring-breathing mode of several fused aromatic rings (more than sixand less than that of graphite), rather than assigning it to E2g modeof graphite [22,23]. The intensity of the D-band is an indicator ofdegree of disorder in a conjugated system of fused aromatic rings.D-band intensity is relatively uniform in the center of inclusion butincreases towards the matrix boundary, indicating slight upturn indisorder of a fused-ring system at the interface with vitrinite. Threedistinct areas (marked with black arrows in Fig. 3) within theinclusion have a 1.75� intensity increase of the D-band showingthe presence of smaller, damaged inclusions. The position of theD-band in the inclusion changes by almost 6 cm�1 (Fig. S2b in Sup-plementary Material), and can be related to the residual stress inthe structure. The width of the D-band remains relatively un-changed, around 60 cm�1, within the area of inclusion. For threesmaller areas, we observe distinctly different changes in D-band.Two of these areas are characterized by a broader D-band, andthe third one had smaller FWHM of the D-peak.

In turn, the G-band is assigned to E2g mode of aromatic ringquadrant breathing vibrations of fused benzene rings and alkene.The G-band position is almost unchanged across the inclusion,but shifts slightly to lower frequency at the boundary with thematrix following changes in D-band (Fig. S2e in SupplementaryMaterial). Similar to the D-band, the FWHM of the G-band isuniform (�80 cm�1) for the most part of the inclusion.

Notably, the ratio of G- to D-band intensities, IG/ID, which is ameasure of crystallinity and/or defect density of carbon structures,stays relatively uniform and slightly above unity across the inclu-sion as seen in Fig. 3c [24,25]. Note that even for three smaller

Fig. 3. Spatially resolved Raman 50 � 50 lm2 maps for the inertodetrinite inclusion. (a) T(c) the ratio of intensities of G- and D-bands. White square marks the reference feature

inclusions marked with black arrows in Fig. 3, the IG/ID ratio doesnot change much, indicating that the crystallinity of smaller inclu-sions is almost the same as the large one. We therefore concludethat smaller inclusions are of the same nature as the large particle,but are more oxidized.

3.2. Near-field microwave microscopy

To explore the structure of the sample at sub-100-nm resolu-tion, the same region was mapped with a near-field scanningmicrowave microscope [26]. The dielectric properties of coals atmacroscopic length scales (�meters) are strongly dependent onrank and are used in coal seam microwave radiation mapping dur-ing exploratory drilling and radar mapping [27]. On smaller scales,the dielectric properties are intrinsically linked to coal structureand behavior. For example, transition from lignite to anthracitealong the coalification pathway leads to the increased formationof polyaromatic rings and ultimately with regions of structuresbecoming graphite-like, with an associated increase of conduc-tance and dielectric constant. The microwave response is alsolinked to water in pores, minerals, etc. Correspondingly, there aremultiple microwave studies of coals [28–36]; as well as studiesof coal transformations during pyrolysis, relevant to dynamicsand mechanism of gasification, liquefaction, and coke formation[37,38]. However, similar to chemical studies, macroscopic micro-wave mapping of coal is subject to multiple uncertainties in inter-pretation due to the complex inhomogeneous nature of thematerial. The measured dielectric properties of coal are the resultof averaging over dielectric properties of all minerals and maceralspresent within a coal specimen. With the ability to map dielectricproperties at the nanometer-length scale and using the differencesin dielectric properties of coal constituents, SMM can be straight-forwardly employed to study coal morphologies at sub-microme-ter-length scales.

Based on the differences in electrical conductivity, dielectricconstant, and density, SMM provides a way for mapping differentmacerals in coal specimens with a spatial resolution of about50 nm, which far exceeds the resolution of the fluorescent micros-copy and micro-Raman spectroscopy. SMM images of the inerto-detrinite fragment and the surrounding vitrinite matrix aredisplayed in Fig. 4 together with corresponding topographicimages. In the SMM images, brighter colors correspond to smallerabsolute values of complex permittivity. The inertodetrinite inclu-sion has a relatively high electrical conductivity associated withhighly conjugated, fused aromatic rings of the material and, there-fore, this inclusion appears darker in the microwave images. Inturn, the dielectric permittivity of the elongated, fiber-like inclu-sions in the surrounding vitrinite matrix which show a strongerfluorescence in Fig. 2a, is lower, corresponding to a less densedielectric material. A higher resolution image over a feature atthe boundary of the inertodetrinite inclusion (marked by the

he integrated intensity of the G- band; (b) the integrated intensity of the D-band andon the sample. Small inclusions within fusinite are marked with black arrows.

Fig. 4. Near-field microwave microscopy of a coal specimen: (a–c) topographic AFM images of the area with the inertodetrinite inclusion in the vitrinite matrix [39]. Panels(d–f) show corresponding SMM images obtained at a frequency of 5.2 GHz. The dashed lines in panels (a and b) outline areas imaged in (b and c), respectively.

36 A. Tselev et al. / Fuel 126 (2014) 32–37

dashed line in Fig. 4a) reveals morphological details at its slope.There are apparent elongated inclusions, appearing as white verti-cal lines in Fig. 4e, which have lower permittivity than inertodetr-inite. The slope of the inclusion is covered by the material of thelow-permittivity vitrinite matrix. The SMM allows imaging withan even higher resolution—the image in Fig. 4f displays furtherinhomogeneities in the structure at the length scale of about100 nm.

As seen, overall, the SMM is able of visualizing morphologicaldetails of the interface between different macerals with a spatialresolution of about 100 nm, and revealing the nature of submi-cron-scale inhomogeneities in the coal specimen. SMM providesa viable novel approach for studies of internal structure of coals,which can be used in studies of maceral reactivity, thermal stabil-ity, graphitization behavior with a spatial resolution unattainablethus far.

4. Conclusions

In summary, while the macroscopic properties of coal havebeen studied for over a century, the intrinsic inhomogeneity of thismaterial has long precluded knowledge-driven studies, and manyfundamental structures of coal remain unresolved. The combina-tion of microwave imaging assisted by Raman microscopy allowsexploration of the coal structure below the 100-nm barrier ofspatial resolution, providing novel tools to obtain fundamentalknowledge of the material organization. These studies can be fur-ther extended to other SPM methods to provide a comprehensivepicture of coal functionality and chemistry. This information isindispensable in developing rapid coal identification based oncomponent mapping, classification of coal’s chemical and oxidativereactivity, and for the predictive modeling to develop novel,environmentally-safe ways to use coal.

Acknowledgements

This research was conducted at the Center for NanophaseMaterials Sciences, which is sponsored at Oak Ridge NationalLaboratory by the Scientific User Facilities Division, Office of BasicEnergy Sciences, US. Department of Energy, as well as Leone FamilyDepartment of Energy and Mineral Engineering, and the EMSEnergy Institute, The Pennsylvania State University, PA.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.fuel.2014.02.029.

References

[1] van Heek KH. Progress of coal science in the 20th century. Fuel 2000;79:1–26.[2] Potgieter-Vermaak S, Maledi N, Wagner N, Van Heerden JHP, Van Grieken R,

Potgieter JH, et al. Raman spectroscopy for the analysis of coal: a review.Raman Spectrosc 2011;42:123–9.

[3] Oliveria MLS, Waanders F, Silva LFO, Jasper A, Sampario CH, McHabe D, et al. Amulti-analytical approach to understand the chemistry of Fe-minerals in feedcoals and ashes. Coal Combust Gasification Prod 2011;3:51–62.

[4] Silva LFO, Sampaio CH, Guedes A, Fdez-Ortiz de Vallejuelo S, Madariaga JM.Multianalytical approaches to the characterisation of minerals associated withcoals and the diagnosis of their potential risk by using combined instrumentalmicrospectroscopic techniques and thermodynamic speciation. Fuel2012;94:52–63.

[5] Taylor GH, Teichmüller M, Davis A, Diessel CFK, Littke R, Robert P. Organicpetrology: A new handbook incorporating some revised parts of Stach’stextbook of coal petrology. Berlin: Gebrüder Borntraeger; 1998.

[6] Morga R. Reactivity of semifusinite and fusinite in the view of micro-Ramanspectroscopy examination. Int J Coal Geol 2011;88:194–203.

[7] Morga R. Micro-Raman spectroscopy of carbonized semifusinite and fusinite.Int J Coal Geol 2011;87:253–67.

[8] Marques M, Suárez-Ruiz I, Flores D, Guedes A, Rodrigues S. Correlationbetween optical, chemical and micro-structural parameters of high-rank coalsand graphite. Int J Coal Geol 2009;77:377–82.

[9] Guedes A, Valentim B, Prieto AC, Rodrigues S, Noronha F. Micro-Ramanspectroscopy of collotelinite, fusinite and macrinite. Int J Coal Geol2010;83:415–22.

[10] Guedes A, Valentim B, Rodrigues S, Costa A, Marques M, Flores D. Micro-Ramanspectroscopy and optical reflectance studies of coals with different rank.Geochim Cosmochim Acta 2007;71. A361-A361.

[11] Livneh T, Bar-Ziv E, Senneca O, Salatino P. Evolution of reactivity of highlyporous chars from Raman microscopy. Combust Sci Technol 2000;153:65–82.

[12] Green PD, Johnson CA, Thomas KM. Applications of laser Raman microprobespectroscopy to the characterization of coals and cokes. Fuel1983;62:1013–23.

[13] Guedes A, Valentim B, Prieto AC, Sanz A, Flores D, Noronha F. Characterizationof fly ash from a power plant and surroundings by micro-Raman spectroscopy.Int J Coal Geol 2008;73:359–70.

[14] Guo Y. Bustin RM, micro-FTIR spectroscopy of liptinite macerals in coal. Int JCoal Geol 1998;36:259–75.

[15] Manoj B, Kunjomana A, Chandrasekharan K. Chemical leaching of low rankcoal and its characterization using SEM/EDAX and FTIR. J Miner Mater CharactEng 2009;8:821–32.

[16] Mastalerz M, Marc Bustin R. Electron microprobe and micro-FTIR analysesapplied to maceral chemistry. Int J Coal Geol 1993;24:333–45.

[17] Sonibare OO, Haeger T, Foley SF. Structural characterization of Nigerian coalsby X-ray diffraction, Raman and FTIR spectroscopy. Energy 2010;35:5347–53.

[18] Mastalerz M, Bustin RM. Variation in elemental composition of macerals; anexample of the application of electron microprobe to coal studies. Int J CoalGeol 1993;22:83–99.

A. Tselev et al. / Fuel 126 (2014) 32–37 37

[19] Tselev A, Strelcov E, Luk’yanchuk IA, Budai JD, Tischler JZ, Ivanov IN, et al.Interplay between ferroelastic and metal-insulator phase transitions in strainedquasi-two-dimensional VO2 nanoplatelets. Nano Lett 2010;10:2003–11.

[20] Tselev A, Meunier V, Strelcov E, Shelton WA, Luk’yanchuk IA, Jones K, et al.Mesoscopic metal-insulator transition at ferroelastic domain walls in VO2. ACSNano 2010;4:4412–9.

[21] Dobigeon N, Moussaoui S, Coulon M, Tourneret JY, Hero AO. Joint Bayesianendmember extraction and linear unmixing for hyperspectral imagery. IEEE TSignal Proc 2009;57:4355–68.

[22] Li X, Hayashi J-I, Li C-Z. FT-Raman spectroscopic study of the evolution of charstructure during the pyrolysis of a Victorian brown coal. Fuel 2006;85:1700–7.

[23] Li C-Z. Some recent advances in the understanding of the pyrolysis andgasification behavior of Victorian brown coal. Fuel 2007;86:1664–83.

[24] Ferrari AC, Basko DM. Raman spectroscopy as a versatile tool for studying theproperties of graphene. Nature Nanotechnol 2013;8:235–46.

[25] Ivanov I, Puretzky A, Eres G, Wang H, Pan Z, Cui H, et al. Fast and highlyanisotropic thermal transport through vertically aligned carbon nanotubearrays. Appl Phys Lett 2006;89:223110–2.

[26] Tselev A, Lavrik NV, Vlassiouk I, Briggs DP, Rutgers M, Proksch R, et al. Near-field microwave scanning probe imaging of conductivity inhomogeneities inCVD graphene. Nanotechnology 2012;23:385706.

[27] Ralston JC, Hainsworth DW, Use of ground penetrating radar in undergroundcoal mining. In: Proc. SPIE, eighth international conference on groundpenetrating radar. 2000; 4084: 731.

[28] Fornies-Marquina JM, Martin JC, Martinez JP, Miranda JL, Romero C. Dielectriccharacterization of coals. Can J Phys 2003;81:599–610.

[29] Marland S, Merchant A, Rowson N. Dielectric properties of coal. Fuel2001;80:1839–49.

[30] Mackinnon AJ, Hayward D, Hall PJ, Pethrick RA. Temperature-dependent low-frequency dielectric and conductivity measurements of Argonne premiumcoals. Fuel 1994;73:731–7.

[31] Mackinnon AJ, Hall PJ, Hayward D, Pethrick RA. Low-frequency dielectricmeasurements on Argonne premium coals. Fuel 1993;72:1077–8.

[32] Chatterjee I, Misra M. Dielectric-properties of various ranks of coal. JMicrowave Power EE 1990;25:224–9.

[33] Nelson SO, Fanslow GE, Bluhm DD. Frequency-dependence of the dielectric-properties of coal. J Microwave Power EE 1980;15:277–82.

[34] Miyasita I, Higasi K, Kugo M. Dielectric investigation on coals 3. On a variety ofcoals. B Chem Soc Jpn 1957;30:550–5.

[35] Miyasita I, Higasi K. Dielectric investigation on coals. 1. Dielectric properties ofJapanese coals. B Chem Soc Jpn 1957;30:513–7.

[36] Groenewege MP, Schuyer J, Vankrevelen DW. Chemical structure andproperties of coal 10. Dielectric constants of low rank and bituminous coals.Fuel 1955;34:339–44.

[37] Zubkova V, Prezhdo V. Change in electric and dielectric properties of someAustralian coals during the processes of pyrolysis. J Anal Appl Pyrol2006;75:140–9.

[38] Peng ZW, Hwang JY, Kim BG, Mouris J, Hutcheon R. Microwave absorptioncapability of high volatile bituminous coal during pyrolysis. Energy Fuel2012;26:5146–51.

[39] Sykorova I, Pickel W, Christanis K, Wolf M. Taylor GH, flores D, ICCP, the newvitrinite classification (ICCP system 1994). Fuel 1994;77:349–58.