understanding the raman spectral features of phyllosilicates€¦ · their source origins, and the...

Research article

Received: 30 September 2014 Revised: 8 February 2015 Accepted: 13 February 2015 Published online in Wiley Online Library

(wileyonlinelibrary.com) DOI 10.1002/jrs.4680

Understanding the Raman spectral featuresof phyllosilicatesAlian Wang,a,b* John J. Freemana,b and Bradley L. Jolliffa,b

Phyllosilicates, one of the five major structural types of silicates, have highly variable structures and very complex chemistry. Thismanuscript reports the first comprehensive Raman spectroscopic study of the five major groups of phyllosilicates. We first dem-
onstrate that phyllosilicates have a unique Raman spectral pattern that can be used to distinguish them from orthosilicate, chainsilicate, ring silicate, and framework silicate with different degrees of silicate polymerization. Second, we distinguish characteris-tics of Raman spectral features associated with phyllosilicates of the different stacking sequences and showhowminor changes inchemical compositions and structural details in some phyllosilicates can affect their Raman spectral patterns, peak positions, andpeak widths. From this study, we extract several empirical rules that can help to identify and characterize phyllosilicates based onin situ Raman spectroscopic measurements. These results are significant for planetary surface exploration, especially Mars, wherethe existence of phyllosilicates as a result of alteration of primary minerals has been indicated by Vis-Near IR spectroscopy used inorbital remote sensing and recently by X-ray diffraction in surface exploration. The spectral data collected from this study are alsouseful for laboratory study of terrestrial phyllosilicate samples. Copyright © 2015 John Wiley & Sons, Ltd.
Additional supporting information may be found in the online version of this article at the publisher’s web site.

Keywords: Raman spectroscopy; phyllosilicates; planetary exploration; Mars

* Correspondence to: Alian Wang, Rudolph Hall, room 110, Department of Earthand Planetary Sciences, Washington University in St. Louis, One Brookings Drive,St. Louis, MO, 63130, USA.E-mail: [email protected]

a Department of Earth and Planetary Sciences, Washington University in St. Louis,St. Louis, MO, USA

b McDonnell Center for Space Sciences, Washington University in St. Louis, St. Louis,MO, USA

Introduction

Phyllosilicates on Earth can be igneous, metamorphic, or sedimen-tary in origin. Over 240 minerals are classified as phyllosilicates bycommon mineralogy databases. In recent years, using orbital re-mote sensing techniques, some phyllosilicates have been reportedto exist at the surface of Mars.[1–5] More recently, trioctahedralsmectites were identified in mudstone-type rocks at YellowknifeBay in Gale Crater on Mars using X-ray diffraction (XRD) techniques(CheMin on Curiosity Rover[6,7]). The occurrence of these mineralsand their geochemical features is indicators of environmentalconditions during their formation in the past and their preservationin the geologic history of the planet. Such findings raise theimportance of definitive identification and characterization ofphyllosilicates during future landed missions, such as the EuropeanSpace Agency 2018 ExoMars and NASA’s 2020 rover mission toMars.

Laser Raman spectroscopy is a powerful technique that has beenproposed for mineral characterization for planetary surface explora-tion missions since 1995.[8] Several laser Raman units for flight havebeen under development. The scientific payload of the 2018ExoMarsmission includes an in situ laser Raman spectrometer using532-nm excitation.[9,10] Two Raman systems, a remote Raman unitusing 532-nm excitation[11] and an in situ Raman unit using 248-nmexcitation,[12] were selected recently for the Mars 2020 missionpayload. A combined team at Washington University in St. Louisand at the Jet Propulsion Laboratory has also been developing anin situ Raman unit [the Mars Microbeam Raman Spectrometer(MMRS) and Compact Integrated Raman Spectrometer (CIRS)[13–16]],which was tested on the Zoë rover in the Atacama Desert[17] forautonomous subsurface materials analysis. This development ef-fort is supported by the NASA MatISSE program to rapidly reachflight readiness.[18]

J. Raman Spectrosc. (2015)

One of the major contributions of the MMRS-CIRS team to Plan-etary Raman Spectroscopy was to fill the knowledge gap for majorRaman spectral features of a variety of mineral groups andmolecu-lar species relevant to planetary surface exploration missions, e.g.Mars, the Moon, Venus, and asteroids.[8,19–21] Among the five majorgroups of silicates, phyllosilicates have been extremely challengingbecause they have themost complex structures and highly variablecompositions, originating from different types of primary melt andfluid compositions and chemical alteration. On the other hand, toidentify and characterize phyllosilicates, in situ during planetary sur-face explorationmissions, is critical to understanding theweatheringprocesses on Mars, asteroids, and other planetary bodies in ourSolar System.

In literature until now, there has been a lack of an overview, acomparison, and an understanding of the major Raman spectralfeatures of phyllosilicates. Raman spectra of some specificphyllosilicates have been published,[22–24] emphasizing the funda-mental vibration modes of the silicate lattice of phyllosilicates.Spectral information is also available on H2O/OH modes.[25–32] Inaddition, the Raman spectra of numerous phyllosilicates are in-cluded in the RRUFF database.[33] We have been systematicallystudying the Raman spectral features of the five major subgroupsof phyllosilicates since 2002, first through a preliminary set of

Copyright © 2015 John Wiley & Sons, Ltd.

A. Wang, J. J. Freeman and B. L. Jolliff

studies[34,35] to demonstrate the feasibility of classifying thephyllosilicates on the basis of their Raman spectral features andthen to work out the details of each subgroup. This manuscriptreports the results from over 12 years of investigations. This workincludes the following five steps:

(1) Obtaining high-quality Raman spectral data on pure mineralspecimens of the five major subgroups of phyllosilicates thathave been previously characterized by other mineralogicaltechniques and thenmatching our spectra with the availableRaman data in the literature (Sections on Phyllosilicate Sam-ples and Experiments);

(2) Identifying the major spectral characteristics of thesephyllosilicates that can be used for identification and classifi-cation when encountered in geological samples (Section onOverview: Raman Spectral Features for Classification ofPhyllosilicates);

(3) Developing an in-depth understanding of the structural andcompositional causes for the major Raman spectral featuresfor the five major subgroups of phyllosilicates (Section onUnderstanding the Raman Spectral Features of Five Groupsof Phyllosilicates);

(4) Developing empirical rules to extract information aboutchemical compositions of some phyllosilicates on the basisof their Raman spectra (Section on Empirical Rules forExtracting Chemical Characters of Some Phyllosilicates onthe Basis of their Raman Spectra);

(5) Understanding the special challenges in characterizingphyllosilicates and potential differences between the physicalcharacteristics of terrestrial and extraterrestrial phyllosilicatesand implications for their identification in robotic planetaryexploration (Section on Challenges for Identifying and Char-acterizing Phyllosilicates on Earth and on Other Planets).

Phyllosilicate samples

Phyllosilicate samples used in this study were selected from themineral collection of the Department of Earth and Planetary Sci-ences (EPSC) at Washington University in St. Louis, from the per-sonal collections of the authors and colleagues in EPSC and fromthe Society of Clay Minerals (CMS). Table S1 lists all samples whoseRaman spectra were presented in the figures of this manuscript,their source origins, and the basis of their identifications. Table S2lists the typical Raman peak positions of major minerals in eachphyllosilicate group. Table S3 lists the chemical compositions ofmost of the samples.These samples fall into three categories on the basis of their

sources:

(1) Standard samples – There are eight standard clay mineralsthat we purchased from the CMS. Their chemistries andstructures have been confirmed by earlier studies.[36,37] Rele-vant data can be found online at the CMS website.

(2) Well-characterized samples from the EPSC collection and per-sonal collections – There are 33 mineral samples collectedby EPSC researchers and by the authors, and previously iden-tified, at a minimum, using optical microscopy and XRD. Forthis study, we have used electron probemicroanalysis (EPMA)and XRD to reconfirm their identities (Tables S1 and S3).

(3) Well-characterized samples provided by the Thermal EmissionSpectroscopy (TES) teamat Arizona State University – two sam-ples are from a collection of pure mineral samples kindly

wileyonlinelibrary.com/journal/jrs Copyright © 2015 J

provided by the TES team led by Professor P. Christensenat Arizona State University (Table S1), selected from theirstandard mineral collections. Most of their samples have adetermined elemental composition and they have beenidentified by XRD. Their thermal emission spectra were mea-sured by the TES team at ASU.

Experiments

Laser Raman spectroscopic measurements

Two laboratory laser Raman spectrometers were used to measurethe Raman spectra. The first is a Hololab 5000 Raman system (KaiserOptical Systems, Inc., Ann Arbor, MI, USA) using 532-nm excitation.The second is a multi-wavelength inVia Raman system (RenishawCompany, New Mills Wotton-under-Edge Gloucestershire GL128JR United Kingdom).

The Hololab 5000 uses a frequency-doubled Nd:YAG solid-statelaser as the excitation source and a transmissive holographic grat-ing spectrometer that covers the Raman Stokes shift range of ~0to 4300 cm�1 relative to the 532-nm laser line. The spectrometerhas a spectral resolution of 4–5 cm�1 and uses a 256×2048 pixelarray Charge-Coupled Device (CCD) camera for recording Ramanspectra. The 532-nm laser radiation is sent through a single-modeoptical fiber (~8μm in diameter) into a probe head attached tothe optical microscope of the HoloLab 5000. A microscope objec-tive condenses the laser beam onto the sample and collects theback-scattered Raman signal. The Raman signal is returned to thespectrograph through a multimode optical fiber (100μm in diame-ter) without the use of a depolarizer. The optical microscope is alsoused for viewing and photographing the sample with either trans-mitted or reflected light. Two objectives were used in this Ramanstudy: (a) 20× (NA 0.4) with a working distance of 12mm that pro-duces a laser beam diameter of ~6μm at focus and (b) 50× (NA0.75) with working distance of 0.5mm that produces a laser beamdiameter of ~2μm at focus. The power of the laser beam at thesample was measured to be ~13mW.

The inVia spectrometer (Renishaw company) is a laser Ramanimaging system with five excitation wavelengths: 785-nm line of adiode laser, 632.8-nm line of a He–Ne laser, 532-nm line of a diodepumped solid state laser, 442-nm line, and 325-nm line of a He–Cdlaser. A Raman system with these multiple excitation laser wave-lengths facilitates the study of a wide variety of materials, becauseRaman cross sections of materials change with excitation wave-lengths. Some laser wavelengths enhance Raman signals by induc-ing Raman resonance effects. Some laser wavelengths minimizethe interference of fluorescence generated in terrestrial samples.For the latter reason, we used the 532-nm, 442-nm, and 325-nm la-ser lines for obtaining the Raman spectra from some phyllosilicatesin this study. We find that changing excitation laser wavelength cansometimes reduce the interference, but not always, because thesources of fluorescence emissions are highly variable. There is nobest wavelength for all cases. However, 532nm, or excitation linesin the green-blue range, has the overall best performance. The inViasystem uses two reflective holographic gratings to cover the RamanStokes shift range of ~50 to 4000 cm�1 relative to three selectedlaser lines. Depending on the selection of laser line and grating,1-cm�1 spectral resolution can be obtained. In this system, the laserbeam is directly coupled intomicroscope, focused onto the sample,and the generated Raman photons are transmitted through thespectrograph to the CCD camera without the use of fiber optics.The objective used for the current study is a 50× (NA 0.5) with a

ohn Wiley & Sons, Ltd. J. Raman Spectrosc. (2015)

Raman spectral features of phyllosilicates

working distance of 8.2mm. The laser power at the sample was ad-justable using combinations of two sets of neutral density filters.

The Raman measurements using both systems were made oneither loose grains of phyllosilicates without preparation or on apressed pellet (13-mm or 25-mm diameter) of sample powder.The calibration of wavelength and spectral response for bothRaman systems was made with a neon emission spectral calibra-tion lamp and a NIST (National Institute of Standards and Technol-ogy) secondary standard white light. The zero Raman-shiftfrequency of the excitation laser was calibrated daily beforeacquiring any sample spectra by measuring and setting theRaman peak position of single crystal of silicon to 520.7 cm�1.These calibrations ensured that measurement errors on Ramanpeak positions were less than ±1 cm�1. The major Raman peakpositions and relative peak intensities for the five groups ofanalyzed phyllosilicates are presented in Table S2.

Electron-microprobe analyses

Individual grains of phyllosilicate samples selected from the samesamples used for Raman spectroscopy were prepared for EPMAby mounting in epoxy on glass slides and grinding/polishing thesurfaces to a final polish using 0.25-μm diamond paste.

The EPMA analyses were carried out using a JEOL 733 Superprobe(JEOL Ltd., Akishima, Tokyo,Japan ) equipped with three wavelengthdispersive spectrometers, a back-scattered electron detector, and

Advanced Microbeam™

automation (Advanced MicroBeam,Vienna, OH, USA). An accelerating voltage of 15 kV, a beam cur-rent of 20 nA, and a defocused beam (10-μm spot size) were usedto prevent loss of Na. We used a combination of oxides and sili-cate standards to calibrate and monitor the EPMA data collection.A modified Armstrong (1988) CITZAF routine incorporated intothe EPMA software was used for X-ray matrix corrections. Molarproportions of the cations were calculated from the measuredweight percentages of the corresponding oxides on the basis often oxygen atoms per formula unit. The elemental chemistry(as oxides) and derived structural formulae for the analyzedphyllosilicate samples are presented in Table S3.

XRD powder-diffraction analyses

Additional XRDmeasurements weremade on selected phyllosilicatesamples that needed structural characterization (Column 9 inTable S1). The samples were prepared as follows: Several smallpieces of each sample were hand ground and slurred with(acetone or alcohol) with a mortar and pestle. The slurry wasplaced onto an XRD powder sample holder, dried, and analyzedas powder mounts. The XRD analyses were carried out with aRigaku Geigerflex D–MAX/A diffractometer. This XRD system usesa CuKa radiation (35 kV, 35mA) and using a Bragg–Brentano fo-cusing geometry with a 1° incident aperture slit, a 0.8° detectorslit, and a scintillation counter as the detector. We used a 2θ rangeof 4–70°, a 2θ step size of 0.04°, and a 1-s dwell time per step. Thedata were reduced and analyzed using the JadeTM software(version 3.1, Materials Data, Inc., Livermore, CA, USA).

Overview: Raman spectral features for classifi-cation of phyllosilicates

It is commonly observed that the major Raman peaks of a mineralarise from the chemical bond or bonds within an ionic group that

J. Raman Spectrosc. (2015) Copyright © 2015 John Wiley

have the highest degree of covalency in a mineral structure, e.g.(NO3)

� in nitrates, (CO3)2� in carbonates, (SO4)

2� in sulfates, (PO4)3�

in phosphates, and (SiO4)4� in silicates, because the degree of

covalence of a chemical bond influences the force constant ofthe bond or the rigidity of the polyhedra in the mineral structure,which are critical factors in determining their Raman spectralfeatures.

Differences in electronegativity between the elements in achemical bond are commonly used to estimate the degree of cova-lence (or polarity) of that chemical bond.[38] For example, comparedwith an Si–O bond in the (SiO4)4� units of silicates that has about51% degree of covalence, the P–O bond in the (PO4)3� unit ofphosphates has ~65%; and the N–O, C–O, and S–O bonds innitrates, carbonates, and sulfates have 92–80%. Thus, in general, sil-icates are medium Raman scatterers when compared with nitrates,carbonates, sulfates, and phosphates. Fe oxides are even weakerRaman scatterers than silicates. Using similar reasoning, the vibra-tional modes of isolated (SiO4)

4� tetrahedral units in orthosilicates,or polymerized (SixOy)

z� in chain silicate, ring silicate, phyllosilicate,and tectosilicate, contribute the major peaks in their Raman spec-tra, whereas the M–O bonds (i.e. Fe–O, Mg–O, Ca–O, and Al–O ) inoctahedral units have much lower degrees of covalency and thusproduce much weaker Raman signals. For these reasons, the de-grees of polymerization of silicates, i.e. the framework of (SiO4)

4�

or (SixOy)z�, determine the major features of their Raman spectral

patterns.[39,40]

Determining the degrees of polymerization of silicates fromthe Raman spectral patterns

The degree of polymerization of a silicate can be quantified by thenumber of bridging oxygen atoms (#Ob) of the Si–Ob–Si bonds in itsstructural framework. Figure 1 shows a set of typical Raman spectraof orthosilicate, chain silicate, double chain silicate, phyllosilicate,and tectosilicate with the #Ob of 0, 2, 2.5, 3, and 4, respectively. Thisfigure demonstrates that phyllosilicates (#Ob=3) can be distin-guished from other silicates based on their Raman spectral patternin a straightforward way. The figure shows that the increase of de-gree of polymerization induces the following changes in Ramanspectra:

(1) A gradual reduction of intensity for peaks in the spectral re-gion 800–1100 cm�1 where the major peaks are assigned tosymmetric stretching vibration of Si–Onb (Onb=non-bridgingoxygen with another Si) bonds in the (SixOy)

z� units oforthosilicate, chain silicate, ring silicate, and phyllosilicate;

(2) A gradual shift to higher wavenumber of the peak positionsin the 800–1100 cm�1 spectral region;

(3) The appearance of a new peak in the spectral region 600–750 cm�1 or the gradual increase of peak intensity in thisregion compared with those in the spectral region 800–1100 cm�1 for chain silicate and phyllosilicate, assigned tothe Si–Ob–Si vibration mode;

(4) The appearance of peaks in the spectral region <600 cm�1

for tectosilicates assigned to the breathing vibration modeof the T–O–T substructure and correlates with the size ofthe ring made by the TO4 tetrahedra, where T= Si orAl.[41–44]

Regarding the aforementioned effect (2), Fig. 2 shows a correlationbetween the major Raman peak position in 900–1100 cm�1 spectralrange against the average Si–O bond length (in Å, over Si–Ob andSi–Onb bonds ) of representative silicates with four different degrees

& Sons, Ltd. wileyonlinelibrary.com/journal/jrs

Figure 1. Raman spectra of silicates with different degrees of polymerization: pyrope (orthosilicate, #Ob = 0), pyroxene (chain silicate, #Ob = 2),cummingtonite (double chain silicate, #Ob = 2.5), muscovite (phyllosilicate, #Ob = 3), and albite (tectosilicate, #Ob = 4).


of polymerization: pyrope (Mg3Al2Si3O12), enstatite (Mg2Si2O6),anthophyllite (Mg7Si8O22(OH)2), and talc (Mg3Si4O10(OH)2). A verysimilar Raman peak position upshift was found in silicate glassesfollowing the increase of polymerization by McMillan.[26] Thiscorrelation suggests a general trend of Raman peak positionupshift following the increased sharing of oxygen(s) in the (SiO4)tetrahedron.Phyllosilicates (#Ob=3) can be distinguished from other silicates

on the basis of their Raman spectral patterns in a straightforwardway (Fig. 1). All phyllosilicates would have their strongest Ramanpeaks in the 600–750 cm�1 spectral range and a set of weak Ramanpeaks in the 800–1100 cm�1 spectral range with peak positionslightly upshifted if compared with similar peaks of orthosilicateand chain silicate. Similar to our previous studies of the Raman spec-tral features of orthosilicate, chain silicate, and tectosilicate,[45,46,44]

the current study involves correlating changes in the Raman spectraof the phyllosilicates with changes of the phyllosilicate frameworkcaused by structural and chemical variations, especially variations

wileyonlinelibrary.com/journal/jrs Copyright © 201

of the cation occupancy in adjacent octahedral sites of the (SixOy)z�

sheets.The major Raman peak positions and relative peak intensities

for the five groups of phyllosilicates that we analyzed are presentedin Table S2. We used the commonly accepted peak assignmentsfor the major peaks in the range of 1100–900 cm�1 for Si–Onb

vibrations and 750–600 cm�1 for Si–Ob–Si vibrations for thosephyllosilicates.

Extracting the chemistry of silicate solid solutions from Ramanpeak positions

Although the Raman spectral pattern and the approximate posi-tions of major Raman peaks of silicates are determined by Si–Obonding parameters in (SixOy)

z� units, the properties of cations(Mg, Fe, Ca, Al, etc.) that occupy adjacent octahedral sites would af-fect the exact position of the Raman peak contributed by (SixOy)

z�

units, because these octahedral sites share coordination oxygen

5 John Wiley & Sons, Ltd. J. Raman Spectrosc. (2015)

Figure 2. A correlation trend exists between the Raman peak positions andthe average Si–O bond length (in Å) in major groups of silicates withdifferent degrees of polymerization.


atom(s) with Si4+, noted as a Si–Onb–M bonding (M¼Mg, Fe, Ca, Al,etc.). For this reason, Raman peak positions of silicates as a functionof cation ratios have been used to develop calibration formulae forderiving chemical information about a variety of silicates and otherminerals that occur as solid solutions.

For example, Raman peak positions were used to extractMg/(Mg+Fe) in olivine,[45] Mg/(Mg+Fe+Ca) in pyroxene,[46] cationratios in X3Al2(SiO4)3 (X¼Mg, Fe, Mn) and in Ca3Y2(SiO4)3 (Y¼Cr,Al, Fe3+) garnets,[47] Fe/(Fe+ Ti +Cr) in Fe oxides,[48] and Mg/(Mg+Fe+Mn+Ca) in carbonates.[49] Raman peak positions of feldsparswere also used to classify feldspars among the K-end, Na-end, andCa-end members (orthoclase, albite, and anorthite).[44] However,the extracted information on the solid solutions of feldspar haslimited accuracy when compared with those from olivine and py-roxene, because in feldspar structures, the distance from the coor-dinated oxygen in (SixOy)

z� units to K+, Na+, and Ca2+ cations islonger in feldspar than those in Mg2+, Fe2+, and Ca2+ in olivineand pyroxene. Therefore, changes among K+, Na+, and Ca2+ in feld-spars have less influence on the vibrational modes of (SixOy)

z�

units. These observations on the effect of cation substitutions onRaman peak position shifts in olivine, garnet, pyroxene, and feld-spar suggest that the effect of nearest neighbors, i.e. the cation inoctahedral sites that directly share the coordinating oxygen with(SixOy)

z� units, should be the first factor to consider for the current

Figure 3. Three major Raman spectral features that can be used to distinguishtalc-pyrophyllite group, and to distinguish OH-bearing and H2O-bearing phyllos


study of Raman features of phyllosilicates (Section on Understand-ing the Raman spectral features of five groups of phyllosilicates).

Distinguishing the dioctahedral and trioctahedralphyllosilicates

Dioctahedral and trioctahedral refer to the two common classifica-tions of phyllosilicates that differ according to the cation arrange-ment in octahedral sites. Dioctahedral phyllosilicates have two-thirds of sites in the octahedral (Oc) layer filled by Al3+ or Fe3+

cations. Trioctahedral phyllosilicates have all three sites in the Oclayer filled by Mg2+, Fe2+, and other divalent cations. For example,kaolinite (Al2[Si2O5](OH)4) and serpentine ([(Mg, Fe)3[Si2O5](OH)4)have the same stacking sequence of layers, a (Si2O5)∞ [where ∞

means a two-dimensional layer made of (Si2O5) units] tetrahedrallayer over a (MO2(OH)4)∞ Oc layer (shown in Fig. 6 for a serpentine),whereas kaolinite belongs to the dioctahedral group because Al3+

cations occupy the Oc layer, and serpentine belongs to trioctahedralgroup because Mg2+ and Fe2+ cations occupy the trioctahedrallayer.

The first criterion to distinguish dioctahedral and trioctahedralphyllosilicates based on their Raman spectra is the position ofthe strongest Raman peak (Si–Ob–Si mode) in the 800–600 cm�1

spectral range. In nearly all phyllosilicate samples measured duringthis study, we found νSi–Ob–Si> 700 cm�1 for dioctahedral andνSi–Ob–Si< 700 cm�1 for trioctahedral groups (Fig. 3a and 3b).

The mass differences of the cations that occupy the octahedralsites in many silicates are found to be the major cause of Ramanpeak position shifts, e.g. peak downshifts caused by the increaseof Fe2+ relative to Mg2+ in olivine and pyroxene in solid solutionseries,[45,46] as well as in carbonates,[49] sulfates,[50] and nitrates.[51]

This characteristic is known as the mass effect. For phyllosilicates,instead of this mass effect of cations, we found the difference in ef-fective ionic radii [52] among Al3+ (0.535 Å), Fe3+ (0.645 Å),Mg2+(0.72 Å), and Fe2+ (0.78 Å) to be one of the major factors thatinfluences Raman peak positions. In dioctahedral phyllosilicates,the M–O bonding lengths of octahedral sites are much shorterbecause of the smaller central cation (Al3+ and Fe3+) than thosein trioctahedral phyllosilicates. Figure 4 shows a distribution ofM–O and Si–O bond lengths in dioctahedral and trioctahedralphyllosilicates (including kaolinite-serpentine, talc-pyrophyllite,micas, smectites, and chlorites). The differences in Si–O bondlengths among the dioctahedral and trioctahedral groups aresubtle, whereas the lengths of M–O are clearly shorter in thedioctahedral group than those in the trioctahedral group. Thisplot confirms that the difference in M–O bond lengths is the major

trioctahedral and dioctahedral phyllosilicates, i.e. in (a) mica group and (b)ilicates, i.e. (c) comparison of OH/H2O vibrational modes in 3000–3800 cm�1.


Figure 4. Metal–O and Si–O bond lengths in octahedral and tetrahedralsites of dioctahedral (blue dots) and trioctahedral (red dots) phyllosilicates.


cause of the difference in Raman peak position of the dioctahedraland trioctahedral phyllosilicates.The second criterion for distinguishing between dioctahedral

and trioctahedral phyllosilicates based on their Raman spectra isthe positions and relative intensities of Raman peaks in the800–1100 cm�1 spectral range. We observed the peak positionsat higher wavemumber in the spectra of dioctahedral phyllosilicates(Fig. 3a and 3b). The weakening of peak intensities is less obviousin some cases.As described in the Determining the Degrees of Polymerization

of Silicates from the Raman Spectral Patterns Section, features (1)and (2), these two Raman peak feature changes were similarly ob-served following the increase of polymerization of silicates, e.g.from an orthosilicate (garnet) to a chain silicate (pyroxene). These

Figure 5. Raman spectra of three kaolinite samples with standard formula ounchanged peak at 3622 cm�1 and the fourth peak with change position betwsilicates. Peaks marked by dotted lines are from kaolinite. Peaks with * marks a


feature changes are the effect of a connected Si4+ onto the vibra-tion modes of the Si–Onb bond in a [SiO4] tetrahedron throughthe linkage of (Onb–)Si

4+–Ob–Si

4+(–Onb) in a polymerized silicate. In-stead of Si4+, when Al3+, or Fe3+, or Fe2+, or Mg2+ is part of a linkageMx+

–Onb–Si–Ob (such as in phyllosilicates), a similar type of spectralfeature change can happen but to a lesser extent. The extent ofspectral changes depends on the difference between the cova-lence of Al3+–O (45%), or Fe3+–O (41%), or Fe2+–O (31%), orMg2+–O (28%) and that of the Si–O bond (51%). For example,Al3+–Onb–Si–Ob in dioctahedral phyllosilicates would induce stron-ger spectral changes in peak intensity and peak position upshift.This understanding is evidenced by the spectral feature differencesin the 800–1100 cm�1 range shown in Fig. 3a and 3b.

Raman features of hydroxyls and structural H2O inphyllosilicates

Hydroxyl groups (OH) in phyllosilicates are normally coordinatedwith the smaller cations (Al3+, Fe3+, Fe2+, and Mg2+) that occupythe octahedral sites, which form an Oc layer of [MOx(OH)6� x]∞(where ∞ means a two-dimensional layer made of MOx(OH)6� x

units) directly linked to the tetrahedral layer (T) of [(SixOy)z–]∞, by

sharing the coordinating oxygen (forming the T–Oc or T–Oc–Tlayers. These (OH) groups, therefore, have well-defined crystallo-graphic sites in the structure. They therefore produce sharp Ramanpeaks in the spectral range of >3600 cm�1 (Figs. 3c, 5, 8, and 9 inthe Understanding the Raman Spectral Features of Five Groups ofPhyllosilicates Section).

In contrast, structural water, H2O, (as well as large cations such asK+, Na+, Ca2+, and Li+) in phyllosilicates mostly enters the largespace between stacked T–Oc–T layers or forms an [Mg(H2O)6] unitin the case of vermiculite.[53,54] In many of these cases, the crystallo-graphic sites for these structural H2O molecules are less well-defined, and the amount of H2O per molecule of a phyllosilicate

f [Al2(Si2O5)(OH)4]. (a) H2O/OH spectral range – dash lines mark the almosteen 3695 and 3685 cm�1; (b) spectral range for fundamental vibrations of

re from TiO2 phases.



can be highly variable. As a result, the Raman spectral peaks ofstructural H2O (in the range of 3700–3000 cm�1) in phyllosilicatesare broad and have variable peak intensities (Figs. 3c, 10–15 inthe Understanding the Raman Spectral Features of Five Groups ofPhyllosilicates Section).

Understanding the Raman spectral features offive groups of phyllosilicates

Phyllosilicates have extremely complex structures and highly vari-able compositions that relate to their specific origins. The purposeof this manuscript is to correlate the major spectral characteristicsof phyllosilicates to the structural and chemical variations withinthis class of minerals and ultimately to determine the feasibility ofidentifying the individual phyllosilicates during a robotic surfaceexploration on a planetary surface.

Using T to represent tetrahedra occupied by Si (and sometimesby Al) as the central cation, and Oc to represent octahedra withMg, Fe2+, Fe3+, or Al as the central cation, the phyllosilicates canbe separated into five different groups based upon different typesof stacking sequences of T and Oc. These groups are the kaolinite-serpentine group (T–Oc), pyrophyllite-talc group (T–Oc–T), micagroup (T–Oc–T–A, A¼K, Na), smectite group [T–Oc–T–B, B¼Na, K,Ca, Li, Mg(H2O)6], and chlorite group (T–Oc–T–Oc–T–Oc–T); we willdiscuss the Raman spectral characteristics of these five majorgroups of phyllosilicates that have distinctively different stackingsequences.

Raman spectral features of kaolinite-serpentine group (T–Oc)

The common structural feature of kaolinite Al2(Si2O5)(OH)4 and ser-pentine (Mg, Fe)3(Si2O5) (OH)4 in this group is the T–Oc stacking, inwhich Al3+ occupies the octahedral sites in the Oc layer of kaolinite,and Mg2+ and Fe2+ occupy the octahedral sites in the Oc layer ofserpentine.

The Raman spectra from three typical kaolinite samples areshown in Fig. 5. Among them, KGa-1 and KGa-2 are both standardkaolinite samples from the CMS. The mineral contents of the sam-ples as determined by Chipera and Bish [37] using XRD analysis are96% kaolinite and 3% TiO2. Kaolinite sample AW, 296 was collectedfrom Kao-ling, a village near Jingdezhen county, Jiangxi province ofChina (where kaolinite got its mineral name).

In the H2O/OH spectral range (3700–3000 cm�1, Fig. 5a), thethree samples all produce four Raman peaks, corresponding tothe four distinct crystallographic sites of OH groups that coordinatewith Al3+ in octahedral sites (AlO2(OH)4) of the kaolinitestructure.[55] Among the three samples, the relative intensities offour peaks and the three of four peak positions (3672, 3653, and3622 cm�1) are the same, but the position of the fourth peak variesamong the samples. It occurs at 3695 cm�1 in the spectrum ofKGa-1 but around 3685 cm�1 in the spectra of AW, 296 andKGa-2. The apparent asymmetric shape of 3685 cm�1 peak of AW,296 suggests the persistence of a 3695 cm�1 peak as a shoulder,which implies that slight variations (structural and/or composi-tional) may have induced more than four crystallographic sites forOH groups in this kaolinite sample.

In the spectral range of the fundamental vibrations of silicates(1200–100 cm�1) (Fig. 5b), the Raman spectra of KGa-1 and KGa-2are dominated by the Raman peaks from TiO2 phases: anatase inKGa-1 and brookite in KGa-2. When excluding the Raman peaksof anatase (marked by *) from the spectrum of KGa-1, all of the


remaining peaks in the <1000 cm�1 spectral range match withthe nine peaks (915, 789, 750, 460, 431, 336, 272, 246, and197 cm�1, marked by dotted lines) in the spectrum of AW, 296.The spectrum of AW, 296 is therefore used as the standard spec-trum of kaolinite in the following discussion.

Serpentine is the trioctahedral complement of kaolinite with asimilar T–Oc stacking sequence. Mg is the major cation in the octa-hedral sites of the serpentine structure, and other octahedralcations can be Fe2+, Mn2+, Al3+, and Ni2+. The ionic radii of the octa-hedral cations determines the ‘matching’ of translational period inthe Oc layer with that in T layer, which in turn influences the mor-phologic shapes of a serpentine mineral species, which can takeplanar, curved modulated, or cylindrical shapes.[56] For example, ahigh Mg (0.72Å) content would increase the curvature of Oc layer,whereas high Fe3+ (0.645 Å) would reduce it. Structurally, serpen-tine can be classified into three subgroups: antigorite, chrysotile,and lizardite.

In the spectral range for fundamental vibrations (1200–100 cm�1),all seven serpentine samples originating from different localitiesproduce very similar Raman spectral pattern (Fig. 6b). Two fea-tures are notable: (1) All spectra have the same pattern that con-tains four major Raman peaks at ~690, 390, 230, and 130 cm�1

with relative intensities of four peaks almost unchanged amongthe samples. This spectral pattern is apparently not influencedby the structural difference between the subgroups (antigorite,chrysotile, and lizardite). (2) The exact positions of these fourmajor Raman peaks among the serpentine subgroups can vary(marked by dotted lines in Fig. 6b). For example, large peakposition downshifts were observed in the spectrum of antigorite(JDP, 38-NZ-62-5) compared with other serpentine samples, e.g.7–11 cm�1 for the peak near 690 cm�1 and 10–16 cm�1 for thepeak near 389 cm�1. Table S3 shows that the composition ofantigorite (JDP, 38-NZ-62-5) (based on Page and Coleman [57]) hasa higher content of Fe2+ in the Oc layer than the other samples. Be-cause these octahedral sites share directly the coordinating oxygenwith (Si2O5)

2� units in the serpentine structure, the heavier cationFe2+ in the Oc layer can influence the vibrations of Si–O bonds bydownshifting the Raman peak positions. Raman spectral featuresof antigorite, chrysotile, and lizardite published in the RRUFF data-base are consistent with the aforementioned observations, i.e. theRaman peaks of antigorite are downshifted 5 and 10 cm�1 fromthe corresponding peaks of chrysotile (at 686 and 383 cm�1).Figure 6b demonstrates that the effect of the nearest neighbors alsoexists in the Raman spectra of phyllosilicates: the Raman spectralpattern (positions and relative intensities of major peaks) are deter-mined by the structural framework of the [(Si2O5)

2�]∞ layer and arefurther influenced by the cations in octahedral sites that directlyshare the coordinating oxygen with (Si2O5)

2� units.On the other hand, in the H2O/OH spectral range (3700–

3000 cm�1, Fig. 6a), seven serpentine samples show quite variablepeak positions (marked by dotted lines in Fig. 6a) and peak pat-terns. In order to understand these variations, we plot a modulatedcrystal structure of antigorite (based on the crystal refinement workby Capotani and Mellini [58]) in Fig. 7, where (SiO4) tetrahedra in Tlayer and (MgO4(OH)2) octahedra in Oc layer are shown, as wellas the waving of coupled T–Oc plates. Figure 7 shows that theorientation of each O–H bond is relatively fixed against each(MgO4(OH)2) octahedron. However, the orientations of O–H bondsrelative to the [(Si2O5)

2�]∞ layer in this structure are quite differentowing to the waving of the Oc and T layers. We anticipate that themultiple OH peaks in Raman spectra of different serpentine sub-groups would reflect the variations in the O–H bond orientation.


Figure 6. Raman spectra of seven serpentine samples with standard formula of [(Mg, Fe)3(Si2O5)(OH)4] but compositionally different (Table S2), fromdifferent localities. (a) The large variation in peak positions in the H2O/OH spectral range; (b) almost unchanged peak positions in the spectral range forfundamental vibrations of silicates.

Figure 7. Themodulated crystal structure of antigorite, plot wasmade using a crystallographic imaging software Diamond (version 3.2i) based on the crystalrefinement data of Capitani and Mellini (2004) for Si34 Mg48 O147 H79.


A more detailed Raman spectroscopic study on structurally well-characterized serpentine samples is needed to address thisassumption.

Raman spectral features of the pyrophyllite-talc group (T–Oc–T)

The common structural feature of pyrophyllite, Al2(Si4O10)(OH)2,and talc, Mg3(Si4O10)(OH)2, in this group is the T–Oc–T stackingsequence, i.e. an Oc layer being sandwiched by two tetrahedrallayers.In the spectral ranges of the fundamental vibrations of silicates

(1200–100 cm�1) and the H2O/OH vibrational modes (3700–3000 cm�1, Fig. 8a and 8b), two pyrophyllite samples from differentlocalities show an exact match of peak positions (707, 360, 261, 193,and 3670 cm�1), relative peak intensities, and spectral patterns. Inaddition to the major Raman peak at 3670 cm�1 of the OH group,there is an extremely weak OH peak at 3644 cm�1 in the spectra


of both samples (it is more obvious in the spectrum of pyrophyllitesample JDP, HB). In the structure of pyrophyllite,[59] the Oc layer ismade of (AlO4(OH)2), where two OH groups occur at the two endsof a shared edge between two octahedra. Crystal refinementstudy suggests that these two OH groups are equivalent, thus, py-rophyllite should have a single Raman OH peak. Because the fre-quency of the OH peak in the Raman spectrum matches theposition of the peak in the IR spectrum (3674 cm�1), crystal fieldsplitting of the OH peak is not apparent; instead, the appearanceof the additional weak Raman OH peak at 3644 cm�1 suggests thatslight structural distortion can induce non-equivalence in structuralsites between the two OH groups.

In both OH and fundamental vibration spectral ranges of threetalc [60] samples originating from different localities (Fig. 9a and9b), exact matches were found for all four major Raman peaks at1051, 675, 361, 193, and 3678 cm�1 and many minor peaks in thespectra, except the appearance of a second small OH peak(3663 cm�1) in one spectrum, probably due to a similar reason as


Figure 8. Raman spectra of two pyrophyllite samples with standard formula of [Al2(Si4O10)(OH)2], from different localities. (a) H2O/OH spectral range, anarrow indicates additional OH peak. (b) Spectral range for fundamental vibrations of silicates.

Figure 9. Raman spectra of four talc samples with standard formula of [Mg3(Si4O10)(OH)2], from different localities. (a) H2O/OH spectral range, with an arrowindicating an additional OH peak; (b) spectral range for fundamental vibrations of silicates. The spectrum of an Fe-rich talc sample, minnesotaite, shows somemajor spectral changes in both spectral ranges.


for additional peaks in pyrophyllite spectra (Fig. 8). Nevertheless,when Fe enters the talc structure, such as the solid solution oftalc Mg3(Si4O10)(OH)2 and minnesotaite Fe3Si4O10(OH)2 (sampleAW,1208, with an actual chemical formulae (Fe2+4.61Mg1.25) Si8O20

(OH)4, Table S3), major spectral differences are observed (bottom


spectrum in Fig. 9a and 9b). First, there are general Raman peak po-sition downshifts in both spectral regions owing to the mass effect,from those peak positions of talc (Table S2) to 1035, 661, 347, 186,and 3639 cm�1. Furthermore, additional peaks appeared, such as ashoulder at 629 cm�1 and three peaks at 545, 445, and 404 cm�1 in



the fundamental vibrational range (Fig. 9b) and a triplet OH peakwith component peaks at 3655, 3639, and 3626 cm�1 (Fig. 9a) inthe OH spectral range. In the structure of this solid solution, bothFe2+ and Mg2+ would occupy octahedral sites, forming [FeO4

(OH)2] and [MgO4(OH)2], which would generate different OH vibra-tional modes. In addition, the effects of Fe2+ and Mg2+ in the Oclayer on neighboring Si–O bonds in T layers can also induce newRaman peaks.

Raman spectral features of themica group (T–Oc–T–A, A¼K, Na)

The common structural feature of muscovite, KAl2(AlSi3O10)(OH)2,phlogopite, KMg3(AlSi3O10)(OH)2, and biotite, KFe3(AlSi3O10)(OH)2,is the stacking sequence T–Oc–T –A, where the large cation A(normally Na+ or K+) is sandwiched in the space between theT–Oc–T layers.[61,62]

In the spectral range for fundamental vibrations (1200–100 cm�1)(Fig. 10b), three muscovite samples from different localities and aparagonite (with slightly higher Na content) have well-matchedRaman spectral patterns. The position shifts of all three majorpeaks (702, 411, and 263 cm�1) are almost negligible (<2–3 cm�1),regardless of changes in K/Na ratio among them (Table S3). Thisobservation again confirms the effect that cations (i.e. K+ and Na+)that occupy a site at a longer distance from the coordinating oxygenof (Si2O5)

2� units (i.e. the A) have a minimal effect on the vibrationof (Si2O5)

2�.Lepidolite [K(Li, Al)3(Si, Al)4O10(F,OH)2] contains Li

+ and Al3+ in oc-tahedral sites.[63] Obvious peak shifts at 711 and 404 cm�1 with anadditional peak at 560 cm�1 were observed in all spectra of four le-pidolite samples from different localities (only one spectrum shownin Fig. 10). A light cation Li+ that occupies octahedral site affects thevibration of the (Si2O5)

2� group that shares a coordinating oxygen.Glauconite [(K, Na)(Fe3+,Al,Mg)3(Si,Al)4O10(OH)2] has Fe3+ and

Al3+ occupying the octahedral site.[64] It has a slightly downshifted

Figure 10. Raman spectra of three muscovite samples with standard formula oNa-rich paragonite, Li-rich lepidolite, Fe3+-rich glauconite, and Ca-rich margaritesilicates. Dotted lines are used to help the reorganization of Raman peak shifts.


701-cm�1 peak; but the major difference from muscovite is the ex-tra peaks at 604 and 553 cm�1, a doublet at 454 and 444 cm�1, anda peak at 275 cm�1 (Fig. 10b). This spectral pattern and peak posi-tions match with the data published by Ospitali et al.[65] The RRUFFRaman spectral database includes a spectrum of glauconite; butidentification of the sample is, at this time, marked as unconfirmed.

When the Ca2+ cation enters the A site, the bivalent cation in-duces an Al/Si ratio change in [(Al,Si)2O5] layer for charge balanceand thus generates very different Raman spectral features. The lastspectrum of Fig. 10 was measured from a margarite (Ca-rich mica[66]) sample, with a formula of [(Ca1.46Na0.58)Al2(Al2Si2O10)(OH)2]based on EPMA analysis (Table S3). The Si–Ob–Si mode ofmargariteappears as a doublet at 711 and 672 cm�1, very different fromthe triplet peaks (near 767, 719, and 676 cm�1) of K-muscovite,Na-muscovite from the same vibrational mode.

This change of spectral pattern is similar to the change observedin the Raman spectrum of Ca-feldspar from those of K-feldspar,Na-feldspar.[44] The Raman peaks of ring-breathing mode fromanorthite (Ca(Al2Si2O8) appear as a doublet near 504 cm�1,while the peaks of same mode from orthoclase (K(AlSi3O8) or albite(Na(AlSi3O8)) appear as a triplet near 513 and 510 cm�1, respec-tively. The change in the number of peaks in feldspars wasinterpreted by the grouping of Si–O–Si and Si–O–Al bond angleswhen the Al/Si ratio changes to 2 : 2 in the Ca-end member.[44]

This interpretation can probably be used to understand theRaman spectral pattern change in these Ca-muscovite, K-muscovite,and Na-muscovite. In the spectrum of margarite, a stronger Ramanspectral peak from the stretching vibration of Si–Onb appears at918 cm�1 (Fig. 10b), because there are two Al–O bonds with a lowerdegree of covalence than Si–O bonds in each (Al2Si2O10) unit ofmargarite. Furthermore, the spectral pattern in the range of500–100 cm�1 of margarite is remarkably different from othermuscovite samples, with the strongest peak occurring at 393 cm�1

instead of 411 cm�1 as in K-muscovite and Na-muscovite. The

f [KAl2(AlSi3O10)(OH)2], from different localities, and four muscovite varieties:. (a) H2O/OH spectral range; (b) spectral range for fundamental vibrations of



Raman spectrum of margarite published in the RRUFF databasehas consistent spectral features as described here.

In the spectral range for H2O/OH vibrationmodes, the peak shiftsamong all K-muscovite and Na-muscovite samples are almost neg-ligible (<2–3 cm�1) for the peak at 3625 cm�1. The same mode inthe spectrum of margarite shifts slightly upward to 3630 cm�1

and has a narrow peak width. In the spectrum of glauconite, thismode splits into three peaks, 3606, 3560, and 3540 cm�1, and isdownshifted owing to the mass effect of the Fe cation (Fig. 10a).

Eastonite [KMg3(AlSi3O10)(OH)2] has a Raman spectral pattern(Fig. 11a and 11b) similar to that of muscovite, with large downshiftsin peak positions of all fundamental modes (681, 351, and 192 cm�1

compared with 702, 408, and 262 cm�1 of muscovite, Table S2).Among the solid solution series of [K(Mg, Fe)3(AlSi3O10)(OH)2],Mg-rich phlogopite has a similar spectral pattern to that ofeastonite. An obvious spectral pattern change occurs for a phlog-opite with an Fe# = 0.l7 [where Fe# is the iron number, i.e.Fe# = Fe2+/(Fe2+ +Mg)] to a nearly end member lepidomelanewith a Fe# = 0.86. Major changes include peak splitting of theOH mode and Si–Ob–Si mode and an additional strong peak near550 cm�1. More importantly, following increase of the iron num-ber in these solid solution micas, some Raman peaks show a sys-tematic position shift towards lower wavenumber as shown inFig. 11c and indicated by dotted lines in Fig. 11a and 11b.

The best correlation between Fe# and peak position was foundfor peaks in the 3800–3500 cm�1 range from the OH vibrationalmode (Fig. 11c). These peak position downshifts are caused bythe mass effect when Mg is substituted by Fe2+ in octahedral sitesin the Oc layer of these micas. The central cation in the octahedralsite is surrounded by O and OH in the form of [(Mg, Fe)O4(OH)2].Obviously, a change from an Mg–OH bond to a Fe–OH bond wouldaffect the OH vibrational mode. Furthermore, because each [(Mg,Fe)O4(OH)2] shares all four of its coordinating oxygens with(Si2O5)

2� units, the change in central cation of the octahedra alsoaffects some of the Si–Ob–Si vibrational modes, which cause peakposition downshifts of the two components of the triplet between

Figure 11. Raman spectra of seven solid solutions of phlogopite-biotite serierange; (b) spectral range for fundamental vibrations of silicates. Dotted lines arpeak positions with Fe# for solid solutions of phlogopite-biotite.


800 and 700 cm�1 (Fig. 11c). Figures 10 and 11 show that using thedifference in Raman spectral patterns and peak positions, classifica-tion of Al-rich, Mg-rich, and Fe-rich micas can be made, and a cali-bration for estimating the Fe# for the phlogopite-biotite seriescan be established.

It is common that H2O molecules enter the space between theT–Oc–T layers in the structure of micas. They induce broadeningof the Raman peaks in the H2O/OH spectral range of muscovite,phlogopite, biotite, and their solid solutions (Figs. 10 and 11). Forthis reason, mica species have broader peaks at 3800–3500 cm�1

than the sharp OH peaks of pyrophyllite and talc (Figs. 8 and 9).Differences in structural sites for H2O can cause a change in peakwidths and peak positions. A Raman spectroscopic study on struc-turally well-characterized micas at low temperature would separateand sharpen these component peaks and would clarify these peakassignments.

Raman spectral features of the smectite group (T–Oc–T–B,B¼Na, K, Ca, Li, Mg(H2O)6)

Montmorillonite (Na,Ca)0.33(Al,Mg)2(Si4O10)(OH)2·nH2O, saponite(Na,Ca/2)0.33(Fe

2+,Mg)3 (Si,Al)4O10(OH)2·4H2O, nontronite Na0.33Fe3+2 (Si,

Al)4O10(OH)2·nH2O, andvermiculite (Mg,Fe3+,Al3+)3(Al,Si)4O10(OH)2·4H2Ohave a common stacking sequence T–Oc–T–B, in which B sites arefilled by large cations such as Na+, K+, Ca2+, Li+, and/or Mg(H2O)6units and sandwiched between T–Oc–T layers.[67,68]

Figure 12 shows the Raman spectra of fourmontmorillonite sam-ples (Ca-rich or Na-rich), taken using two excitation laser lines (532and 442nm). All four samples (STx-1, SAz-2, SWy-1, and SWy-2)were obtained from the CMS as clay standards, for which thechemistry[36] and powder XRD analyses[37] have been published.Three of the four samples were previously studied using laserRaman and IR spectroscopy.[28] When comparing the Raman spec-tra in the 1200–100 cm�1 range of fundamental modes (Fig. 12b),similarities in spectral patterns are evident. They all have threestrong peaks near 705, 440, and 205 cm�1; a sharp peak near

s, [K(Mg, Fe)3(AlSi3O10)(OH)2], from different localities. (a) H2O/OH spectrale used to help the reorganization of Raman peak shifts. (c) Shifts of Raman


Figure 12. Raman spectra of four standard montmorillonite samples with standard formula of [(Na,Ca)0.33(Al,Mg)2(Si4O10)(OH)2·nH2O], measured using 532-nm (first spectrum) and 442-nm excitations (last three spectra). (a) H2O/OH spectral range; (b) spectral range for fundamental vibrations of silicates. Dottedlines are to help observe the almost non-change positions of OH and Si–Ob–Si peaks in these montmorillonite samples.


290 cm�1; and many weak peaks in a range of 1100 to 600 cm�1.Among the four samples, the strongest peak shifts from 709–710 cm�1 for STx-1 and SAz-2 to 702–701 cm�1 for SWy-1 andSWy-2, and these shifts could arise from the mass effect (Fe–Mgsubstitutions in the Oc layer) on the Si–Ob–Si vibration, becauseSWy-2 has slightly higher Fe, Al, and Na than STx-1 but lower Mgand Si.[36] Because Ca and Na occupy the large site B among theT–Oc–T layers, these cation variations have very little effect on thevibration modes of (Si2O5)

2� in the spectral range <1200 cm�1.In the spectral range of H2O/OH vibrational modes (3900–

3000 cm�1, Fig. 12a), there are two types of Raman peaks: sharp,strong peak(s) and/or peak shoulders above 3600 cm�1 assignedto OH vibrations and the broad ‘envelope of peaks’ with multiplecomponents and variable intensities below 3600 cm�1 assignedto the OH vibration in H2O molecules. On the basis of montmoril-lonite crystal structure refinements,[67,68] two OH groups are partof the [(Al,Mg)O4(OH)2] octahedron, i.e. they have well-defined crys-tallographic sites. There is no data on the crystallographic site ofthe H2O molecules in the montmorillonite structure, but H2O isexpected to enter the B sites between the T–Oc–T layers similarlyto other large cations (Na+, K+, Ca2+, and Li+). The broad Ramanpeaks (at <3600 cm�1) with multiple components (top spectrumin Fig. 12a) and variable peak intensities (among four spectra inFig. 12a) all suggest irregularities in the crystallographic sites forH2O as well as in the number of H2O molecules per unit cell.Raman spectra of montmorillonite, saponite, and nontronite in

the H2O/OH spectral range (Fig. 13a) bear some similarities, butthere are obvious peak position shifts. The OH peak in the spectrumof nontronite shows a strong downshift to 3570 cm�1 comparedwith the similar peak of montmorillonite at 3625 cm�1. A similarpeak in the spectrum of saponite occurs at 3678 cm�1. In addition,


themuch narrower OHpeakwidth in the spectrumof saponite sug-gests a highly ordered crystallographic site for OH groups in itsstructure.

The Si–Ob–Si Raman peaks in the spectra of montmorillonite(709 cm�1) and saponite (676 cm�1) have peak positions thatmatch with the criteria used to distinguish dioctahedral andtrioctahedral phyllosilicates (Fig. 13b). The same peak (682 cm�1)in the spectrum of nontronite (dioctahedral) seems more affectedby the mass effect induced by Fe3+. The Raman spectrum ofnontronite published in the RRUFF database is consistent with theone described here.

Vermiculite is a group of phyllosilicates that has similar T–Oc–T–Bstacking sequence but with highly variable chemistry and structure,especially in the B site. Crystal structural refinements were carriedout for vermiculite with Ca occupying B sites,[54] with Mg(H2O)6 oc-cupying B sites,[53] and with organic species C–N–O occupying Bsites.[69] Nevertheless, we do not anticipate a change of cation orionic group in B sites to have a noticeable effect on the majorRaman peak positions of (Si2O5)

2� units in T layers of vermiculite.The Raman spectra (Fig. 14a and 14b) of three vermiculite samples,including jeffersite and zonolite, two vermiculite varieties, confirmthis hypothesis. The peaks at 1090, 900, 673, 550, 430, 350, 280,and 190 cm�1 appear in the Raman spectra of all three vermiculitesamples (Fig. 13b), and very minor peak shifts (2–3 cm�1) amongthe samples were observed. A strong peak centered around550 cm�1 among three samples was observed and will bediscussed in the Empirical Rules for Extracting Chemical Charactersof Some Phyllosilicates on the Basis of their Raman Spectra Section.A Raman spectrum of vermiculite published in the RRUFF databaseconfirmed these observations. In contrast, the Raman spectral pat-terns in the H2O/OH spectral range vary widely among the samples


Figure 13. Raman spectra of three end members of smectite group, montmorillonite, saponite with standard formula of [(Na,Ca/2)0.33(Fe2+,Mg)3(Si,Al)4O10

(OH)2·4H2O] and nontronite with standard formula of [Na0.33Fe3+2 (Si,Al)4O10(OH)2·nH2O]. (a) H2O/OH spectral range; (b) spectral range for fundamental

vibrations of silicates. Dotted lines are to help observe the position shifts. * marks the Raman peaks from hematite.

Figure 14. Raman spectra of three vermiculite samples from different localities. The general vermiculite formula is [(Mg,Fe2+,Al)3(Al,Si)4O10(OH)2·4H2O]. (a)H2O/OH spectral range; (b) spectral range for fundamental vibrations of silicates. Dotted lines are to help observe the changed peak positions.


(Fig. 14a). Apparently, these Raman spectral features reflect thehighly variable sites for OH groups and H2O molecules in the ver-miculite structure.

Electron probe microanalyses of three vermiculite samplesshow high concentrations of K in EPSC, 480-4 (jeffersite) andEPSC, N480-2 (zonolite), which would place them in the category


of hydrobiotite. Hydrobiotite has an ideal formula (Mg,Fe3+,Al3+)3(Al,Si)4O10(OH)2·4H2O, essentially the same as that of vermiculite,and has been considered as an intermediate hydration productof biotite. In the structure of hydrobiotite, the K cation in the Asite is substituted partially by H2O.

[70] The similarity of three Ra-man spectra in Fig. 14b suggests that H2O substitution, although



incomplete in samples EPSC, 480-4 and EPSC, N480-2, has a greateffect on the structure of T–Oc–T stacking layers such that the Ra-man peaks of (Si2O5)

2� vibrational modes occur at the same po-sitions as a standard vermiculite sample (EPSC, 1069). On theother hand, the large differences in Raman spectral patternsand peak positions in 3800–3500 cm�1 H2O/OH spectral range(Fig. 14a) reveal complete or partial substitution of K by H2O inthese samples (Table S3).

Raman spectral features of the chlorite group (T–Oc–T–Oc–T–Oc–T)

Cookeite LiAl2(OH)6Al2(AlSi3O10)(OH)2 is the Al-rich end member ofthis group. Chlorite (Mg,Fe,Al)3(OH)6(Mg,Fe,Al)3[(Si,Al)4O10](OH)2 isthe (Mg, Fe)-rich end member. The general structure of chloriteshas T–Oc–T layers with layers of (Mg,Fe,Al)(OH)6 octahedrasandwiched in between. Depending on the ratio of Fe2+/R2+

(R=other divalent cations) and the ratio of Si/Al3+, the chlorite min-erals can be classified into three subgroups [71] according to theFe#; (a) penninite, clinochlore, sheridanite; (b) diabentite,brusnvigite, ripidolite (prochlorite); and (c) delessite, chamosite,and thuringite.We do not have a cookeite sample that can be confirmed by EPMA

or XRD. The RRUFF database published two Raman spectra (using532-nm and 785-nm excitations), but both spectra have extremelylow S/N.We thus concentrate on the Raman spectral features of chlo-rites in this section, which occur more commonly in terrestrial rocks.We measured the Raman spectra of eight chlorite samples from

different localities. Among them, wemade EMP analyses on six. Un-fortunately, the Fe/(Fe+Mg) ratios of most of these samples arewithin the range of 0.05–0.12 and thus belong to the sub-categoryof penninite, clinochlore, and sheridanite. The only Fe-rich chlorite

Figure 15. Raman spectra of four chlorite samples from different localities. TheH2O/OH spectral range; (b) spectral range for fundamental vibrations of silicate


sample is a ripidolite (prochlorite) CCa-1, purchased from the CMS,with a composition of Ca0.05(Mg4.44 Al0.60 Fe3+3.47 Fe2+3.02Mn0.01Ti0.06)(Si4.51 Al3.49)O20(OH)16. Four Raman spectra from thosechlorites are shown in Fig. 15.

In the spectral range of silicate fundamental vibrations (Fig. 15-b, 1200–100 cm�1), three Mg-rich chlorite samples have very sim-ilar Raman spectral patterns, i.e. three major Raman peaks at 683,552, 357, and 203 cm�1 with similar relative peak intensities. Thepeak shifts among the three samples are within a 2–3 cm�1

range. The most obvious change in the spectrum of Fe-richripidolite is the peak position downshift of the Si–Ob–Si modeto 671 cm�1 from 681-683 cm�1 of Mg-rich chlorites. Two Ramanspectra of chlorites published in RRUFF database show thesame type of peak downshift to 670 cm�1 for a compositionof (Mg2.82Fe

2+1.80A11.24 Mn0.06) (Si2.91Al1.09) O10(OH)8 and to

662 cm�1 for a composition of (Fe2+2.53Mg2.15Al1.30Mn0.02)(Si2.70Al1.30)O10(OH)8. Another common spectral feature of Fe-richchlorites is the intensity reduction of a peak near 350 cm�1 ob-served in the spectrum of CCa-1 (Fig. 15b) and in the spectraof two chlorites in RRUFF database.

In the spectral range of H2O/OH vibrations (Fig. 15a), the Ra-man spectral patterns from all four chlorite samples show widevariations, which we believe are caused by the following threefactors:

(1) There are two types of Oc layers in the chlorite structurewhere OH groups occur. The first type is the Oc layer insandwiched T–Oc–T layers where both OH and O are co-ordinated with octahedral cations. The second type isthe Oc layer that occupies the space between the T–Oc–T layers in which the octahedral cation coordinates onlywith OH.

general chlorite formula is [(Mg,Fe,Al)3(OH)6(Mg,Fe,Al)3[(Si,Al)4O10](OH)2]. (a)s. Dotted lines are to help observe the peak position shifts.



(2) Cation substitutions (Mg, Fe, and Al) in these Oc layerschange the site symmetry for OH groups, which in turn couldcause the non-equivalency of OH groups (as in the case ofpyrophyllite and talc) generating additional Raman peaks.

(3) The entrance of structural H2O into chlorite with variations inthe amount of H2O per chlorite molecule and in the locationwithin the structure can cause additional Raman peaks andbroadening of Raman peak widths.

Empirical rules for extracting chemical charac-ters of some phyllosilicates on the basis oftheir Raman spectra

We showed in the Determining the Degrees of Polymerization ofSilicates from the Raman Spectral Patterns Section that thephyllosilicates can be distinguished from other silicates based ontheir Raman spectral pattern in a straightforward way. We alsoshowed in the Distinguishing the Dioctahedral and TrioctahedralPhyllosilicates Section that dioctahedral and trioctahedralphyllosilicates can be distinguished from each other based on theirRaman peak positions. Nevertheless, the detailed spectral analysesof the five major types of phyllosilicates presented in theUnderstanding the Raman Spectral Features of Five Groups ofPhyllosilicates Section demonstrates a wide range of spectralchanges owing to complex structural stacking sequences, a widerange of chemical substitutions, and corresponding structuralvariations.

In addition to developing a Raman spectral database for thephyllosilicates (Section on Challenges for Identifying and Character-izing Phyllosilicates on Earth and on Other Planets), we find thatsome empirical rules can be applied for extracting additional com-positional information for the phyllosilicates from their Ramanspectra. This is especially important when conducting a stand-aloneRaman analysis during robotic surface exploration on a planetarysurface.

The first rule is to use spectral patterns in the H2O/OH range todistinguish between simple stacking sequences (T–Oc; T–Oc–T;Figs. 5a, 6a, 8a, and 9a) from much more complicated ones(T–Oc–T–A; T–Oc–T–B; T–Oc–T–Oc–T–Oc–T; Figs. 10a, 11a, 12a,13a, 14a, and 15a), because in the latter cases, structural H2O canenter the space between T–Oc–T layers of mica, smectites, andchlorites that will induce the changes in the spectral patterns,peak positions, peak widths, and relative peak intensities in the3800–3500 cm�1 range.

The second rule is to use the appearance and the relative inten-sities of three sets of peaks between 600–300 cm�1 to separateFe-rich, Al-rich, and Mg-rich phyllosilicates. The key indicator peakfor Fe-rich phyllosilicates occurs near 550 cm�1 in the spectra ofFe-talc (minnesotaite, Fig. 9b), glauconite (Fig. 10b), and nontronite(Fig. 13b). This peak also shows increased peak intensities in thespectra of solid solutions of phlogopite-biotite following increaseof the Fe# (Fig. 11b) and appears with variable peak intensities inthe spectra of vermiculites (Fig. 14b). The exception is the sharppeak near 550 cm�1 in the spectra of chlorites, which is contributedby the (Si2O5)

2� vibrational mode. Iron enrichment (Fe-richripidolite CCa-1, Fig. 15b) causes an additional peak on the lowwavenumber side of this 550-cm�1 peak, accompanied by reduc-tion of intensity of the peak near 350 cm�1. The distinguishing peakfor Mg-rich phyllosilicates occurs near 350 cm�1. It appears in thespectra of talc (Fig. 9b) and saponite (Fig. 13b), and it shows


reduced peak intensities in the spectra of solid solutions ofphlogopite-biotite with increase of Fe# (Fig. 11b). This peak also ap-pears with variable peak intensities in the spectra of vermiculites(Fig. 14b). This peak is very strong in the spectra of Mg-rich chloritesbut shows reduced peak intensity in the spectra of Fe-rich chlorites(Fig. 14b). The distinguishing peak for Al-rich phyllosilicates occursnear 430 cm�1, and it appears in the spectra of muscovite (Fig. 10b)and montmorillonite (Figs. 12b and 13b). We emphasize that usingthese three sets of peaks to predict a phyllosilicate to be Fe-rich,Mg-rich, or Al-rich is based on observations from the current study.The rule is empirical based on the Raman spectra we have collectedto date.

The third rule, to be investigated in detail in future studies, is thatthe Fe/(Fe+Mg) (where Fe is 2+) ratios of some phyllosilicatesmight be extracted from the Raman peak position shifts in a waysimilar to the calibrations we developed for olivine[45] andpyroxene.[46] Figure 11c gives the best demonstration of this corre-lation for phlogopite-biotite solid solutions. Similar correlationsmayexist for chlorites and other phyllosilicates whose octahedral cat-ions aremainly Mg and Fe2+. With our current data, only a rough es-timation on high, median, and low Fe# can be determined from theRaman peak positions.

Challenges for identifying and characterizingphyllosilicates on Earth and on other planets

As discussed in the Overview: Raman Spectral Features for Classifi-cation of Phyllosilicates Section, silicates are moderately strongRaman scatterers compared with nitrates, carbonates, sulfates,and phosphates. Among the silicates with different degrees of po-lymerization (orthosilicate, ring silicate, chain silicate, double chainsilicate, phyllosilicate, and tectosilicate), there should be no specificdifficulty in distinguishing the Raman spectra of phyllosilicates fromthe silicates with other degrees of polymerization. Nevertheless, thecomplicated layered structure of phyllosilicates, the large spacesbetween the layers, and the variability of cations and ionic groups(especially H2O and OH) that can enter the spaces between layers,can cause complexities in the Raman spectra (as described in theUnderstanding the Raman Spectral Features of Five Groups ofPhyllosilicates Section).

Because of these complexities in a Raman spectral database ofphyllosilicates, it is important to include a large number of Ramanspectra from samples of each phyllosilicate mineral originatingfrom different localities and affected by different geologic pro-cesses. The information on localities, chemistry, and structure willhelp in understanding the major spectral characteristics of aphyllosilicate and especially potential spectral variations arisingfrom structural distortions and chemical substitutions.

In addition to the complications discussed earlier, clay mineralsbring three more difficulties in their identification and characteriza-tion by laser Raman spectroscopy. The first difficulty arises from theultrafine grain size (normally considered to be less than 2μm onstandard particle size classification). Unlike Vis-NIR and thermalemission IR (TES), the Raman spectral pattern and Raman peakpositions of a mineral are unchanged for the ultra-fine grain size(including sub-micron grain sizes).[72] However, from an aggregateof fine particles, the Raman signal strength (i.e. the number ofRaman photons entering the collecting optics of a Raman spec-trometer) is greatly reduced by the intense scattering of the laserphotons and of the generated Raman photons at grain surfaces



and grain boundaries. For this reason, the clay minerals are intrinsi-cally very weak Raman scatterers.The second difficulty is the weak Raman signal of clays arising

from low degrees of crystallinity. Many clays were formed by in situchemical weathering of igneous materials at low temperatureresulting gels or mixed layer types that have poor crystallinity.[73,74]

The lack of long-range translational symmetry in the structure ofamorphousmaterials (gels) would cause a Raman signal reductionof 2–3 orders of magnitude.[75] For example, a study on saponiteshows a very low S/N Raman spectrum from the sample synthe-sized at 120 °C and much improved S/N for a sample synthesizedat higher temperature (280 °C).[27] A study on montmorillonitesamples using both Raman and XRD demonstrated the reductionof Raman signal strength for the samples where the long-rangeorder is absent or the order is randomized with layer spacing.[26]

These studies revealed that high crystallinity generates a Ramanspectrum with high S/N. The natural clays, especially thoseformed in a low temperature environment (as on Mars) or lack-ing post-formation metamorphism, are normally low in crystal-linity and can appear as an amorphous phase in mineralidentifications.[26,27,73,74,76,77] Furthermore, when co-existing withcrystalline phases (igneous minerals such as olivine and pyrox-ene), the broad and weak Raman peaks of gel-like clays wouldbe concealed by strong Raman peaks of the crystalline minerals.A third difficulty that can interfere with terrestrial clay detection

by laser Raman spectroscopy involves the formation and preserva-tion of clays in terrestrial field sites. Contamination from organicspecies in terrestrial clays is extremely common and this can pro-duce a high fluorescence background that obscures detection ofthe already weak Raman signals from clays.Nevertheless, this fluorescence interference does not occur in

almost all extraterrestrial materials that have been examined,hitherto, by Raman spectroscopy and by UV-stimulated fluores-cence microscopy. These observations include the following: (1) Adetailed data analysis [78] of optical microscope results for thePhoenix Mars Lander reported no detection of UV (360–390nm)-stimulated luminescence from the soil grains at the Phoenix land-ing site; (2) a similar analysis of a UV (365nm)-stimulated lumines-cence image of Sayunei at Yellowknife Bay taken by MAHLI onthe Curiosity rover reported no detected fluorescence;[79] (3) twoUV (365nm)-fluorescence studies [17,80] on five Martian meteorites:Lafayette, Los Angeles, Nakhla, EETA79001, and Zagami show nodetected fluorescence, except those from epoxy and from amerrillite (a phosphate, known to be enriched in Rare Earth Ele-ments in some lunar samples that emits sharp fluorescence peaks)grain in Zagami; (4) good quality Raman spectra (using cw 532-nmexcitation), without obvious fluorescence interferences, were ob-tained in the study of four Martian meteorites: Zagami,[19] LosAngles,[81] EETA79001,[48] and MIL03346;[82,83] (5) a quantitativestudy [84] on the fluorescence emitting properties of a variety of ex-traterrestrial materials (lunar and Martian meteorites, achondrites,carbonaceous chondrites, and H, L, and LL type ordinary chon-drites) found an average of 10–100 times weaker fluorescenceemitting strength than four terrestrial clay samples from the CMS(SWx-1, WWy-1, KGa-1, and KGa-2).Based on aforementioned observations at the surface of Mars

and on extraterrestrial materials, we do not anticipate fluores-cence being a threat for in situ Raman measurements on Marsand other planetary bodies. During a planetary surface explora-tion, the major causes for potential weak Raman signal from clayminerals would be their fine-grain size and low degree ofcrystallinity.


Summary

This study demonstrates that phyllosilicates have a distinctiveRaman spectral pattern that can be used to distinguish them fromother silicates having different degrees of polymerization. More-over, the five major subgroups of phyllosilicates all have character-istic Raman spectral features. The empirical rules extracted from thisstudy should help to identify and characterize these minerals fromstand-alone Raman spectroscopic measurements during roboticsurface exploration of the planetary bodies in our Solar System.

Acknowledgements

This complex and lengthy investigation has been carried out duringthe development of a flight laser Raman system (MMRS-CIRS)and through the studies of its applications to planetary explorationof the Moon, Mars, Venus, and asteroids. We appreciate the finan-cial support from various NASA programs, including PIDDP(NAG5-10703), MIDP (09-030), ASTEP (NNX09AE80A), ASTID(NNG05GM95G), MFRP (NNX07AQ34G, NNX10AM89G), MER(39361-6444) and MoO (1295053). This study has been a team ef-fort. The past and current teammembers of Planetary SpectroscopyGroup include L. A. Haskin, K. Kuebler, P. Sobron, Z. C. Ling, W. G.Kong, Y. H. Zhou, Y. L. Lu, K. Connor, and J. Wei; all have made con-tributions to various aspects of this study, for which the authorsexpress their deep gratitude.

USA-NASA; PIDDP program (NAG5-10703); MIDP program (09-030); ASTEP program (NNX09AE80A); ASTID program (NNG05GM95G);MFRP program (NNX07AQ34G, NNX10AM89G); Mars ExplorationRover mission (39361-6444); MoO for ExoMars mission (1295053).

References[1] F. Poulet, J.-P. Bibring, J. F. Mustard, A. Gendrin, N. Mangold,

Y. Langevin, R. E. Arvidson, B. Gondet, C. Gomez, the Omega Team,Nature 2005, 438, 623.

[2] B. L. Ehlmann, J. F. Mustard, C. I. Fassett, S. C. Schon, J. W. Head III,D. J. Des Marais, J. A. Grant, S. L. Murchie, Nat. Geosci. 2008, 11, 355.

[3] J. L. Bishop, E. Z. Noe Dobrea, N. K. McKeown, M. Parente, B. L. Ehlmann,J. R. Michalski, R. E. Milliken, F. Poulet, G. A. Swayze, J. F. Mustard,S. L. Murchie, Science 2008, 321, 830.

[4] B. L. Ehlmann, J. F. Mustard, S. L. Murchie, J.-P. Bibring, A. Meunier,A. A. Fraeman, Y. Langevin, Nature 2011, 479, 53.

[5] J. Carter, F. Poulet, Icarus 2012, 219, 250.[6] D. T. Vaniman, D. L. Bish, D. W.Ming, T. F. Bristow, R. V.Morris, D. F. Blake,

S. J. Chipera, S. M. Morrison, A. H. Treiman, E. B. Rampe, M. Rice,C. N. Achilles, J. Grotzinger, S. M. McLennan, J. Williams, J. Bell III,H. Newsom, R. T. Downs, S. Maurice, P. Sarrazin, A. S. Yen,J. M. Morookian, J. D. Farmer, K. Stack, R. E. Milliken, B. Ehlmann,D. Y. Sumner, G. Berger, J. A. Crisp, J. A. Hurowitz, R. Anderson,D. DesMarais, E. M. Stolper, K. S. Edgett, S. Gupta, N. Spanovich,Science 2013. DOI:10.1126/science.1243480.

[7] D. L. Bish, D. F. Blake, D. T. Vaniman, S. J. Chipera, R. V. Morris,D. W. Ming, H. Treiman, P. Sarrazin, S. M. Morrison, R. T. Downs,C. N. Achilles, A. S. Yen, T. F. Bristow, J. A. Crisp, J. M. Morookian,J. D. Farmer, E. B. Rampe, E. M. Stolper, N. Spanovich, Science 2013.DOI:10.1126/science.1238932.

[8] A. Wang, B. L. Jolliff, L. A. Haskin, J. Geophys. Res. 1995, 100, 21189.[9] F. Rull, J. Martinez-Frias, Spectrosc. Eur. 2006, 18, n° 1.

[10] T. Belenguer, M. Fernandez-Rodriguez, M. Colombo, E. Diaz-Catalá,J. Sanchez-Páramo, ICSO 2010 Rhodes, Greece, International Conferenceon Space Optics, 2010.

[11] S.M. Clegg, S.K. Sharma, R.C. Wiens, S. Maurice, O. Gasnault, A.K. Misra,R. Newell, S. Bender, O. Forni, J. Lasue, M.D. Dyar, K.L. Nowak-Lovato,11th International GeoRaman Conference, 2014, abstract# 5082.

[12] L.W. Beegle, R. Bhartia, L. DeFlores, M. Darrach, R.D. Kidd W. Abbey,S. Asher, A. Burton, S. Clegg, P.G. Conrad, K. Edgett, B. Ehlmann,



F. Langenhorst, M. Fries, W. Hug, K. Nealson, J. Popp, P. Sorbon,A. Steele, R. Wiens, K. Williford, 45th Lunar and Planetary ScienceConference, 2014, abs# 2835.

[13] A. Wang, L. A. Haskin, E. Cortez, Appl. Spectrosc. 1998, 52, 477.[14] A. Wang, L. A. Haskin, A. L. Lane, T. J.Wdowiak, S. W. Squyres, R. J. Wilson,

L. E Hovland., K. S. Manatt, N. Raouf, C. D. Smith, J. Geophys. Res. 2003,108, 5005.

[15] A. Wang, J. L. Lambert, International Workshop on Instrumentation forPlanetary Missions, 2013, abstract #1157.

[16] A. Wang, J. L. Lambert, Low Cost Planetary Mission Conference, 2013,Abstract #220.

[17] A. Wang, L. A. Haskin, J. J. Gillis, 34th Lunar and Planetary ScienceConference, 2003, abstract #1753.

[18] A. Wang, B. L. Jolliff, J. L. Lambert, R. Menzies, I. Hutchinson, J. Wei,W. Goezt, 2nd International Workshop on Instrumentation for PlanetaryMissions, 2014, Abstract #1090

[19] A. Wang, B. L. Jolliff, L. A. Haskin, J. Geophys. Res. 1999, 104, 8509–8519.[20] A. Wang Venus geochemistry: Progress, Prospects, and NewMission 2009,

51.[21] A. Wang, W. G. Kong, First Moscow Solar System Symposium (1M-S3),

2010,[22] R. L. Frost, Clays Clay Miner. 1995, 43, 191.[23] R. L. Frost, J. T. Kloprogge, Appl. Spectrosc. 2000, 54, 402.[24] N. Wada, W. A. Kamitakahara, Phys. Rev. B 1990, 43.[25] C. Rinaudo, D. Gastaldi, E. Belluso, Can. Minerals. 2003, 41, 883.[26] R. L. Frost, L. Rintoul, Appl. Clay Sci. 1996, 11, 171.[27] J. T. Kloprogge, R. L. Frost, Vib. Spectrosc 2000, 23, 119.[28] J. L. Bishop, E. Murad, J. Raman spectroscopy 2004, 35, 480.[29] J. J. Blaha, G. J. Rosasco, Anal. Chem. 1978, 50, 892.[30] R. L. Frost, S. J. Van Der Gaast, Clay Miner. 1997, 32, 471.[31] V. C. Farmer, Clay Miner. 1998, 22, 601.[32] R. Petry, R. Mastalerz, S. Zahn, T. G. Mayerhofer, G. Volksch,

L. Viereck-Gotte, B. Kreher-Hartmann, L. Holz, M. Lankers, J. Popp,ChemPhysChem 2006, 7, 414.

[33] R. T. Downs, Program and Abstracts of the 19th General Meeting of theInternational Mineralogical Association in Kobe, Japan. 2006, O03-13.

[34] A. Wang, J. J. Freeman, K. E. Kuebler, 33 th Lunar and Planetary ScienceConference, 2002, Abstract #1374.

[35] A. Wang Alian, R. B. Valentine, 33rd Lunar and Planetary ScienceConference, 2002, Abstract #1370.

[36] A. R. Mermut, A. F.Cano, Clays Clay Miner. 2001, 49, 381.[37] S. J. Chipera, D. L. Bish, Clays Clay Miner. 2001, 49, 398.[38] A. D. McNaught, A. Wilkinson, IUPAC, Compendium of Chemical

Terminology, 2nd ed., Blackwell Scientific Publications, Oxford, 1997.[39] A. Wang, J. Han, L. Guo, J. Yu, P. Zeng, Appl. Spectrosc. 1994, 48, 959.[40] P. McMillan, Am. Mineral. 1984, 69, 622.[41] S. K. Sharma, J. F. Mammone, M. F. Nicol, Nature 1981, 292, 140.[42] S. K. Sharma, B. Simons, H. S. Yoder Jr., Amer. Mineral. 1983, 68, 1113.[43] D. W. Matson, S. K. Sharma, J. A. Philpotts, Amer. Mineral. 1986, 71, 694.[44] J. J. Freeman, A. Wang, K. E. Kuebler, L. A. Haskin, Can. Mineral. 2008, 46,

1477.[45] K. E. Kuebler, B. L. Jolliff, A. Wang, L. A. Haskin, Geochim. Cosmochim.

Acta 2006, 70, 6201.[46] A. Wang, B. L. Jolliff, L. A. Haskin, K. E. Kuebler, K. M. Viskupic, Am.

Minerals. 2001, 86, 790.[47] D. Bersani, S. Andò, P. Vignola, G. Moltifiori, I.-G. Marino, P. P. Lottici,

V. Diella, Spectrochim. Acta A Mol. Biomol. Spectrosc. 2009, 73, 484.[48] A. Wang, K. E. Kuebler, B. L. Jolliff, L. A. Haskin, American Minerals. 2004,

89, 665.[49] N. Rividi, Astrobiology 2010, 10, 293.[50] A. Wang, Y. Zhou, ICARUS 2013, 234, 162.[51] M. H. Brooker, D. E. Irish, Can. J. Chem. 1970, 48, 1183.[52] R. D. Shannon, Acta Cryst 1976, 32, 751.[53] H. Shirozu, S. W. Bailey, Am. Mineral. 1966, 51, 1124.[54] C. de la Calle, H.Pezerat, M. Gasperin, J. Geophys. Res. 1977, 38, 128.[55] D. L. Bish, R. B. Von Dreele, Clays Clay Miner. 1989, 37, 289.


[56] F. J. Wick, D. S. O’Henley, in Hydrous Phyllosilicates, Vol. 19 in Reviews inMineralogy Mineralogical Society of America, (Ed: P. Bailey), 1991,BookCrafters, Inc., Chelsea, Michigan, 91–159.

[57] N. J. Page, R. G. Coleman, Geological Survey Research 1967, paper575-B, pp. B103-107.

[58] G. Capotani, M. Mellini, Am. Mineral. 2004, 89, 147.[59] J. H. Lee, S. Guggenheim, Am. Mineral. 1981, 66, 350.[60] B. Perdikatsis, H. Burzlaff, Zeitschrift fur Kristallographie 1981, 156, 177.[61] S. M. Richardson, J. W. Richardson, Am. Mineral. 1982, 67, 69.[62] G. J. Redhammer, A. Beran, J. Schneider, G. Amthauer, W. Lottermoser,

Am. Mineral. 2000, 85, 449.[63] S. Guggenheim, Am. Mineral. 1981, 66, 1221.[64] V. A. Drits, B. B. Zviagina, D. K. McCarty, A. L. Salyn, Am. Mineral. 2010,

95, 348.[65] F. Ospitali, D. Bersani, G. D. Lonardo, P. P. Lottici, J. Raman Spectroscopy

2008, 39, 1066.[66] W. Joswig, Y. Takeuchi, H. Fuess, Zeitschrift fur Kristallographie 1983,

165, 295.[67] D. Gournis, A. Lappas, M. A. Karakassides, D. Tobbens, A. Moukarika,

Phys. Chem. Miner. 2008, 35, 49.[68] A. Viani, A. Gualtieri, G. Artioli, Am. Mineral. 2002, 87, 966.[69] F. Kanamaru, V. Vand, Am. Mineral. 1970, 55, 1550.[70] G. W. Brindley, P. E. Zalba, C. M. Bethke, Am. Mineral. 1983, 68, 420.[71] M. D. Foster, U.S. Geological Survey Professional paper, 1962, 414-A, 33

pp.[72] K. E. Kuebler, A. Wang, K. Abbott, L. A. Haskin, 32nd Lunar and Planetary

Science Conference, 2001, abstract #.[73] N. J. Tosca, F. A. Macdonald, J. V. Strauss, D. T. Johnston, A. H. Knoll,

Earth Planet. Sci. Lett. 2011, 306, 11.[74] N. J. Tosca, A. L. Masterson, Clay Miner. 2013, 49, 195.[75] W. B. White, in The Infrared Spectra of Minerals (Eds: VC Farmer) ,

Mineralogical Society, London, 1975 p, 87–110.[76] A. Wang, R. L. Korotev, B. L. Jolliff, L. A. Haskin, L. Crumpler,

W. H. Farrand, K. E. Herkenhoff, P. de Souza Jr., A. G. Kusack,J. A. Hurowitz, N. J. Tosca E02S16, J. Geophys. Res. 2006, 111.

[77] D. W. Ming, D. W. Mittlefehldt, R. V. Morris, D. C. Golden, R. Gellert,A. Yen, B. C. Clark, S. W. Squyres, W. H. Farrand, S. W. Ruff,R. E. Arvidson, G. Klingelhöfer, H. Y. McSween, D. S. Rodionov,C. S. P. A. de Souza Jr., A. Wang E02S12, J. Geophys. Res. 2006, 111.

[78] W. Goetz, M. H. Hecht, S. F. Hviid, M. B. Madsen, W. T. Pike, U. Staufer,M. A. Velbel, N. H. Harrit, E. Zych, K. S. Edgett, Planet. Space Sci. 2012,70, 134.

[79] M. Minitti, K.S. Edgett, R.A. Yingst, P.G. Conrad, M. Fisk, C.J. Hardgrove, K.E. Herkenhoff, L.C. Kah, M.R.Kennedy, G.M. Krezoski, M.T. Lemmon,L. Lipkaman, S.R. Kuhn, M.L. Robinson, V.V. Tompkins, A. Treiman, K.H. Williford, 44th Lunar and Planetary Science Conference, 2014,abstract #2029.

[80] M. Minitti, T. McCoy, 43rd Lunar and Planetary Science Conference, 2012,abstract #2349.

[81] A. Wang, K. E. Kuebler, J. J. Freeman, B. L. Jolliff, 32nd Lunar andPlanetary Science Conference, 2001, abstract #xxx.

[82] Z. C. Ling, A. Wang, Secondaryminerals in Martianmeteorite MIL 03346as detected by Raman imaging spectroscopy. 11th GeoRamanInternational Conference, 2014, abstract #5089.

[83] A. Wang, R. L. Korotev, B. L. Jolliff, Z. C. Ling, Planetary and SpaceScience 2014. http://dx.doi.org/10.1016/j.pss.2014.10.005

[84] J. Wei, A. Wang, W. Goetz, K. Connor, 2nd International Workshop onInstrumentation for Planetary Missions, 2014, Abstract #1112.

Supporting information

Additional supporting information may be found in the onlineversion of this article at the publisher’s web site.


http://dx.doi.org/10.1016/j.pss.2014.10.005

understanding the raman spectral features of phyllosilicates€¦ · their source origins, and the...

Documents