Mon. Not. R. Astron. Soc. 367, 577–591 (2006) doi:10.1111/j.1365-2966.2006.09894.

Quantitative infrared spectra of hydrosilicates and related minerals

A. M. Hofmeister1 and J. E. Bowey2

1 Department of Earth and Planetary Science, Washington University, 1 Brookings Drive, St Louis, MO 63130, USA
2 Department of Physics and Astronomy, University College London, Gower Street, London WC1E 6BT

Accepted 2005 November 18. Received 2005 November 9; in original form 2005 January 20

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

ABSTRACT
Absorption coefficients associated with atomic motions of species expected in astronomical
environments are determined from infrared measurements of various hydrosilicates, hydrated
magnesium oxide, and the Al-bearing chain silicate, sapphirine. Band types measured include
O–H stretching modes near 3 μm, Si–O stretching motions near 10 μm, Si–O–Si bends near
14 μm, O–Si–O bends near 20 μm, and translations of cations such as Mg and Ca near
50–200 μm. We obtain data from films of varying thickness and use a ratioing method. First,
bandstrengths of O–H fundamentals were determined from spectra obtained from films of
controlled thicknesses, generally 6 μm. The O–H absorbance strength was then used to ac-
curately determine thickness for a thinner film of each mineral (found to be <1 μm), thus
providing bandstrengths of all other absorptions. Thin films were prepared such that the fun-
damental lattice modes showed intrinsic behaviour (i.e. band shapes were unchanged upon
further thinning) and O–H modes are well resolved above the spectral noise. Bandstrengths
were found to depend weakly on structure and should be applicable to other silicate minerals,
allowing estimation of elemental concentrations independent of knowing the speciation of dust
in astronomical environments. Comparison with observational data of NGC 6302 suggests that
lizardite and saponite could be present in addition to refractory minerals.
Key words: line: identification – methods: laboratory – techniques: spectroscopic – infrared:
general.

crystals contain additional reflections from grain boundaries, which

1 INTRODUCTION
affect extraction of the complex indices of refraction and thus the
Mid-infrared and far-infrared spectra obtained with the Infrared inferred bandstrengths (Hofmeister, Keppel & Speck 2003). Band-
Space Observatory Short- and Long-Wavelength Spectrometers strengths are generally based on sample weight and spectra produced
(SWS and LWS, respectively) have revealed emission bands at- by finely ground powder dispersed in a transparent medium such as
tributable to crystalline silicate dust. The environments include KBr (e.g. Koike & Shibai 1990). However, dispersions can con-
comets (e.g. Wooden et al. 1999), young stellar objects (YSOs; tain agglomerates of particles that are large enough to be virtually
e.g. Malfait, Waelkens & Waters 1998) and oxygen-rich dust in out- opaque at frequencies where the solid strongly absorbs (i.e. sizes of
flows and discs associated with late-type stars and planetary neb- a few μm block light near 10 μm). Thus, some unknown proportion
ulae (e.g. Waters et al. 1996; Cohen et al. 1999; Sylvester et al. of the mass does not contribute to the IR spectrum (e.g. Hofmeister
1999; Molster et al. 2001). Hence, dust plays an important role in et al. 2000), causing bandstrengths to be underestimated.
astronomical processes, and characterizing the crystalline silicates To address these problems, we are compiling a data base of IR
comprising the dust is essential. spectra of minerals and inorganic solids that might exist in space.
Identification is made through comparison with infrared (IR) lab- The series is organized around mineral structure and includes a vari-
oratory data, yet appropriate spectra in electronic form are available ety of chemical compositions. Hofmeister et al. (2003) presented IR
for a relatively small number of species, mainly olivines (e.g. Fabian data for simple oxides and sulphides. Data on calcium aluminates
et al. 2001) and pyroxenes (e.g. Koike et al. 2000; Chihara et al. were compared to the protoplanetary nebula NGC 6302 (Hofmeister,
2002). To infer abundances of cations from the astronomical data, Wopenka & Locock 2004). The present effort centres on hydrosili-
bandstrengths are obtained by several approaches, yet each has an as- cates and provides a new means of inferring bandstrengths.
sociated drawback. Quantitative analysis of IR reflectivity data from
single crystals is accurate, but sufficiently large crystals of appropri-
1.1 Relevance of hydrosilicates to astronomy
ate minerals are not always available. Reflectivity data from poly-
Laboratory data longwards of 13 μm have been presented in the
 E-mail: hofmeist@levee.wustl.edu astronomical literature for three hydrous silicates (Koike & Shibai


C 2006 The Authors. Journal compilation 
C 2006 RAS
578 A. M. Hofmeister and J. E. Bowey
Table 1. Sample descriptions. The synthetic samples (those from Alfa or Merc) each have a purity near 99 per cent and a formula represented by the ideal. For
the natural samples, formulae are computed from electron microprobe analyses obtained from similar samples at the same locality, as referenced. The Mg and
Fe content of the mineral examined here may vary by a few per cent from the published analyses. BMNH denotes the British Museum of Natural History. See
the Acknowledgments for donators of samples.

Mineral and Class Chemical formula Locality/Reference for composition/Notes Source

Brucite Mg(OH) 2 Synthetic N. Johnson

Chain silicates
Tremolite (Ca 1.72 Mg 0.14 Mn 0.04 Fe 0.01 )(Mg 4.92 Al 0.08 )Si 8 O 22 (OH) 2 Balmat, NY, USAa Washington Univ.
Tremolite K 0.01 Na 0.04 Ca 1.93 (Mg 4.90 Fe 0.02 )(Al 0.03 Si 7.97 )O 22 (OH) 2 Canaan, CT, USAb Wards
Tremolite K 0.01 Na 0.09 Ca 1.94 (Mg 4.96 Fe 0.02 )(Al 0.07 Si 7.98 )O 22 F 0.54 (OH) 1.46 Campolungo, Switzerlandc,d N. Johnson
Actinolite (Ca 1.98 Na 0.02 )(Mg 4.14 Fe 0.86 )(Al 0.09 Si 7.91 )O 22 (OH) 2 Cumberland, RI, USAe Harvard 9959

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

Sapphirine (Mg 3.53 Fe 0.23 Al 4.22 Cr 0.01 Ti 0.01 )(Al 4.4 Si 1.6 )O 20 Fiskennasset, W. Greenlandc,f R.F. Dymek

Sheet silicates:
Chrysotile Mg 2.95 Fe 0.05 Si 2 O 5 (OH) 4 Thetford, Quebec, Canadag Wards
Lizardite Mg 2.95 Fe 0.05 Si 2 O 5 (OH) 4 Thetford, Quebec, Canadag Wards
Talc, synthetic Mg 3 Si 4 O 10 (OH) 2 Synthetic Alfa/Aesar
Talc, natural (Mg 2.91 Fe 0.08 Ni 0.01 )Si 4 O 10 (OH) 2 Unknown Washington Univ.
Saponite (Na,Ca/2) 0.3 (Mg,Fe) 3 (Al,Si) 4 O 10 (OH) 2 · 4H 2 Oh Old Kilpatrick, Dunbartonshire, Scotland BMNH 54285
Saponite, mixture (Na,Ca/2) 0.3 (Mg,Fe) 3 (Al , Si) 4 O 10 (OH) 2 · 4H 2 Oh Texas, Lancaster Co., PA, USA BMNH 40259
Montmorillonite Na 0.05 Ca 0.02 (Al 1.64 Mg 0.17 Fe 0.18 )(Al 0.10 Si 3.90 )O 10 (OH) 2 · nH 2 O Cook County, WY, USAc,i Washington Univ.
Montmorillonite (Na,Ca) 0.3 (Al,Mg) 2 Si 4 O 10 (OH) 2 · nH 2 Oh Hampden Park borehole, Eastborne, BMNH 1983,583
Sussex, UK
a Deer (1963). See Williams (1984) for chemical compositions of similar samples. b Johnson & Fegley (2003a). This sample has up to 0.05 atoms of F per

formula unit. c Type locality. d Johnson & Fegley (2003b). e See the appendix. f R. F. Dymek (private communication). For analyses of other samples from this
locality, see Mustard (1992). g Cogulu & Laurent (1984). h These samples were not analysed (see text) and formulae reported are typical of these minerals.
i Weaver & Pollard (1973). Wyoming montmorillonite has small amounts of quartz and carbonate.

1990). Ground-based searches for hydrous silicates have been con- smectite (i.e. Mg 3 Si 2 O 7 and Mg 3 Si 4 O 11 ) form in gas-phase com-
ducted in the 3.5–5 μm band (Knacke et al. 1985). Although bustion experiments (Nuth, Rietmeijer & Hill 2002). If such precur-
these searches failed to measure strong features, Bowey & Adam- sors are available in astronomical environments, hydration could
son (2002) suggested that a narrow feature near the peak of the produce the phases examined here.
10-μm silicate feature in some evolved stars and YSOs could be Double chain silicates are a focus in this paper, because of possible
due to a mineral similar to montmorillonite or talc. Malfait et al. identification of tremolite in the ice-rich environments of deeply-
(1998) associated a broad band at 100 μm with montmorillonite. In embedded YSOs (Bowey & Hofmeister 2005). Our selection of sam-
these environments, the proportion of hydrous silicate could be as ples (Table 1) addresses the effect of various chemical substitutions.
much as 10–20 per cent by mass.
Many of the minerals studied here (Table 1) are known in me-
1.2 Improvements in methodology
teorites, and are major components of carbonaceous chondrites
(Zolensky & McSween 1988; Rubin 1997; Lodders & Fegley 1998). In this paper we devise a method to ascertain the thickness of our
Many are formed by aqueous alteration of mineral species identi- thin films and we use this to provide the strengths of bands involving
fied in space (forsterite and pyroxenes; Molster et al. 2001; Kemper atoms known to exist in space. Directly measuring strengths of Si–O
et al. 2002a,b), and have end-member chemistries appropriate to stretching modes is particularly difficult because films that provide
astronomical environments (e.g. Palme & Fegley 1990). Brucite is good spectral data of these peaks are thin (t  1 μm) and difficult
formed by hydration of MgO, which is expected in some environ- to measure directly (e.g. Hofmeister et al. 2003). Therefore, an in-
ments (Gail & Sedlmayr 1999). Although large-scale hydration of direct means is sought. Obtaining accurate bandstrengths through
silicates is thought to occur within asteroidal bodies, distorted and comparison requires non-overlapping bands with diverse strengths,
localized (30 nm) regions of protophyllosilicate consistent with a such that the weaker band is distinct from the noise in the spec-
pre-accretionary origin have been found on the surfaces of submi- tra obtained from the thinner films. These criteria are infrequently
crometre grains within a primitive (CO) chondritic meteorite. Lat- met. For example, the lowest frequency bands in Mg-rich olivines
tice spacings of the protophyllosilicates on CO-matrix olivines are are weak and distant from all other fundamentals (e.g. Bowey et al.
consistent with those of serpentine (Keller & Buseck 1990) whilst 2001), but the window materials needed to produce thick films in the
spacings on amorphous CO-matrix silicates indicate alteration to far-IR are either opaque in the visible (Si), which prohibits visual
serpentine or talc with interleaved brucite (Brearley 1993). Perhaps inspection of the thick film, or are too soft (plastic), which makes
the most compelling evidence for hydration occurring in circumstel- formation of a uniform film difficult, or are prohibitively expensive
lar (evolved or protostellar) or interstellar environments is the broad in large sizes (diamond). Overtone-combination bands (Bowey &
range of D/H ratios seen in pyroxene chondrules from LL3 mete- Hofmeister 2005) are generally too weak (with intensities ∼1/100th
orites (Deloule, Robert & Douknan 1998); the authors suggested those of the fundamentals) to be used in this manner.
that the isotopic anomaly originated within the hydroxyl ions of Phases with stoichiometric hydroxyl are well suited to determine
precursor phyllosilicates which were hydrated at low temperatures bandstrengths using a comparative method. (i) The O–H stretch-
in interstellar environments. Also, dehydoxylates of serpentine and ing bands have about 1/10th the intensity of the other fundamental


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
 IR spectra of hydrosilicates 579
modes and are distant from the other IR modes. These modes are It is related to emissivity (ξ ) through the Kirchhoff law:
reasonably well resolved both in spectra from films that are suffi-
ξ (ν, d) = 1 − exp[−t A(ν)]. (2)
ciently thin to provide undistorted band shapes for the Si–O modes
and in spectra obtained from films with thickness (t) ∼6 μm, which Absorbance measurements (a = At) need correcting for reflections.
is thick enough to measure accurately. Comparison then allows de- If the surface is parallel to the beam
termination of bandstrengths of the Si–O modes. (ii) Hydrous phases
At = − ln(Imeas /I0 ) + 2 ln(1 − R). (3)
are soft and compressible, thereby allowing preparation of films of
uniform thickness in the t = 6 μm range. (iii) Hard and transparent
window materials are available for high-frequency O–H modes. In this paper we report on the true absorbance, computed from
Extinction coefficients provided for the various types of modes natural logarithms. Reflections at the sample–diamond interfaces
(e.g. Si–O stretching near λ = 10 μm) should be generally appli- were neglected in converting the raw data to true absorbance shown
cable. Transferability is indicated by derivation of reasonably accu- in the figures, as the function is unknown. The value of R due to
rate OH− concentrations in various minerals from the calibration of diamond–diamond reflections is high, but is removed by using the

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

Paterson (1982), which is based on glasses and silica, and by the empty DAC as the reference spectrum and should not be included
fact that 0.3 mg of mineral dispersed in 40 mg of KBr gives simi- in computing a or A. However, diamond–diamond reflections are
lar absorbance values, irrespective of phase (e.g. Nyquist & Kagel coherent, as a result of the polished, parallel surfaces, and are man-
1971). ifest as interference fringes in the transparent region. Fringes were
eliminated from the final results as discussed below. Mass extinction
2 E X P E R I M E N TA L M E T H O D S coefficients (κ) are obtained using the density (ρ):

2.1 Sample characterization

κ = A/ρ. (4)
Chemical formulae were computed from published analyses of sam-
ples from the same locality, or from electron microprobe analysis.
We used a JEOL-733 with Advance Microbeam automation, 15- The spacing of the fringes in the thin films (ν) suggests a gap (d)
kV accelerating voltage, 30-nA beam current, 1-μm beam diame- of 1–3 μm, from d = 1/(2nν) where n is the index of refraction,
ter, and silicates or oxides as standards. X-ray matrix corrections which is attributed to the spacing between the diamonds. The film
were based on a modified Armstrong (1988) CITZAF routine. Hy- thickness is smaller, below about 1 μm from relief observed in the
droxyl contents are determined by assuming that the weight per binocular microscope and from the diamond spacings indicated by
cent of all the oxides (measured) and of H 2 O (inferred) sum to 100 the fringes.
per cent. Analyses are included as an appendix to the electronic Thick films obtained between BaF 2 discs show a gradual rise
version of the paper and are available on the web (http://galena. in absorbance from frequencies just above the fundamental region
wustl.edu/∼dustspec/specpubs.html/). towards the visible. Transparent BaF 2 reflects only ∼5 per cent of
the light in the near-IR to visible, which is insufficient to account for
2.2 Infrared measurements the enhanced absorbance in Fig. 2(a) (see equations in Hofmeister
IR absorption spectra were acquired using an evacuated Bomem et al. 2003). Most likely, absorbance increasing with ν is a result
DA 3.02 Fourier transform spectrometer with an SiC light source of scattering. We therefore obtained peak parameters after baseline
at 1-cm−1 resolution. Instrumental accuracy is ∼0.01 cm−1 . An Si subtraction.
bolometer and a coated Mylar beamsplitter were used in the far- For a > 6, spectra are dominated by noise because the amount
IR, ∼50–650 cm−1 (∼200–15 μm). A 12-μm Mylar beamsplitter of light transmitted is comparable to the noise level of the detec-
was used for lower frequencies. An HgCdTe detector with a KBr tor. The sample is opaque for all practical purposes. For absorbance
beamsplitter covered the mid-IR from ∼450 to 4000 cm−1 (∼23– near 4, non-linear effects are expected due to a combination of de-
2.5 μm). Between 500 and 2000 scans were acquired. tector response and light leaking through cracks in the films. There-
Optically thin (∼0.2–2 μm) films were made by crushing particles fore, peak parameters were obtained when a < 3, and preferably
in a diamond anvil cell (DAC). Spectra were obtained from films of below 2, but large enough that the peak is distinct from spectral
various thicknesses to confirm the existence of weak peaks, to ascer- noise.
tain that optically thick conditions hold for the most intense peaks,
and to provide spectra free of fringes for the transparent regions. For 2.2.2 Determination of thickness through merging spectral
some measurements, water vapour remained in the sample cham- segments
ber, producing narrow, rotational peaks which were subtracted from
the mineral data. Very thick (d = 6 or 12 μm) films were made by To remove unwanted interference fringes and to provide high signal-
placing a small amount of powder in the centre of a 25-cm diameter to-noise ratios for all peaks, we merged spectral segments obtained
disc of BaF 2 . Thickness was controlled using a tin gasket (NSG from various thicknesses of each mineral. We used spectral segments
Precision Cells, Inc.) as a spacer between a pair of BaF 2 windows. with a < 2, and minimal interference fringes. Spectral segments
Lizardite spectra were obtained from a mat of fibres, 5 μm thick, were scaled in pairs so that the frequency regions common to both
made by compression. For details, see Hofmeister et al. (2003, 2004) segments matched. The process was repeated as needed to join far-,
and Bowey & Hofmeister (2005). mid- and near-IR segments. Comparison of mid-IR data obtained
using the DAC to near-IR data obtained using BaF 2 discs with con-
2.2.1 Infrared data analysis trolled spacing provided thicknesses of the mid-IR thin films, which
in turn provided the thickness of the far-IR thin film. Some of our
The absorption coefficient (A) is tied to the imaginary part of com-
raw data are shown to visually depict the process of merging and
plex index of refraction (k) through
scaling, whereas the final products are provided in the appendix and
A(ν) = 4πνk(ν). (1) are available on the web.


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
580 A. M. Hofmeister and J. E. Bowey
Uncertainties in t are connected with the degree of overlap among
4.1 Brucite
the spectral segments, and our confidence in avoiding distortion
when splicing spectral regions. More spectra were collected for Brucite has a simple IR spectrum (Fig. 2a). The O–H region is
some minerals, affording better matches of the segments and lower dominated by a sharp, strong peak. The water bending region is
uncertainties in thickness. Some O–H bands are too intense for relatively complex. Because brucite should lack fluid water, the
accurate representation in the thick films. In the spectra of the thick trio of peaks near 1600 cm−1 is assigned to Mg–O–H bending.
films, the fundamental modes are off-scale, and only the profiles near Overtone frequencies are low, ∼900 cm−1 , because fundamental
the base of these peaks are trustworthy. In some cases, the profiles frequencies occur below ∼500 cm−1 . Two far-IR peaks exist, as
near the base of the Si–O stretching mode are more reliable in the expected (Table 2).
thick film, than in the thin film spectra, because of the presence of For the thin-film spectrum, both narrowly spaced and widely
interference fringes in the latter. spaced interference fringes are superimposed in the transparent re-
gions (∼800–3000 cm−1 ). These fringes alter the peak shape of the
main band at high frequency, and thus the scaled thick film is a

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

better representation of the high-frequency side of the main Mg–O
3 CONNECTION OF INFRARED SPECTRAL
stretching band.
PAT T E R N S W I T H M I N E R A L S T RU C T U R E
The number of intense bands is consistent with symmetry analysis
The characteristic IR intensity patterns used to identify dust in space which assumes that the O–H dipoles have one crystallographically
largely arise from the different crystal structures possible for min- controlled orientation (Table 2; appendix). Additional weak bands
erals. Atomic arrangements of the hydrosilicates are more compli- (bars in Fig. 2a) may be more prominent in observational data, as
cated than those of pyroxene (Fig. 1a) and olivine now familiar to discussed below.
astronomers. Most are layered structures that can be described in
terms of two essential elements: ‘T’ layers containing corner-linked
4.1.1 Determination of thickness
SiO2−4 tetrahedra (Fig. 1c), and ‘O’ layers containing cations in edge-
linked octahedra (Figs 1d and e). Multiple ways of stacking these Brucite has few peaks, adding uncertainties in establishing film
layers create many different structures (Figs 1f–h). The complex thickness through comparison. The thickness of the thin film is ob-
chain silicate sapphirine (Fig. 1b) is included to distinguish transla- tained from comparison of three different peaks. (i) Comparing the
tions due to Ca from those due to Mg in amphiboles (Fig. 1i). For absorbances of the weak O–H stretching modes from the thick and
details see, for example, Wyckoff (1963), Smith & Bish (1988) or thin films suggests that t = 0.3 μm for the merged spectrum shown
Klein (2002). in Fig. 2(a). However, the weak O–H mode is barely resolved in
Ascertaining cation abundances through comparison of obser- the thin-film spectrum so this thickness is inaccurate. (ii) Compar-
vational and laboratory IR measurements requires knowing band- ing bending modes suggests t ∼ 0.8 μm. This estimate is rough
strengths which, in turn, requires correlating atomic motions with IR because the trio is poorly resolved in the thin-film spectrum. (iii)
peaks. Band assignments are suggested by peak positions as these Kramers–Kronig analysis of single-crystal reflectivity data (Mitra
depend largely on cation mass (e.g. Si–O tetrahedral stretching oc- 1969) gave A = 2.8 μm−1 for the intense O–H band in E||c, suggest-
curs near 10 μm whereas H–O stretching is near 3 μm). Band assign- ing that the thin film is ∼1.2 μm thick, assuming randomly oriented
ments are better constrained by determining the number, type, and grains. The later two estimates together indicate a film thickness of
polarization of the fundamental IR modes from symmetry analysis 1.0 ± 0.2 μm.
of the atomic sites (e.g. Cotton 1971; Fateley et al. 1972). Symmetry
analyses are deposited at the Washington University web site (and
included in the appendix). 4.1.2 Origin of ‘extra’ hydroxyl bands in brucite
The polarizations are mixed in absorption spectra of thin films. The O–H region has one more peak than is expected for OH dipoles
Fewer modes are observed than predicted because broad, over- parallel to the c-axis of the crystal structure. The intensity of the extra
lapping peaks are difficult to resolve. Talc units alternate with mode (Fig. 3, Table 2) increases with pressure whilst the strong
cation layers in the clay minerals, so the IR modes associated with mode at 3696 cm−1 weakens (Kruger, Williams & Jeanloz 1989;
talc should be recognizable in the spectra of montmorillonite and Shinoda & Aikawa 1998). The results at elevated pressure pertain
saponite (Figs 1g–h). to astronomical environments, as decreasing temperature similarly
contracts the lattice. Decreasing temperature also sharpens IR bands
(e.g. Bowey et al. 2001), which could permit resolution of these
4 I N F R A R E D S P E C T RO S C O P I C R E S U LT S weaker peaks in observational data.
Shinoda & Aikawa attributed the weak mode at 3645 cm−1 to
Fig. 2 shows the full spectral range for each type of hydrosilicate some OH− dipoles being tilted with respect to the c-axis (see
examined. The chemical composition nearest to the end member Fig. 1d). Given this configuration, we assign the trio of features
is shown for each structure. Because the various types of atomic near 1500 cm−1 and a weaker mode at 1650 cm−1 (Fig. 2a) to four
motions absorb in different frequency (wavelength) regions and Mg–O–H bending motions (Table 2).
because astronomical observations are frequently made over rel-
atively narrow spectral regions, we discuss our results in order
of wavelength: O–H stretching bands (λ near 2.7 μm), overtone- 4.2 Chain silicates
combination bands and water bending modes such as H–O–H,
4.2.1 Tremolite
Si–O–H or Mg–O–H (all near 5–8 μm), Si–O stretching (∼10 μm),
Si–O–Si and O–Si–O bending modes (∼12–25 μm), and lattice The large number of IR peaks (Fig. 2b) is connected with many
modes (>25 μm). Additional data on other compositions or related atoms being in the tremolite unit cell. Tremolite is characterized by
structures are presented, and comparisons are made. a strong, narrow band in the O–H region, a series of weak bands


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
 IR spectra of hydrosilicates 581

a b c

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

d e f g

h i

Figure 1. Essential structural elements in hydrous silicates. Triangles denote SiO4− 4 tetrahedra. Grey boxes denote the outline of an MO 6 octahedra, where M
is a cation such as Mg2+ . Black dots denote H+ ions. Open circles denote O2− anions. Dark grey circles denote Mg2+ sites. Light grey circles denote sites
for Ca+ or Na+ . At the top is shown a downward-looking view of the linkages of the SiO 4 tetrahedra: (a) single chain; (b) winged chain in sapphirine; (c)
tetrahedral ‘T’ sheet in the layer silicates. The double Si chains in tremolite can be derived by breaking the ‘T’ layer as shown by the grey bands, which allows
incorporation of OH− . In the middle section are edge-on schematic views of the three types of tightly bonded layers found in hydrous minerals. (d) Atomic
arrangement in the unit cell of brucite emphasizing the placement of the H atoms (after Shinoda & Aikawa 1998). The thin arrow shows the bending motion
of H+ tilted from the c-axis (thick arrow) of the crystalline lattice. (e) Octahedral ‘O’ brucite layers. The dashed line outlines the unit cell shown in (d), but
the H atoms are not shown. (f) T–O layers in lizardite or chrysotile, members of the serpentine group, showing placement of the H atoms. (g) T–O–T layers in
talc. In the lower section are edge-on views of more complex stacking sequences. (h) Layers in the clay minerals montmorillonite or saponite. The amount of
cations between the layers depends on the amount of Al3+ substituting for Mg2+ in the O layer. (i) Pseudo-layers in tremolite. Because the ‘T’ layer is broken
into stripes, so is the ‘O’ layer. The structure of sapphirine (not shown) has similar attributes.

in the overtone region, strong stretching and bending modes of the much higher frequencies typical of silicates (Table 4; Bowey &
SiO 4 tetrahedron, sharp, moderately intense peaks for the Si–O–Si Hofmeister 2005). We observed 40 out of the 57 predicted IR lattice
bending modes, and by a host of weak peaks in the far-IR. modes (Table 5; appendix).
The thickness of the mid-IR film is 0.62 μm. The uncertainty in
thickness is low in part because the O–H band is well resolved in 4.2.2 Effects of minor substitution
both thick- and thin-film spectra (Fig. 3).
All of the predicted hydroxyl modes (see the appendix) were Several samples of tremolite were examined to understand the
observed; the parameters are given in Table 3. Overtones occur at effect of chemical substitutions on IR peaks and to make band


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
582 A. M. Hofmeister and J. E. Bowey

λ, μm λ, μm
100 20 10 5 4 2.5 100 20 10 5 4 2.5
3
Mg-(OH) Brucite a c 5

Absorbance (thick film)

Mg(OH) 6 Chrysotile and
2
2 Lizardite 4

Absrobance (thick film)

Absorbance (thin films)
Absorbance (thin films)

t = 5 μm
2 t = 6 μm 4 3
H2O

M-O-H 2
C-H O-H
2
1

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

O.T. 1
1 Mg-O-H O.T.
t = 1 μm
f f
0 f 0
O-H f
* x t = 0.5 μm
F 0.8 < t < 1.2 μm -1
F F
x x f f f
0 -2 0 x
7 Absorbance (thick film) 2.5 -2
6
Si-O Tremolite b Talc d
6

Absorbance (thick film)

Canaan
2 2
5 f f
4
Absorbance (thin films)

Si-O-Si
4 t =6 μm
Absorbance (thin film)

O-Si-O t = 6 μm 1.5 O.T.

O.T.
3 2
1 1
O-H 2

1 x8 O-H 0
f 0.5 f
f
t = 0.62 μm 0 f t = 0.41 μm
x
F f f F
x F F
0 -1 0 -2
500 1500 2500 3500 500 1500 2500 3500
-1
Wavenumber, cm-1 Wavenumbers, cm

Figure 2. IR spectra over the full range for selected minerals. Thick-film spectra (t = 6 μm, grey line and left y-axis) have an increasing baseline due to
reflections from the BaF 2 windows. The true absorbance is near zero between the overtone-combination modes, which occur near 1800 cm−1 , and the O–H
stretching modes near 3600 cm−1 . Other spectra (black lines) are keyed to the right y-axis. The thin line denotes the thick-film spectrum scaled to match the
thickness of the thin-film spectrum. The medium line denotes the thin-film spectrum. The dotted line denotes the approximate linkage of spectra where the
merge was equivocal. An open circle denotes the point where far- and mid-IR spectra are merged. F or f denote two superimposed sets of interference fringes.
H 2 O denotes absorbed water. O.T. denotes overtone-combination bands. Vertical lines point to fundamental IR modes that may be difficult to view at this
scale. x–x denotes the section removed where diamonds strongly absorb. (a) Brucite. Grey triangle indicates poor resolution of the three peaks assigned to
Mg–O–H bending in the thin-film spectra. The mid-IR segment was divided by 1.3 to match the far-IR. (b) Tremolite from Canaan. The far-IR was divided
by 1.6 to match the mid-IR segment. (c) Lizardite (grey and light line) and chrysotile (thick line). The flat top (*) resulted from truncating the data. Some
organic contamination exists, marked as C–H. The mid-IR segment was multiplied by 1.28 to match the far-IR. (d) Talc. The dotted line denotes the spectrum
of natural talc obtained from a thickness of 2.62 μm, divided by 6.4 to match absorbance of the thinner film in the O–H region. The synthetic sample was used
for the 6-μm film; the other spectra are of the natural talc. The mid- and far-IR regions of the thin films were merged without scaling. (e) Saponite. The dotted
line denotes the PA sample, which appears to also contain Ni-rich talc. The PA spectra were offset by 0.2 absorbance units for clarity. The mid-IR of the pure
saponite was divided by 2.4 to merge with the far-IR. Thicknesses were determined by comparing the three bands near 1500 cm−1 ; the PA sample has t ∼ 1.3
μm. (f) Montmorillonite. The dotted line denotes the Wyoming sample, which has a small amount of quartz, the thin line denotes quartz, and the dashed line
denotes the thick film of Wyoming clay. The left y-axis is plotted to provide the measured absorbance for the thick-film spectrum.

assignments. The Canaan sample is closest to end-member tremo- ment with Robert, Ventura & Hawthorne (1999). A few of the
lite. Although the chemical substitutions involve different sites lattice modes are reduced in intensity by the partial F substitu-
and different amounts (Table 1), the IR spectra vary little tion, and this observation is used to deduce which of the ob-
(Fig. 4). The F-rich sample from Campolungo has lower inten- served peaks is connected with the bending of the Mg–O–H unit
sity in the OH region, but the same peak position, in agree- (Table 5).


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
 IR spectra of hydrosilicates 583

λ, μm Table 2. Brucite peak parameters.

100 20 10 5 4 2.5
5 ν (cm−1 ) A (1 μm−1 ) Assignment
thick
Saponite e

Absorbance (thick film)

367a 2.35a Mg–(OH) stretch ||c
3 t = 1.3 μm O-H 442a 2.5a Mg–(OH) stretch ⊥c
4
PA H2O 805.7 0.008 Overtone ∼
= 367 + 442
Scotland 853.7 0.013 Overtone
O.T. water 887.3 0.01 Overtone ∼
= 2 × 442
3
1118 ∼0.006 Overtone ∼
= 2× LAb
Absrobance (thin films)

t = 6 μm 1419.6 0.23 Mg–O–H bend straightc

2
1481.1 0.19 Mg–O–H bend straightc
O.T.? 2 1516 ∼0.06 Mg–O–H bend tiltedc
1656 ∼0.02 Mg–O–H bend tiltedc

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

3450 Variable Sorbed H 2 O
1 3652 ∼0.048 O–H stretch tilted
1 O-H 3696 1.4d O–H stretch straight
a Parameters from the thin film. The thickness of the thin film is inferred
f t = 1 μm 0 to be 1 μm from comparison of peak heights (see text). A for the other
peaks was taken from the 6 - μm film. b Suggests a resonance at 560 cm−1 ,
PA probably arising from the longitudinal acoustic mode. c Straight and tilted
0.84 μm
02 -1 refer to the two orientations of O–H dipoles parallel to c and tilted from
the c-axis (Fig. 1d). Symmetry analysis after Buchanan, Caspers & Murphy
Montmorillonite f (1963). d This is half the absorbance determined by Mitra (1969) from
10 reflectivity ||c, to account for a random orientation.

1.5 WY
t = ~1.2 μm
accompanied by many weak peaks at lower frequency (Table 4).
Absrobance (thin film)

Absorbance (thick film)

Our mid-IR data agree with dispersion spectra (Christy et al. 1992),
U.K. except that the latter has strong peaks near 1100 and 1200 cm−1 ,
1 t = ~0.75 μm which are probably longitudinal optic (LO) modes.
O-H Weak, far-IR modes occur at the same frequencies as seen for
HO 5 tremolite, but peak positions for the silicate modes in the mid-IR
2
O.T. WY region from the two minerals do not match (Fig. 4b, Table 5). These
Q HO observations are consistent with sapphirine having Al in some tetra-
0.5 2
t = 1 μm U.K. hedral sites and a different arrangement of the chains. The far-IR
is produced by translations of the Mg and Ca cations in tremolite,
and of Mg in sapphirine. Because of similarly sized sites for Mg (cf.
Moore 1969; Smith & Bish 1988), the far-IR frequencies are quite
0 quartz 0 close. Far-IR modes in tremolite that are not replicated in sapphirine
500 1500 2500 -1 3500 are assigned to Ca translations (Table 5). The number of Ca peaks is
Wavenumbers, cm consistent with symmetry analysis of tremolite (see the appendix).
One weak mode was seen in the overtone region (Fig. 4b, λ =
Figure 2 – continued
7 μm, Table 4), other peaks appear to be fringes. We could not easily
join these spectral segments. As sapphirine is hard, its thick film has
4.2.3 Effect of major Fe substitution air spaces, and thus its absorbance values are underestimated.

Significant substitution of Fe for Mg in actinolite little alters its peak

positions from those of tremolite (Fig. 4a). The presence of multiple 4.3 Serpentine minerals
O–H peaks and their relative intensities (Fig. 3) are connected with
Our serpentine sample contains both lizardite and chrysotile poly-
the number of Fe atoms in the four surrounding sites (e.g. Hawthorne
morphs (Cogulu & Laurent 1984). IR spectra of the platy (lizardite)
1981).
and fibrous (orthorhombic chrysotile) samples are virtually the
Mid-IR modes (ν > 350 cm−1 ) assigned to motions involving
same, although some of the weaker peaks are better resolved in
Si and O are little affected by Fe substitution, as expected. Far-IR
the chrysotile spectrum, as shown in Fig. 2(c). In the rest of the
modes are generally less intense than those of tremolite, and more
paper therefore we do not distinguish between these polymorphs.
modes are observed at low frequency as a result of the greater mass
Two O–H stretching modes exist (Table 3), as expected (see
of Fe ions. The mode at 238 cm−1 is not affected, suggesting that
the appendix). The overtone region has few modes, as a result of
this is connected with Ca ions.
the small number of IR fundamental modes. The band near 1583
cm−1 is assigned to Mg–O–H bending (Table 7), because this fea-
ture resembles the bending modes in brucite (cf. Figs 2a and c),
4.2.4 Comparison with an anhydrous chain silicate
and because OH sites in these minerals are similar (cf. Figs 1d
Translatory modes of Ca in tremolite are assigned through com- and f).
parison with sapphirine. The IR spectrum of sapphirine contains The thin-film spectra have strong fringes in the transparent region
multiple overlapping bands from about 400 to 1100 cm−1 (Fig. 4b), above 1000 cm−1 and near 700 cm−1 which were removed when


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
584 A. M. Hofmeister and J. E. Bowey

λ, μm
2.8 2.75 2.7
1 0.6

Absorbance (actinolite, tremolite)

OH region
0.8
t = 1 μm 0.4

actinolite
0.6

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

0.2
Absrobance

tremolite

0.4 0
saponite
chrysotile
0.2 -0.2
brucite
* talc
montm.
0 -0.4
3560 3600 3640 3680 3720
-1
Wavenumbers, cm
Figure 3. Expanded view of the O–H region plotted for thicknesses of 1 μm. The actinolite spectrum is scaled so that intensities of its Si–O–Si bending modes
match those in the tremolite (from Canaan) spectrum. Raw data are shown. Because saponite and montmorillonite have fluid H 2 O, their O–H peak shapes are
asymmetrical, because these peaks overlap with the tail from the H 2 O IR band. The very weak O–H stretching band in brucite is shown by an asterisk.

Table 3. Parameters for the O–H stretching region. ily) have broader and shifted peaks (Mellini et al. 2002). These
differences are probably due to higher Fe content of 15–20 atom per
Phase ν FWHM A cent, compared to <5 atom per cent in the present samples. More
(cm−1 ) (cm−1 ) (1 μm−1 ) Fe means more disorder (broader peaks) and shifts to lower fre-
Brucite 3696 15 1.4a quency (larger bond distances and/or greater cation mass). Our data
3652 ∼6 ∼0.048 resemble measurements of serpentine (a mixture of polymorphs)
Tremolite 3674.5 3.0 0.396 by Koike & Shibai (1990), but reveal more far-IR modes. Band-
Actinolite 3674.3 5.0 0.17 strengths of the main peaks (near 10 and 20 μm) measured here are
3661.1 3.4 0.092 almost double those of Koike & Shibai (1990), suggesting that their
3658.0 ? Shoulder dispersion contains some large grains or clumps of particulates that
3643.0 <6 0.021 do not transmit light.
3624 ? ∼0.005
Lizardite 3686.5 27 0.95
∼3650 26 0.254
Talc 3660.3b 3.0 0.045 4.4 Talc
3677.1 3.3 0.35
Saponite 3670 36c 0.112 Talc has a relatively simple spectrum (Fig. 2d) with very sharp peaks.
Montmorillonite 3627 ∼74c 0.21 Because of the small number of bands, many interference fringes
∼3400 ∼400 0.5 are seen. As occurred for chrysotile, the thickness could be well
a From
constrained.
Mitra (1969). b Natural talc (#2) has an additional O–H peak at
One strong O–H stretch exists, in accord with symmetry, but
3662 cm−1 (see text). c Twice the HWHM.
is accompanied by a much weaker peak, which is stronger in the
natural specimen than in the synthetic. This peak could be due to Fe
merging. Peaks are sharp, especially in the far-IR (Fig. 5). Because impurities or a tilted orientation.
of the narrow widths, peaks are resolved and 19 of the 21 modes We observed 18 of the 27 expected fundamental modes (Table 6;
predicted by symmetry analysis were observed (Table 6). appendix). The tetrahedral silicate layers dominate because the spec-
Compared to our results, mid-IR dispersion spectra of lizardite trum of talc closely resembles that of chrysotile (Figs 2c and d). Talc
and antigorite (the third polymorph of the serpentine mineral fam- lacks the OH− dipole between the layers (Figs 1e and f), leading to


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
 IR spectra of hydrosilicates 585
Table 4. Overtone peak positions from the thick (t = 6 μm) films. Sum is Table 5. Peak positions (ν in cm−1 ) and absorption coefficients (A in μm−1 )
the possible combination of fundamental modes. An assignment of all bands for the chain silicates. Above 375 cm−1 , the match between the amphibole
cannot be made (indicated as ‘?’) because Raman-active species can produce and sapphirine spectra is poor, consistent with these modes arising from
overtones in IR measurements. ‘Water’ indicates the H–O–H bend of fluid different types of chains.
water.
Tremolite Actinolite Sapphirine
ν(cm−1 ) Sum ν (cm−1 ) Sum Assignment ν A ν ν
Tremolite Talc T(chain) 84 0.034 70 –
1243 470+761 1253 466+780 T(Mg 3 O) 144 0.147 140 139
1295 527+761 1318 541+780 T(chain) 158 0.19 155 –
1354 2×687 1359 2×670 T(Mg) 170 0.01 – 170
1499 509+992 1398 385+1020 T(Mg) 182 0.10 177 181
1630 Water 1526 ? T(Mg) 190 0.16 191 187

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

1694 761+922 1645 ? T(Al) – – 205
1784 687+1114 1673 ? T(Ca) 220 0.10 216 –
1830 2×922 1710 452+466+790 T(Ca) 227 0.10 225 –
1871 922+960 1765 ? T(Ca) 238 0.28 238 –
1906 2×960 1803 780+1020 Mg–O 244 0.2 248
1956 960+992 1834 789+1020 Mg–O 254 0.14 ∼250 257
2079 960+1114 1868 394+452+1020 Mg–O 283 0.09 278 282
1920 2×452+1020 ? 288 0.09 289 –
1952 2×466+1020 ? 303 0.15 300 –
Sapphirine Saponite Mg–O 318 0.28 315 315
1469 486+981 1350 2×674 quest; 338 0.09 338 –
1415 ∼415+1014 R? 362 0.59 357 –
Lizardite/chrysotile 1457 448+1014 R? 375 0.2 373 370
1201 2×614 1559 534+1014 R? – 381
1268 305+964 1624 Water R? 392 0.46 389
1306 411+437+462 Montmorillonite O–Si–O 401 0.25 401 412
1463 964 1625 Water O–Si–O 420 0.25 417 431
1583 614+964 1728 845+881 O–Si–O 444 sh 443 449
1624 667+964 1872 845+1038? O–Si–O – 466 486
1705 374+2×667 1980 920+1038? O–Si–O 470 1.79 476 502
O–Si–O 499 0.01 530
O–Si–O 509 ∼1.3 508 544
O–Si–O 527 sh 530 561
a series of weak overtones occurring near 1700 cm−1 , as occurred O–Si–O – 587 589
for tremolite. Al–O – – 610
Mg–O–H 645 0.15 642 –
Chain 665 0.15 660 633
4.5 Clay minerals Chain 687 0.60 688 692
Chain 725 0.057 724 725
X-ray diffraction analyses of saponite and montmorillonite indicated Mg–O–H 740.5 0.017 737.6
single phases, but IR analysis revealed small amounts of quartz im- Mg–O–H 761 0.92 759 752
purities. The fine grained nature of clays is not amenable to physical Si–O 833 772
separation, so comparison was made to spectra of quartz (essentially Si–O 884 823
stoichiometric SiO 2 from Lisbon, Portugal) to ascertain the spectra Si–O 922 1.6 922 851
of the pure clays. Si–O 960 ∼2.0 959 878
Si–O 992 ∼3.3 988 903
Si–O 1039 sh 1035 949
4.5.1 Saponite Si–O 1079 1.5 1066 981
Si–O 1114 1.8 1111 1053
More peaks are seen in the spectrum of the Pennsylvania (PA) sam-
ple (Fig. 2e) than of the Scottish sample. Willemseite, which is talc
with Ni substituting for Mg ions (Shepard 1848), probably con-
4.5.2 Montmorillonite
taminated our sample. Consequently, the remainder of the paper
concerns the Scottish sample. The Wyoming (WY) sample has a significant amount of quartz im-
The O–H region consists of a broad band from about 3200 to purities, roughly 10 per cent of the volume (Fig. 2f). The UK clay has
3600 cm−1 , due to interlayer water molecules. The band is asym- fewer impurities, ∼2 per cent. Spectral subtraction provided unsat-
metric because it also includes a contribution from stoichiometric isfactory results probably because the quartz is unevenly distributed.
OH− near 3600 cm−1 . The 1630-cm−1 peak is indicative of molecu- For the deposited spectra, we therefore removed the affected spec-
lar H 2 O (e.g. Rossman 1988). The overtone region has strong peaks, tral segment of the UK sample near 800 cm−1 . The spectra resemble
as occurs in brucite. The thickness of the thin film was determined both talc and saponite, mainly differing in the weakness of the peaks
by comparison to this region, rather than to the O–H region. in the Si–O–Si bending region. Structurally, the main difference be-
The fundamental region of the Scottish saponite resembles that tween these clays is that saponite has Mg in the octahedral layer,
of talc, being broader with fewer peaks (Table 6). whereas montmorillonite has Al.


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
586 A. M. Hofmeister and J. E. Bowey

Amphiboles a

Absorbance (Actinolite, Campolungo)

3

Absorbance (Canaan, Balmat)

4
Canaan x10
2

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

1
2
actinolite

0
Campolungo
t = 1 μm

-1 Balmat Canaan
0

b
2 Sapphirine
Absorbance

1 merge

x5

0
20 40 60 80 100 120 140
Wavelength, μm
Figure 4. Lattice modes of chain silicates using a wavelength scale. (a) Amphiboles. The grey line denotes the Canaan sample, plotted for t = 10 μm. The
thick line denotes actinolite. The thin line denotes F-rich tremolite. The Campolungo sample was very thin, and the spectrum is mostly noise for λ > 55 μm.
The thick line (far-IR only) denotes tremolite from Balmat. (The Balmat mid-IR spectrum does not differ from the Canaan mid-IR spectrum, and is not shown.)
The dotted line denotes Canaan tremolite, plotted at 1-μm thickness. Weak peaks are marked with vertical bars. (b) Sapphirine. The thick line denotes merged
thin-film data, and the thin line the expanded view at low frequency.

Montmorillonite has a broad band from 3200 to 3500 cm−1 , as ties. The weak band observed by Koike & Shibai (1990) near
in saponite, but its hydroxyl peak near 3600 cm−1 is better resolved 110 cm−1 appears to be intrinsic, as we observed this in spectra
(Fig. 2f). The amount of interlayer water is lower, consistent with obtained from different film thicknesses. This band was used by
the lower intensity of the overtone-combination bands. Because the Malfait et al. (1998) to identify dust in space. Low-temperature
interlayer water content varies, the thickness of the thin film is not spectra are needed to better resolve this weak, broad feature.
well constrained.
Our data resemble previous mid-IR powder reflectance and trans- 5 DISCUSSION
mission spectra (Esposito, Colangeli & Palomba 2000). Since
Wyoming montmorillonite was examined, the classical dispersion All minerals absorb moderately to weakly below ∼400 cm−1 , where
analysis of Esposito, Colangeli & Palomba (2000) may include lattice vibrations involving the cations occur, but strongly at frequen-
quartz peaks. Our bandstrengths are 20 per cent below those of cies between ∼400 and 1200 cm−1 , where bending and stretch-
Koike & Shibai (1990), possibly due to experimental uncertain- ing modes of the SiO2− 4 tetrahedron occur, and moderately near


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
 IR spectra of hydrosilicates 587

Wavenumbers, cm-1
400 200 125 100 80
0.5
Far-IR
t = 1 μm
0.4

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

0.3
Absorbance

sapphirine

tremolite
0.2
saponite
F F

0.1 chrysotile

montm.
talc
0
40 60 80 100 120
Wavelength, μm
Figure 5. Expanded view of the far-IR region using a wavelength scale. Spectra are plotted for thickness of 1 μm. The thin line denotes sapphirine, dashed
curve Canaan tremolite, thick curve saponite, medium solid line chrysotile, grey curve montmorillonite, and dotted curve talc. Bars mark the longest wavelength
peak of each mineral.

3600 cm−1 , where O–H stretching occurs. As discussed in Hofmeis- 5.1 Thickness and bandstrengths
ter et al. (2003, 2004), thin-film spectra appear to closely represent
the intrinsic absorption spectra of the solids in the absorbing regions Variations in intensity with spectral region and the softness of
when the bands are not extremely broad. Except for brucite, peaks the hydrous samples permit determining thicknesses of the thin
in the minerals studied here meet this criterion. films through comparison with thick films. This method is not
Although common features exist, IR spectra of the seven mineral tractable for hard, anhydrous species, e.g. sapphirine. Optically thin
types presented here are distinguished by the number of vibrational films of brucite and the hydrosilicates created by compression have
modes, their distribution over frequency and their relative intensi- t  1 μm, which is similar to thicknesses inferred from comparing
ties. In fact, the signatures in each spectral range are distinct. Far-IR measured absorbance to calculated absorption coefficients of MgO
patterns (Fig. 5) show the greatest variety. The overtone (∼1000– (Hofmeister et al. 2003) and of hibonite (Hofmeister et al. 2004).
2500 cm−1 ; Fig. 2) and O–H stretching regions (∼3000–3800 cm−1 ; Better data are obtained in the spectral region containing the strong
Fig. 3) have intensities similar to those in the far-IR and display Si–O fundamental modes by using films with thickness ∼0.5 μm,
similar variety, but the overtones are both broad and weak and the whereas resolving weak modes at low frequency from the noise
O–H bands are generally confined to the narrower spectral region requires thickness of several μm.
of 3300–3700 cm−1 (e.g. Rossman 1988). Less variety occurs in Weak overtone-combination bands in the region ∼1200–
the mid-IR, with intense, generally structured bands near 400 and 2500 cm−1 and H–O–H bending of water molecules near 1600 cm−1
1000 cm−1 , because of the internal bending and stretching motions are well resolved in the t = 6 μm films of the chain silicates, serpen-
of the SiO4− 4 tetrahedra. Sapphirine (Fig. 4) has a different mid-IR
tines and talc, but are obscured by interference fringes in thin-film
pattern than the hydrosilicates (Fig. 2), as a result of Al extensively data. Thus, overtones are not useful in determining bandstrengths
replacing Si in the tetrahedra. of the fundamentals, except for phases containing layers that are
On the basis of pattern alone, the far-IR affords the best opportu- hydrogen bonded to other layers in the structure (brucite) or to hy-
nity to identify mineral structure by comparing remotely sensed to drogen molecules between the layers (clays), both of which have
laboratory data. strong overtones for t < 1 μm.


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
588 A. M. Hofmeister and J. E. Bowey
Table 6. Peak positions and absorption coefficients for the layered silicates. Band assignments are approximate.

Lizardite Talc Saponite Montmorillonite

Thickness 0.50 μm Thickness 0.41 μm Thickness 1.0 μm Thickness 0.75 μm
Assignment ν (cm−1 ) A (μm−1 ) ν (cm−1 ) A (μm−1 ) ν (cm−1 ) A (μm−1 ) ν (cm−1 ) A (μm−1 )

T(Ca, Na) 110 0.013

T(oct) 131 0.046 170 0.061 157 0.050
T(M1) 179 0.043
204 0.007 196 0.008
oct. 230 0.20 232 0.0038
T(M1)? 260 0.073 251 sh 254 sh
Mg–O? 305 0.67 293 0.083 304 sh
Mg–O? 321 0.55 313 0.025 319 ∼0.01

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

Mg–O? 348 sh 345 0.126 342 0.08
R(SiO 4 ) 374 sh
R(SiO 4 ) 386 0.8 385 0.8 384 0.60
405 1.5 394 1.9
422 ∼0.3 425 0.9 ∼415 sh 424 sh
O–Si–O 437 3.7 448 sh
O–Si–O 456 ∼0.7 452 1.3 458 2.0 467 1.04
466 4.3
O–Si–O 502 sh 508 sh 534 sh 524 0.83
553 0.5 541 0.02 575 sh
Si–O–Si 607 1.2 625 0.02
Si–O–Si 667 0.7 670 1.93 674 0.52
780 0.048 770 sh
789 0.036
860 sh 845 0.02
881 0.080
Si–O 965 2.82 960 sh 943 sh 920 0.16
Si–O 1016 ∼1.8 1020 5.80 1014 2.3 1038 1.5
LO 1073 1.1 1130 1120 1124 0.5

Table 7. Transferrable bandstrengths obtained from the tables and figures. True absorption coefficients (in μm−1 ) are given for the dominant band in the
various spectral regions. Mass extinction coefficients can be obtained using equation (4).

ν(cm−1 ) Brucite Tremolite Lizardite Talc Saponite Montmorillonite Average

Formula – Mg(OH) 2 Ca 2 Mg 5 – Mg 3 Si 2 O 5 (OH) 4 Mg 3 Si 4 O 10 (OH) 2 Na 0.3 Mg 3 (Al 0.3 Si 3.7 )– Na 0.3 (Mg 0.3 Al 1.7 )–
Si 8 O 22 (OH) 2 O 10 (OH) 2 · 4H 2 O Si 4 O 10 (OH) 2 ·nH 2 O
Density (g cm−3 ) – 2.377 3.01 2.58 2.776 2.10 2.06 2.48
O–H stretch 3500 1.4 0.40 0.95 0.45 0.11 0.21 0.6
Mg–O–H 1480 0.2 0.9 0.045 – 0.2 0.08 0.3
Overtones 2000 0.01 0.05 0.045 0.025 0.1 0.014 0.04
Si–O stretch 1000 – 3.3 2.8 5.8 2.3 1.6 3.1
Si–O–Si bend 670 – 0.6 1.2 1.9 0.5 ? 1.0
O–Si–O bend 450 – 1.8 3.7 4.3 2.0 1.0 2.5
Mg–O stretch 300 2.4 0.3 0.67 – – 1.0
Ca translation 200 – 0.3 – ? 0.01 0.15
Mg translation 200 – 0.1 0.05 0.08 0.05 0.01 0.06

As shown in Fig. 2, observing undistorted peak maxima for the cable bandstrengths, we did not attempt detailed fitting, but instead
Si–O stretching and O–Si–O bending fundamentals requires thick- report on rough bandstrengths in Tables 2, 3, 5 and 6. Table 7 summa-
nesses below 1 μm and, for some cases, below 0.5 μm. Larger thick- rizes the bandstrength for the dominant peaks of each mineral. For
nesses round the profiles, leading to bands that appear artificially each type of vibrational mode, a consistent range of bandstrengths
broad, and sometimes shifted. Such shifts occur in IR measure- is obtained. Talc, for which most bands are singlets, has high inten-
ments utilizing the dispersion method, and are due to overly large sities compared to the other minerals with multicomponent peaks.
particles or particle clumping, rather than the use of a KBr ma- This difference is expected, because peak area actually represents
trix (see Speck, Hofmeister & Barlow 1999; Bowey & Hofmeister the number of oscillators producing the peak. Given the variations
2005). in numbers of peaks and overlap for the diverse silicate minerals,
Strengths of well-resolved bands were determined from splicing we simply average the bandstrengths of the five hydrosilicates to
spectral segments for each mineral. As our interest is in widely appli- provide values that can be applied to any silicate mineral, not just


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
 IR spectra of hydrosilicates 589

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

Figure 6. Matches of laboratory spectra to features in the observed spectrum of the planetary nebula NGC 6302 after subtraction of a continuum comprised
of blackbodies at 80, 65 and 50 K after Molster et al. (2001). Features in the 17–50 μm region matched with serpentine are indicated with vertical lines. The
broad 63-μm feature is most closely matched by saponite. The laboratory spectra have been multiplied by Planck functions at temperatures of 40 or 60 K as
indicated; the scaling of these spectra is arbitrary, but is the same in both wavelength ranges. Spectral artefacts at 25.8 μm and gas-phase emission lines are
omitted from the nebula spectrum.


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
590 A. M. Hofmeister and J. E. Bowey
hydrosilicates. These values are appropriate only under optically can be saponite and/or montmorillonite) are present in substantial
thin conditions. amounts in high-temperature vapour-phase combustion experiments
(summarized by Nuth et al. 2002). Possibly, the high density of this
nebula could have promoted formation of dehyroxalates along with
5.2 Comparison with NGC 6302
the already identified refractories. At long wavelengths the spectra
NGC 6302 is a strongly bipolar planetary nebula which formed of lizardite (or saponite) and its dehydroxylate should be almost
from an intermediate mass progenitor star; the nebula contains a identical.
dense dusty torus and ionized polar regions. The 2.4–197 μm spec-
trum of the nebula was obtained with the LWS and SWS on the
6 CONCLUSIONS
Infrared Space Observatory and published by Molster et al. (2001);
shortwards of 17 μm the spectrum is dominated by polycyclic aro- In this paper we document IR spectral patterns, peak positions and
matic hydrocarbon (PAH) bands. However, at longer wavelengths intensities for hydrosilicates and a few other phases at room tem-

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

the spectrum is characterized by the narrow features of crystalline perature over a wide frequency range. In space, dust exists at low
silicates (Fig. 6). Molster et al. identified many bands with Mg end temperatures, which reduces peak widths and shifts peak positions.
members of olivine (forsterite = Mg 2 SiO 4 ; Fo) and of pyroxene Variation in optical path-length (connected with a range of grain
(enstatite = MgSiO 3 ; En). The broad 63.6-μm feature has pre- size for dust in space) broadens and rounds the peaks, as can be
viously been identified with diopside (CaMgSi 2 O 6 ; Koike et al. seen in some of our spectra. It is the combination of these opposing
2000), or a blend of diopside, crystalline H 2 O ice with enstatite effects which makes comparison of observational data to laboratory
(Molster et al. 2001), or with a mixture of H 2 O ice and dolomite measurements viable, but with some uncertainties. For robust iden-
[MgCa(CO 3 ) 2 ; Kemper et al. 2002b], or of grossite (CaAl 4 O 7 ) and tification, additional measurements are needed at low temperatures,
hibonite (CaAl 12 O 19 ; Hofmeister et al. 2004). Molster et al. also as well as an improved understanding of the effects of grain-size
suggest that both the 43.54- and 52.04-μm features are due to crys- distributions, dilutions and mixtures on peak profiles. On the basis
talline H 2 O ice. of our present and previous examinations of NGC 6302, we suggest
Following the method of Molster et al. (2001), who matched that laboratory studies of dehydroxylates are worthwhile, as well as
the observations with forsterite and enstatite spectra multiplied by a wide range of calcium bearing minerals.
65- and 70-K blackbodies, we compared our hydrous silicates with
their continuum-subtracted spectrum of NGC 6302 by multiplying
AC K N OW L E D G M E N T S
our laboratory spectra by 20–100 K blackbodies. We found that the
best matches were obtained with blackbody temperatures of 60 or This work was supported by NSF-AST-9805924 and NASA-
40 K for the different minerals (Fig. 6). This effectively increases APRA04-0000-0041. We thank N. Johnson and B. Fegley (Wash-
the strength of their far-IR features with respect to bands shortwards ington University) for samples of brucite, lizardite, chrysotile and
of ∼30 μm the tremolites; R. F. Dymek (Washington University) for sapphirine
In the 17–50 μm range (Fig. 6a), lizardite bands at 22.88, 24.69, and a chemical analysis; Carl Francis (Harvard University) for acti-
32.79 and 43.48 μm best match features in NGC 6302. Amphiboles nolite; M. Grady (British Museum of Natural History) for clay
(tremolite and actinolite) and brucite might also contribute to fea- minerals. Campolungo tremolite was provided by P. Brack (ETH,
tures in the 20–45 μm range. However, no features in this region Zürich). We thank D. M. Jenkins (University of Binghamton) for
can be identified with talc, or the clays because the low tempera- discussions. Special thanks to G. Cressy (British Museum of Natural
tures suppress their strong 10–30 μm bands. The 50–85 μm range History) who identified phases using X-ray diffraction methods.
of NGC 6302 (Fig. 6b) is dominated by a ∼15 μm wide 63.6-μm
feature, which is closely matched by our spectrum of saponite, al-
REFERENCES
though lizardite might also contribute. The presence of substantial
quantities of tremolite and talc is precluded by the absence of narrow Armstrong J., 1988, in Newbury D. E., ed., Microbeam Analysis. San
bands in the 50–60 μm range of NGC 6302. Francisco Press, San Francisco, p. 469
The nebula spectrum also has a feature at 91 μm. Of the materials Bowey J. E., Adamson A. J., 2002, MNRAS, 334, 94
examined here, only actinolite and montmorillonite have features Bowey J. E., Hofmeister A. M., 2005, MNRAS, 358, 1383
near this wavelength. The features are not well enough resolved Bowey J. E., Lee C., Tucker C., Hofmeister A. M., Ade P. A. R., Barlow
M. J., 2001, MNRAS, 325, 886
above the noise to provide a trustworthy peak shape, so fitting was
Brearley A. J., 1993, Geochim. Cosmochim. Acta, 57, 1521
not attempted. Kemper et al. (2002a,b) suggested calcite as the ori- Buchanan R. A., Caspers H. H., Murphy J., 1963, Appl. Opt., 2, 1147
gin, whereas Hofmeister et al. (2004) suggest that hibonite, grossite Chihara H., Koike C., Tsuchiyama A., Tachibana S., Sakamoto D., 2002,
and melilite minerals (e.g. Ca 2 MgSi 2 O 7 ) are likely as these match A&A, 391, 267
some of the shorter wavelength peaks as well. Christy A. G., Phillips B. L., Guttler B. K., Kirkpatrick R. J., 1992, Am.
In short, of the samples examined here, lizardite and saponite may Mineral., 77, 8
be present as dust particles within NGC 6302. In terrestrial environ- Cogulu E., Laurent R., 1984, Can. Mineral., 22, 173
ments, lizardite forms by hydration of forsterite at temperatures Cohen M., Barlow M. J., Sylvester R. J., Liu X. W., Cox P., Lim T., Schmitt
below 400 ◦ C (e.g. Deer, Howie & Zussman 1992). As forsterite is a B., Speck A. K., 1999, ApJ, 513, L135
major constituent of finer dust grains in the nebula, lizardite could be Cotton F. A., 1971, Chemical Applications of Group Theory.
Wiley-Interscience, New York
derived from some of these refractory dust particles by hydrous al-
Deer W. A., Howie R. A., Zussman J., 1963, Rock-Forming Minerals. Long-
teration. Saponite forms from extensive, low-temperature alteration mans, London
of rocks with olivine, pyroxene and calcic feldspars, and therefore Deer W. A., Howie R. A., Zussman J., 1992, An Introduction to the Rock-
seems likely to be present, unless this can form from alteration of Forming Minerals. Wiley, New York
hibonite or grossite. Alternatively, dehydroxalates of both serpen- Deloule E., Robert F., Douknan J., 1998, Geochim. Cosmochim. Acta, 78,
tine (which is lizardite and its polymorphs) and smectite (which 957


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591
 IR spectra of hydrosilicates 591
Esposito F., Colangeli L., Palomba E., 2000, J. Geophys. Res., 105, 17 643 Smith J. R., Bish D., 1988, Crystal Structures and Cation Sites of the Rock-
Fabian D., Henning T., Jaeger C., Mutschke H., Dorschner J., Wehrhan O., Forming Minerals. Allen & Unwin, Boston
2001, A&A, 378, 228 Speck A. K., Hofmeister A. M., Barlow M. J., 1999, ApJ, 513, L87
Farmer V. C., 1974, The Infrared Spectra of Minerals. Mineralogical Society, Sylvester R. J., Kemper F., Barlow M. J., de Jong T., Waters L. B. F. M.,
London Tielens A. G. G. M., Omont A., 1999, A&A, 352, 1034
Fateley W. G., Dollish F. R., McDevitt N. T., Bentley F. F., 1972, Infrared Waters L. B. F. M. et al., 1996, A&A, 315, L361
and Raman Selection Rules for Molecular and Lattice Vibrations: the Weaver C. E., Pollard L. D., 1973, Dev. Sediment., 15, 55
Correlation Method. Wiley-Interscience, New York Williams H. R., 1984, Can. Mineral., 22, 417
Gail H.-P., Sedlmayr E., 1999, A&A, 347, 594 Wooden D. H., Harker D. E., Woodward C. E., Butner H. M., Koike C.,
Hawthorne F. C., 1981, Rev. Mineral., 9, 103 Witteborn F. C., McMurty C. W., 1999, ApJ, 517, 1034
Hofmeister A. M., Keppel E., Bowey J. E., Speck A. K., 2000, in Salama Wyckoff R. W. G., 1963, Crystal Structures, Vol. 1. Wiley, New York
A., Kessler M. F., Leech K., Schulz B., eds, Proc. Conf. ISO beyond the Zolensky M., McSween H. Y., 1988, in Kerridge J. F., Matthews M. S.,
Peaks, ESA SP-456. ESA Publications Division, Noordwijk, pp. 343– eds, Meteorites and the Early Solar System. University Arizona Press,
346 Tucson, pp. 45–63

Downloaded from https://academic.oup.com/mnras/article-abstract/367/2/577/1012099 by guest on 05 April 2020

Hofmeister A. M., Keppel E., Speck A. K., 2003, MNRAS, 345, 16
Hofmeister A. M., Wopenka B., Locock A. K., 2004, Geochim. Cosmochim.
Acta, 68, 4485 S U P P L E M E N TA RY M AT E R I A L
Johnson N. M., Fegley B. Jr, 2003a, Icarus, 165, 340 The following supplementary material is available for download as
Johnson N. M., Fegley B. Jr, 2003b, Icarus, 164, 317
part of the full-text version of the article from http://www.blackwell-
Keller L. P., Buseck P. R., 1990, Geochim. Cosmochim. Acta, 54, 1155
Kemper F., Jaeger C., Waters L. B. F. M., 2002a, Nat, 415, 295
synergy.com.
Kemper F., Molster F. J., Jaeger C., Waters L. B. F. M., 2002b, A&A, 394, Appendix. Supplementary chemical analyses, symmetry analy-
679 ses and spectra. Includes Tables A1–A6.
Klein C., 2002, Mineral Science. Wiley, New York Table S1. Actinolite is a tab-delimited text of the merged spectra
Knacke R. F., McCorkle S., Puetter R. C., Erickson E., 1985, AJ, 90, 1828
shown in Fig. 4a.
Koike C., Shibai H., 1990, MNRAS, 246, 332
Koike C. et al., 2000, A&A, 363, 1115
Table S2. Sapphirine.txt is a tab-delimited text of the merged
Kruger M. B., Williams Q., Jeanloz R., 1989, J. Chem. Phys., 91, 5910 spectra shown in Fig. 4b.
Lodders K., Fegley B., 1998, The Planetary Scientist’s Companion. Oxford Table S3. Brucite1.0.txt is a tab-delimited text of the merged
Univ. Press, Oxford spectra shown in Fig. 2a, with fringes and artefacts removed, but
Malfait K., Waelkens C., Waters L., Vandenbussche B., Huygen E., limited to 1800 cm−1 .
de Graauw M. S., 1998, A&A, 332, L25 Table S4. Tremmerge.txt is a tab-delimited text of the merged
Mellini M., 1982, Am. Mineral., 67, 587 mid- and far-IR spectra shown in Fig. 2b, with fringes and artefacts
Mellini M., Fuchs Y., Lemaire C. L., Linares J., 2002, Euro. J. Mineral., 14, removed.
97 Table S5. Tremolitefullrang.txt is a tab-delimited text of the entire
Merlino S., 1980, Zeit. Kristallog., 151, 91
range of Figure 2b, formatted as above.
Mitra S. S., 1969, in Nudelman S., Mitra S. S., eds, Optical Properties of
Solids. Plenum Press, New York, pp. 333–452
Table S6. Lizardite0.50.txt is a tab-delimited text of the merged
Molster F. J. et al., 2001, A&A, 372, L165 mid- and far-IR spectra shown in Fig. 2c, with fringes and artefacts
Moore P. B., 1969, Am. Mineral., 54, 31 removed.
Mustard J. F., 1992, Am. Mineral., 77, 345 Table S7. Talc0.41.txt is a tab-delimited text of the merged mid-
Nuth J. A. III., Rietmeijer F. J. M., Hill H. G. M., 2002, Meteoritics Planet. and far-IR spectra shown in Fig. 2d, with fringes and artefacts re-
Sci., 37, 1579 moved.
Nyquist R. A., Kagel R. O., 1971, Infrared Spectra of Inorganic Compounds. Table S8. Saponite0.84.txt is a tab-delimited text of the merged
Acamemic Press, New York mid- and far-IR spectra shown for the Scottish sample in Fig. 2e,
Palme H., Fegley B., 1990, Earth Planet. Sci. Lett., 101, 180 with fringes and artefacts removed.
Paterson M. S., 1982, Bull. Mineral., 105, 20
Table S9. Montm0.75.txt is a tab-delimited text of the merged
Robert J. L., Ventura G. D., Hawthorne F. C., 1999, Micro. Microanalysis,
84, 86
mid- and far-IR spectra shown for the UK sample in Fig. 2f, with
Rossman G. R., 1988, Rev. Mineral., 18, 193 fringes and artefacts removed.
Rubin A. E., 1997, Meteor. Planet. Sci., 32, 231
Shepard C. U., 1848, Am. J. Sci., 6, 249
Shinoda K., Aikawa N., 1998, Phys. Chem. Mineral., 25, 197 This paper has been typeset from a TEX/LATEX file prepared by the author.


C 2006 The Authors. Journal compilation 
C 2006 RAS, MNRAS 367, 577–591