Available online at www.sciencedirect.

com

Geochimica et Cosmochimica Acta 75 (2011) 3011–3023

www.elsevier.com/locate/gca

Arsenic uptake by natural calcite: An XAS study

F. Bardelli a, M. Benvenuti b, P. Costagliola b,⇑, F. Di Benedetto b,c, P. Lattanzi d,
C. Meneghini e, M. Romanelli c, L. Valenzano f
a
Institut des Sciences de la Terre, Université Joseph Fourier, Maison des Géosciences, 1381 rue de la Piscine, 38400 Grenoble, France
b
Dipartimento di Scienze della Terra, Università di Firenze, via G. La Pira 4, 50121 Firenze, Italy
c
Dipartimento di Chimica, Università di Firenze, via della Lastruccia 3, 50019 Sesto Fiorentino, Italy
d
Dipartimento di Scienze della Terra, Università di Cagliari, via Trentino 51, 09127 Cagliari, Italy
e
Dipartimento di Fisica “E. Amaldi”, Università degli Studi Roma Tre, via della Vasca Navale 84, 00146 Roma, Italy
f
Michigan Technological University, Physics Department, 1400 Townsend Drive Houghton, MI 49931-1295, USA

Received 26 May 2010; accepted in revised form 23 February 2011; available online 6 March 2011

Abstract

As-bearing travertine rocks from Tuscany (Italy), where previous studies suggested the existence of a CO32 , AsO33
substitution in the calcite lattice, were investigated with X-ray Absorption Spectroscopy (XAS) at the As K-edge
(11,867 eV). In two of the studied samples, XANES indicates that As is in the 5+ oxidation state only, and EXAFS analysis
reveals a local environment typical of arsenate species. For these samples, the lack of detectable second shell signals suggests a
poorly ordered environment, possibly corresponding to an adsorption onto oxide and/or silicate phases. On the other hand, in
the third sample XANES reveals a mixed As oxidation state (III and V). This sample also presents evident next nearest neigh-
bor coordination shells, attributed to As–Ca and As–As contributions. The occurrence of next neighbor shells is evidence that
part of As is incorporated in an ordered lattice. Furthermore, the local structure revealed by EXAFS is compatible with As
incorporation in the calcite phase, as further supported by DFT simulations. The observation of next neighbors shells only in
the As(III)-rich sample suggests the substitution of the arsenite group in place of the carbonate one (CO32 , AsO33). The
conclusion of this work is that uptake of As by calcite is in general less favored than adsorption onto iron oxhydroxides, but
could become environmentally important wherever the latter phenomenon is hindered.
Ó 2011 Elsevier Ltd. All rights reserved.

1. INTRODUCTION 2002; Sherman and Randall, 2003; Dixit and Hering,

2003; Ouvrard et al., 2005). In particular, adsorption of
Arsenic contamination of groundwater and food repre- arsenic by Fe- and Mn-oxyhydroxide surfaces is considered
sents a serious environmental problem in many parts of among the most important and widespread mechanisms
the world (e.g., Matschullat, 2000; Smedley and Kinnin- controlling arsenic mobility. The destabilization of these
burg, 2002). A wide range of studies have recently been de- phases, or the desorption of As oxyanions from their sur-
voted to the mobility of arsenic in natural media, and much faces, is considered one of the pathways to the human food
attention has been paid to sorption–desorption and copre- chain (Nriagu, 1994; Tufano et al., 2008 and reference
cipitation reactions between arsenic and mineral surfaces, therein). Under some conditions, this mechanism represents
especially Fe- and Mn-oxyhydroxides (e.g., Lenoble et al., a transient immobilization for arsenic.
The possible role of calcite in sequestering As has been
⇑ Corresponding author. Tel.: +39 55 2757476; fax: +39 55 considered in a number of studies (Goldberg and Glaubig,
1988; Sadiq, 1997; Cheng et al., 1999; Le Guern et al., 2003;
2757455x284571.
E-mail addresses: pilario.costagliola@unifi.it (P. Costagliola), Roman-Ross et al., 2006; Di Benedetto et al., 2006; Alexan-
francesco.dibenedetto@unifi.it (F. Di Benedetto). dratos et al., 2007; Fernández-Martı́nez et al., 2008;

0016-7037/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.gca.2011.03.003
3012 F. Bardelli et al. / Geochimica et Cosmochimica Acta 75 (2011) 3011–3023

Sø et al., 2008; Yokoyama et al., 2009). The consequences possible replacement of carbonate groups in the calcite
of As uptake by calcite in natural systems are of consider- lattice.
able relevance because of the ubiquity of this mineral in
the Earth’s crust and its stability in a variety of geologic 2. EXPERIMENTAL
environments; therefore, calcite could represent an effective
agent for As immobilization. 2.1. Background information for natural samples
Most studies are focused on the possibility that calcite
may adsorb As-oxyanions onto its surface. In the last four An anomalous As-rich travertine sequence, located
years, however, it was demonstrated that As oxyanions may along the valley of the Pecora River (PRV, Southern Tus-
substitute for the carbonate group in the calcite structure cany, Italy) was sampled. For a detailed mineralogical
(Di Benedetto et al., 2006; Roman-Ross et al., 2006; Alex- and geological description, the reader is referred to Costa-
andratos et al., 2007). This replacement should be possible gliola et al. (2004, 2008, 2010), Di Benedetto et al. (2006),
in spite of the remarkable size and geometry differences be- and Benvenuti et al. (2009).
tween the carbonate group (C–O distance 1.3 Å, planar Travertines were formed in a lacustrine environment and
shape) and As oxyanions (As–O 1.8 Å in arsenite, and make up a sequence having a thickness up to 15–20 m of
1.7 Å in arsenate, both of pyramidal shape), because of carbonatic tufa of meteogenic travertine (sensu Pentecost,
the known flexibility of the calcite structure (e.g., Tang 1995). In the paleolake, where travertine accumulated, there
et al., 2007). There are, however, a number of significant was a concomitant, intermittent deposition of siliciclastic
gaps in the knowledge and understanding of the phenome- sediments. As a consequence, travertines are composed
non. Firstly, most of the studies refer to laboratory condi- mostly of calcite in micro-crystals forming a vacuolar tex-
tions, and the extension of their results to natural ture (Fig. 1), and by variable amounts of other minerals
environments is only inferred. Roman-Ross et al. (2006) (mainly silicates), including Fe-oxyhydroxides, conferring
indicate that As(III) oxyanions replace CO32 groups up to the bulk rock a brownish-yellow color. Non-carbonate
to concentrations on the order of 30 mM (2 g/kg) As, minerals (essentially quartz, mica, clay minerals, and Mn
whereas Alexandratos et al. (2007) provide a model in and Fe-oxyhydroxides) may be found as discrete thin layers
which the incorporation involves As(V)-oxyanions. In a re- (Fig. 1). In addition to Ca, travertine cation composition
cent study, Sø et al. (2008) cast some doubt on the results includes minor amounts of Si, Al, Mg and Fe (Di Benedetto
by Roman-Ross et al. (2006), concluding that arsenite up- et al., 2006). Bulk As-contents range between 127 and
take at the calcite surface can occur only at comparatively 257 mg/kg. Application of an ad hoc implemented sequen-
high pH and arsenite concentrations. The feasibility of tial extraction procedure demonstrated that a significant
the CO32 , AsO33 replacement mechanism in natural part (tens to hundreds of mg/kg) of this element is specifi-
calcite was first inferred by Di Benedetto et al. (2006) in a cally linked to the calcite fraction (Costagliola et al.,
study of Quaternary travertine deposits in Southern Tus- 2007). Arsenic may have been introduced in the paleolake
cany (Italy). These authors applied a spectroscopic tech- along with the siliciclastic sediments (Di Benedetto et al.,
nique, Electron Spin Echo Envelope Modulation 2006; Costagliola et al., 2008, 2010; Benvenuti et al., 2009).
(ESEEM) spectroscopy, whereby the presence of As is de-
tected from its nuclear modulation of the spectrum of the
paramagnetic ion Mn(II), which substitutes for Ca in the
calcite lattice. The estimated As–Mn distance is comparable
with the Ca–C distance in calcite. However, the employed
technique does not allow the estimation of the amount of
arsenic sequestered by calcite, nor determination of As oxi-
dation state. Thus, an estimate of the actual role played by
calcite in natural processes of As-sequestration awaits the
full assessment of the process through which As is included
in the calcite lattice.
X-ray Absorption Spectroscopy (XAS) is a well estab-
lished technique for structural investigation (Lee et al.,
1981) and chemical speciation (Wilke et al., 2001; Isaure
et al., 2002) and, with the development of beamlines able
to measure highly diluted elements (down to ppb) at high
spatial resolutions, it is gaining increasing importance in
the fields of environmental science and geochemistry (Fen-
Fig. 1. SEM image (backscattered electrons) of a travertine
ter et al., 2002; Alexandratos et al., 2007; Fernández-Martı́-
sample. It is evident the vacuolar structure of the rock (dark
nez et al., 2008; Chakraborty et al., 2010; Aurelio et al.,
holes). Calcite is in medium grey, whereas small crystals of Fe-
2010). oxyhydroxides are present at the top (white). The medium grey-
In this study, we used XAS to examine samples from the dark area at the bottom is presumably composed by a mixture of
same travertine sequence studied by Di Benedetto et al. Mn(Fe)-oxyhydroxides, calcite and silicates. The semi-quantitative
(2006), with the aim of investigating the oxidation state, EDS analysis of one of the latter regions (white rectangle) is
geometry, and local structure of As-oxyanions, and their provided in the table.
 Arsenic uptake by natural calcite: An XAS study 3013

In this study three different aliquots (F1b, F1bM and

F1bNM) of a travertine sample (F1b, studied by Di Bened-
etto et al., 2006) were examined. This sample was selected
since it had the highest As-content of the sequence
(257 mg/kg; see Table 1 of Di Benedetto et al., 2006), and
was already studied by these authors via EPR-ESE spec-
troscopy. The three different aliquots are the result of a sep-
aration of the mineral phases carried out by means of a
magnetic isodynamic Frantz separator (F1b: raw sample;
F1bM and F1bNM: “magnetic” and “non-magnetic” frac-
tions of F1b, respectively). We expected a significant sepa-
ration of magnetic Fe-oxyhydroxides from diamagnetic
minerals (including calcite); however, even if qualitative
X-ray Fluorescence (XRF) spectra acquired at 12 keV on
the Italian beamline at the European synchrotron (ESRF),
confirmed Fe to be enriched in the magnetic F1bM fraction
(Fig. 2), X-ray Diffraction (XRD) patterns of the bulk sam-
ple and of the two fractions did not show differences. Nev-
ertheless, it has to be noted that detection limit of XRD Fig. 2. Qualitative fluorescence spectrum of the F1bM and
technique (3–5 wt.%) and poor crystallinity of the unique F1bNM samples: Fe-content is observed to increase in the
magnetic fraction F1bM.

Table 1
Coordination numbers (CN), bond distances (R) and Debye–Waller factors (r2) of all reference compounds and travertine samples.
Coordination numbers of reference compounds were kept fixed to their crystallographic values.
As(III) samples + F1bNM Path CN R (Å) r2 (Å2)  103 S0 2
Na-arsenite As–O(1) 1 1.69(3) 1.9(9)a 1.0(2)
NaAsO2 As–O(2) 2 1.80(2) 5.5(9)
As(III) As–As 2 3.22(2)
Arsenic (III) oxide As–O(1) 3 1.80(1) 1.7(9) 0.8(1)
As2O3 As–O(2) 3 3.00(5) 4.7(9)
As(III) As–As(1) 3 3.25(1) 1.1(9)
As–As(2) 6 3.93(4) 12(1)
F1bNM As–O 3.3(5) 1.74(1) 6(1) 1b
As(V)–As(III) As–Ca(1) 1.5(7) 3.07(1) 4(2)a
As–Ca(2) 2.1(8) 3.36(1)
As–As 4(2) 3.93(2) 12(5)
As(V) samples Path CN R (Å) r2 (Å2)  103 S0 2
F1B As–O 4 1.68(2) 2 1.0(1)
As(V) .4(2)
F1BM As–O 4 1.68(3) 2.4(9) 1.0(2)
As(V)
Ferrihydrite As–O 4 1.69(2) 1.3(2) 1.0(1)
Fe(OH)3 As–As 1.1(5) 3.34(3) 2.6(6)a
As(V) As–Fe 1.7(9) 3.26(2)
Scorodite As–O 4 1.69(1) 1.1(8) 0.9(1)
Fe(AsO4)(H2O)2 As–Fe 2 3.33(3) 3.8(9)
As(V) As–Fe 2 3.51(6) 8.5(9)
Na-arsenate As–O(1) 4 1.69(1) 0.8(1) 0.8(1)
NaH2(AsO4)(H2O) As–O(2) 1 3.21(4) 1.0(1)
As(V) As–Na(1) 2 3.69(4) 4.9(6)
As–Na(2) 4 3.93(4) 5.0(7)
Arsenic (V) oxide As–O(1) 2 1.67(3) 2.7(2)a 0.9(1)
As2O5 As–O(2) 3 1.80(4)
As(V) As–As(1) 2 3.10(3) 7.2(6)
As–As(2) 2 3.23(4) 7.9(6)
Ca-arsenate As–O 4 1.68(1) 2.0(4) 1.0(1)
Ca3(AsO4)2 As–Ca(1) 1 3.27(6) 6(1)
As(V) As–Ca(2) 3 3.56(8) 9(1)
As–Ca(3) 3 3.71(6) 9(2)
a
The same Debye–Waller factor parameter was used for these paths.
b
S02 was arbitrarily fixed to 1 to avoid strong correlation with CN. Numbers within parenthesis represent the uncertainty on the last digit.
3014 F. Bardelli et al. / Geochimica et Cosmochimica Acta 75 (2011) 3011–3023

magnetically active phases of the investigated sample (i.e., surface in the actual region. The default maximum value for
Fe-, Mn-oxy-hydroxides) could have prevented their minimization is 0.5. At each optimization cycle, the step is
identification. computed by means of a Newton-like scheme, while the
length is determined by linear minimization along an
2.2. SEM-EDS extrapolated quadratic polynomial.
Since both total energy and gradients are affected by the
Morphological, mineralogical and semi-quantitative integrals classification (truncation of infinite Coulomb and
chemical microanalyses were performed on graphite-coated exchange series), a single-point energy calculation needs al-
polished sections using a Philips SEM515 Scanning Elec- ways to be run with the final structure, and integrals have to
tron Microscope (SEM) equipped with an AnalySis-ADDA be re-classified according to the new final geometry to cal-
image grabber and an Edax Falcon X-ray Energy Disper- culate correct total energy and gradients. If during this final
sive Microanalytical System. run procedure the convergence test on the forces is not sat-
isfied, optimization is restarted, keeping the integrals classi-
2.3. DFT simulations fication based on the new geometry. For this reason, the
FINALRUN option has to be used. Among the different
Numerical simulations of the arsenite inclusion in the levels of choice, we chose the tighter, which computes sin-
calcite structure were carried out using periodic Density gle-point energy and gradient followed by a new optimiza-
Functional Theory (DFT) calculations employing the hy- tion process starting from the final geometry of the previous
brid B3LYP functional (Becke, 1993; Lee et al. 1988). All one (used to classify the integrals); if the convergence test is
calculations were performed with the CRYSTAL06 code not satisfied the procedure iterates until full stable optimi-
(Dovesi et al., 2006). zation is reached.
The CRYSTAL package performs ab initio calculations The experimental unit cell of pure calcite structure
of the ground state energy, energy gradient, electronic wave (Markgraf and Reeder, 1985) was used as starting point
function, and properties of periodic systems at both Har- for the model. A single unit cell was used for the calcula-
tree–Fock and Kohn–Sham level of theory. The fundamen- tions where one of the CO32 ions, placed in the bulk of
tal approximation made is the expansion of the single the cell, was replaced with an HAsO32 unit to avoid any
particle wave functions (crystalline orbitals) as a linear surface effect in the volume expansion. Despite the apparent
combination of Bloch functions defined in terms of local simplicity of the model, we believe that this approach is suf-
functions (atomic orbitals). The local functions are, in turn, ficient to provide a comparison with the local structure
linear combinations of Gaussian type functions whose experimentally obtained by means of the XAS technique.
exponents and coefficients are defined by input. Functions Nevertheless, we want to stress here that the model used
of symmetry s, p, d and f can be adopted, and also available is not suitable for a comparison with the long range struc-
are sp shells. ture. Such a comparison would have been possible only
For calcium, oxygen, and carbon atoms, all-electron using a large supercell, in order to let the structure around
Gaussian-type basis sets were used, as reported in Valenz- the inserted As defect relax.
ano et al. (2006); for arsenic, the Durand and Barthelat The presence of the hydrogen atom was necessary to en-
effective core pseudo-potential was used, as reported in sure the charge neutrality. The geometrical optimizations
Causà et al. (1991). For numerical integration of the ex- were made at zero temperature and removing symmetry
change–correlation term, a (75,974)p pruned grid was constraints (space group: P1).
adopted, and the condition for the SCF convergence was
set to 107 Hartrees during geometry optimization. A full 2.4. XAS setup
relaxation of the structures (both lattice parameters and
atomic positions) was performed. For geometry optimiza- X-ray Absorption Spectroscopy (XAS) measurements at
tion, the default convergence criteria were used. In particu- the As K-edge (11,867 eV) were performed at the GILDA
lar, an estimated model Hessian is adopted in the program (BM-08) beamline (Pascarelli et al., 1996) at the European
and the matrix is stored at each optimization step, allowing Synchrotron Radiation Facility (ESRF, Grenoble –
the possibility, for example, to restart the calculation at an- France). The main beamline optical features include a fixed
other level of theory. Different Hessian updating schemes exit monochromator with a pair of Si [3 1 1] crystals, and a
are available for minimization; in the reported calculations, pair of Pd coated mirrors for efficient harmonic rejection
the so called BFGS Hessian update scheme by Broyden– and (vertical) focusing of the X-ray beam. A dynamic sag-
Fletcher–Goldfarb–Shanno (Broyden, 1970a,b; Fletcher, ittal (horizontal) focusing of the X-ray beam is realized by
1970; Goldfarb, 1970; Shanno, 1970) was employed. This bending the second monochromator crystal. This provides
is the default method used in CRYSTAL to update the Hes- a sub-millimetric (250 lm horizontal, 150 lm vertical)
sian matrix. X-ray spot on the sample, with a photon flux at 12 keV
The convergence criteria on the RMS of the gradient of about 109 ph s1. Considering the monochromator crys-
and the displacement were set to 0.0003 and 0.0012, respec- tals and the vertical aperture of slits at the source (1 mm),
tively. Nevertheless, a more sophisticated and accurate the energy resolution was calculated to be about 0.5 eV at
technique to control the step size was adopted: the trust ra- the As K edge. The energy sampling interval in the higher
dius region scheme that limits the step length of the displace- resolution near edge region (11,840–11,920 eV) was
ment at each cycle, according to the quadratic form of the 0.2 eV. In order to reduce the dumping of the signal due
 Arsenic uptake by natural calcite: An XAS study 3015

to the thermal contribution and to prevent possible beam- 2002) as input. Theoretical amplitude and phase functions
induced redox reactions, all samples were measured at li- were generated using the FEFF8 code (Ankudinov et al.,
quid nitrogen temperature (77 K). 1998). The FITEXA code (Monesi et al., 2005), used for
Reference compounds included: pure synthetic refinement of the EXAFS signal, exploits the MINUIT rou-
arsenic(III) and arsenic(V) oxides, As2O3 and As2O5 (Alfa tines of CERN libraries (James and Winkler, 2004) for least
Aesar), calcium and sodium arsenate, Ca3(AsO4)2 square minimization. Errors on structural parameters were
and Na2HAsO47H2O (Alfa Aesar), sodium arsenite, calculated using the MINOS subroutine from the MINUIT
(NaAsO2)x, and natural scorodite, FeAsO42H2O, kindly package, and takes into account the correlation between
provided by the Museo di Storia Naturale (Mineralogy parameters.
Section) of the Università di Firenze. In addition, an As- The number of floating parameters during minimization
bearing Fe-oxyhydroxide was synthesised in the laboratory never exceeded 12, which is lower than the maximum num-
by precipitation from a As-, Fe-acidic solution through pH ber of independent points (Nind = (2  Dk  DR/p + 1 = 17,
increase (Jia et al., 2007). Phase and crystallinity check of Dk and DR being the k (reciprocal space) and R (direct
natural scorodite and synthetic Fe-oxyhydroxide was per- space) ranges of useful data). For a maximum useful exten-
formed by XRD. Both samples were found monophasic, sion of the spectra of kmax = 14 Å1, the theoretical spatial
and the Fe-oxyhydroxide revealed the characteristic two- resolution of our data is 0.02 Å (dR  p/(2  kmax)). Actual
line structure (Schwertmann et al., 2004). resolution is lower due to the noise contribution to the spec-
All reference samples were prepared for XAS measure- tra, and was estimated to be about 0.05 Å.
ments as follows: we calculated the amount suitable to For XANES (X-ray Absoption Near Edge Structure)
reach an edge step (Dlt) close to 1 while keeping the total semi-quantitative analysis, all spectra were normalized to
absorption (lt) in the range 1–2.5; this amount was mixed a region far from the edge step (>200 eV), and the energy
and homogenized with cellulose in an agate mortar. The scale was carefully calibrated using the reference absorption
mixture was then pressed at 10 bars into 25 mm diameter spectra from As2O3 simultaneously acquired at each scan.
pellets that were then mounted on a copper sample holder, Structural information (number and type of neighboring
placed in a vacuum chamber (105 mbars), and cooled atoms, bond distances and structural disorder) around the
down to liquid nitrogen temperature (77 K). absorber atom were extracted from the analysis of the ex-
All reference compounds spectra were collected in trans- tended region of the absorption spectra (Extended X-ray
mission geometry, i.e., measuring the intensity of the X-ray Absorption Fine Structure: EXAFS).
beam upstream (I0) and downstream (It) of the sample by Data refinements were performed in the back-trans-
means of ionization chambers filled with suitable gas and formed space, applying the standard XAFS formula (Rehr
pressure (20% absorption in the upstream chamber, 80% and Albers, 2000). For direct comparison all the spectra
in the downstream one). A third ionization chamber was (travertine and references samples) were treated and fitted
placed after a second experimental chamber containing an using the same parameters: Fourier transform (FT) range
As2O3 standard which spectrum was acquired simulta- 1–14 Å1 (13.3 Å for F1bNM sample) and weight 3,
neously with each sample spectra and used for accurate en- back-Fourier transform (BFT) range 1–3.5 Å (uncorrected
ergy calibration. Depending on the quality of the spectra, a for phase shift). The fitting range was 4–14 Å1 (in the
minimum of two scans were collected, interpolated and back-transformed space) and the fitting weight 3. Such a
averaged to increase the signal-to-noise ratio. large starting point (4 Å1) for minimizations was chosen
Powdered travertine samples were either pressed in pel- to avoid strong near edge multiple scattering effects, and be-
lets, or diluted in ethanol and deposited onto microporous cause theoretical amplitude and back-scattering functions
membranes. Because of the low arsenic content (200 mg/ calculated by ab initio multiple scattering based codes, such
kg), travertine samples were measured in fluorescence mode as FEFF, are less accurate in the low k-region.
by means of a 13-elements solid state (high purity Germa- For all samples, only single-scattering paths were found
nium) detector (ORTEC). A minimum of three spectra were relevant to fit the spectra. Experimental energy shift (DE)
collected for each travertine sample, and averaged to im- was optimized through a first shell analysis and kept fixed
prove the statistics of the data. Higher integration times during final full spectra refinements. Coordination numbers
(up to 30s) per energy point were preferred to a larger num- (CN) of all reference compounds were kept fixed to their
ber of scans, because of the long dead-time of the acquisi- crystallographic values in order to avoid correlation with
tion system and the limited beamtime available. the amplitude reduction (S02) and Debye–Waller (r2) fac-
Standard procedures (Rehr and Albers, 2000) were fol- tors. More details on the strategy followed for refinements
lowed to extract the structural EXAFS (Extended X-ray of travertine samples are given in the EXAFS section.
Absorption Fine Structure) signal (k  v(k)): pre-edge back-
ground removal, spline modeling of bare atomic back- 3. RESULTS
ground, edge step normalization using a far above the
edge region, and energy calibration. 3.1. XANES
Model atomic clusters centered on the absorber atom
were obtained by ATOMS (Ravel, 2001) using crystallo- XANES absorption spectra of all measured samples
graphic structures reported in the literature (Menary, (reference compounds and travertine samples) are reported
1958; Gopal and Calvo, 1971; Ferraris et al., 1974; Kitaha- in the left panel of Fig. 3. Edge crests (i.e., the main peak
ma et al., 1975; Pertlik, 1978; Jansen, 1979; Jansen et al., after the absorption edge) and absorption edges energies
3016 F. Bardelli et al. / Geochimica et Cosmochimica Acta 75 (2011) 3011–3023

Fig. 3. Calibrated and normalized XANES spectra of travertine samples (F1b, F1bM and F1bNM) and of all measured reference compounds
(left panel). Note the similarity of the position of the main edge crest of F1b and F1bM and the all the other As(V) compounds (11,873 eV),
while F1bNM have a double edge crest which positions match As(V) and As(III) edge crests (11,870 and 11,873 eV). Right panel: first
derivatives of the absorption spectra of all spectra reported in the left panel. The peaks in the derivatives indicate the position of the main
inflection point, which, by convention, is considered to be the absorption edge energy (E0). It can be seen that all As(III) compounds have an
absorption edge at 11,868.6 eV, and all As(V) compounds at 11,871.6 eV (1.6 and 5.6 eV above the elemental As edge, respectively). Again
F1bNM show two peaks at energies corresponding to As(III), As(V) edges (11,868.6 and 11,872 eV). Vertical dashed lines are a guide to the
eye to highline As(V) and As(III) edge crests. Spectra are vertically shifted for ease of view.

(position of the first peak of the derivative of the absorption

spectra, Fig. 3, right panel) are separated in two groups:
As(V) reference compounds and F1b, F1bM travertine
samples (average edge crest, 11,873 eV, absorption edge
11,871 eV), and As(III) reference compounds (average edge
crest, 11,870 eV, absorption edge 11,869 eV). These values
are in agreement with the literature (Jönsson and Sherman,
2008), reporting energies about 1.5 and 5.5 eV above the
elemental As K-edge (11,867 eV) for As(III) and As(V) spe-
cies, respectively.
A double edge crest occurs in F1bNM’s spectrum, at
energies corresponding to those of As(III) and As(V) refer-
ence compounds (Fig. 3, left panel); this appears in two
peaks in the first derivative at energies corresponding to
As(III) and As(V) adsorption edges (Fig. 3, right panel).
This is clear evidence of a mixed As oxidation state (III/
V) in F1bNM.
The XANES features of F1b and F1bM travertine sam- Fig. 4. Linear square fit of F1bNM (solid line) using a linear
ples are similar to those of both scorodite and arsenate combination of the As(V)- and As(III)-oxide spectra. The exper-
imental F1bNM spectrum is reported in points + line style together
spectra, showing a smooth trend after the edge crest,
with the relative contributions of As(V)-(dashed line) and As(III)-
whereas Ca- and Na-arsenates have slightly more struc-
oxides (solid line).
tured spectra (Fig. 3), representing poorer matches to F1b
and F1bM.
F1bNM more structured spectrum originates from the tor of the linear combination (x) was optimized using the
superposition of signals arising from As(III) and As(V) lo- XPSFIT code (Bardelli et al., 2009) and indicates a ratio
cal environments. This is supported by the fact that of As(III) to As(V) of roughly 3:1 (Fig. 4). It is worth not-
F1bNM spectrum can be reproduced using a normalized ing that this procedure was only used to estimate the
linear combination of As2O5 and As2O3 spectra lsam- As(III)/As(V) ratio in F1bNM and does not imply that
ple = x  l(As2O5) + (1  x)  l(As2O3). The weighting fac- F1bNM is a mixture of the arsenic oxides.
 Arsenic uptake by natural calcite: An XAS study 3017

3.2. EXAFS (around 1.4 Å, uncorrected for phase shift) is always related
to the first oxygen coordination shell (As–O paths), while
Back-Fourier transformed (BFT) EXAFS oscillations of peaks at higher R values correspond to As–Fe, As–Ca,
As(III) standards, plus F1bNM travertine sample, are re- As–Na, and/or As–As contributions, depending on the
ported in the left panel of Fig. 5 (upper panel), together with compound considered.
the corresponding fit curves, while BFT and fit curves of In the following we describe the refined local structure
As(V) standards plus F1b and F1bM travertine samples around the arsenic atom of F1b and F1bM together, and
are shown in the lower left panel of the same figure. Fourier consider the F1bNM sample in a separate paragraph.
transform moduli (|FT|) of the EXAFS oscillations of the
standards and samples reported in the left panels are shown 3.2.1. F1b–F1bM
in the right panels (As(III) plus F1bNM, upper panel; As(V) As can be seen from Fig. 5 (lower left panel), F1b and
plus F1b and F1bM, lower panel) together with the corre- F1bM have an almost single-frequency oscillation reflecting
sponding fit curves. |FT|s represent a radial distribution in a single well defined peak in the real (R) space (Fig. 5,
function of distances centered on the absorber atom. Each lower right panel). This means that virtually no contribu-
peak corresponds to an average length (provided a correc- tions from coordination shells higher than the first are
tion for a phase function) arising from a single scattering detectable, suggesting a poor crystalline structure of adsor-
path or, due to the limited resolution, is the convolution bate. Structural refinements reveal that F1b and F1bM
of peaks corresponding to different paths. The first peak have nearly identical local structure: four As–O bonds at

Fig. 5. Left panels: back-transformed (1  3.5 Å) EXAFS signals (points) of the As(III) plus F1bNM (upper panels), and As(V) standards
plus F1bNM (lower panels), and corresponding fit curves (solid grey line). Fits were performed in the k-space (4  14 Å1). Right panel: FT
(4  14 Å1) of the signals and fit curves reported in the right panel (experimental, points + line; fit curves, solid gray line). Values in the real
space (R) are not corrected for the phase shift. Note the difference between less structured spectra of As(V) compounds, which reflects in low
or negligible peaks at R > 2 Å in the FTs (contributions from shells higher than the first), and more structured spectra of As(III) compounds
and F1bNM, which, conversely, have relevant higher shell contributions. Spectra are vertically shifted for ease of view. (For interpretation of
the references to colour in this figure legend, the reader is referred to the web version of this article.)
3018 F. Bardelli et al. / Geochimica et Cosmochimica Acta 75 (2011) 3011–3023

a distance of about 1.68 Å. Coordination numbers were set 3.2.2. F1bNM
to four in final refinements to avoid the strong correlation Since the EXAFS signal in F1bNM originates from an
with the EXAFS amplitude reduction factor S02. This admixture of As sites (arsenite and arsenate), having differ-
assumption is supported by XANES results, revealing As ent local structures, it was not possible to fix CNs to theoret-
in the 5+ oxidation state (arsenate groups, AsO43, have ical values as done for F1b and F1bM. The lack of a reliable
tetrahedral structure, Fig. 6c). Further support to this mod- model to fit such an unknown structure led us to arbitrarily
el comes from the facts that: setting the amplitude reduction factor (S02) to 1, to avoid the
strong (anti)correlation with CNs. Considering that S02 lies
(1) F1b and F1bM XANES spectra are similar to scoro- in the range 0.8–1.0 for all structures refined in this work
dite, ferrihydrite, Na- and Ca-arsenate reference (Table 1), we estimated that this assumption introduces an
compounds spectra, where As is present as arsenate error on the refined CN not larger than 20%, which is
(see Fig. 3); the usual error attributed to CNs in EXAFS. Conversely,
(2) low values of the Debye–Waller factors ( 2  103 there was no way to avoid correlation between the Debye–
Å2), suggest equal As–O bond distances and a well Waller factors (r2) and CNs, a common problem in EXAFS
ordered first coordination shell, as occurring for the refinements (Lee et al., 1981). To reduce the correlation, the
arsenate group (see Table 1); same r2 parameters were used for the split As–Ca distances.
(3) fixing the coordination number to 4, a stable mini- Furthermore, correlations between parameters are taken
mum for the amplitude reduction factor (S02) is into account by the minimization routine increasing the er-
found in the range of physically meaningful values ror bars according to the correlation matrix.
(0.7  1.0), close to that of all other As(V) reference Refined structural parameters of F1bNM reveal a first
compounds (see Table 1). oxygen coordination shell with CN lying between 3 and 4

Fig. 6. (a) Representation of the incorporation of an arsenite unit HAsO32 in the calcite unit cell. Simulation leads to As atoms displaced out
of the arsenite oxygen’s plane (along the c-axis) while CO32 groups remain planar. (b) Molecular geometry of arsenite (AsO33) and (c)
arsenate (AsO43) oxyanions in solution (Ramı́rez-Solı́s et al., 2004). Local structure around the As(III) atom as resulting from EXAFS
refinements projected along the y direction (d) and along the z direction (e); the arrows labeled d1, d2, d3 and d4 correspond to the As–O,
As–Ca1, As–Ca2, and As–As distances, respectively.
 Arsenic uptake by natural calcite: An XAS study 3019

(3.3), and an As–O bond length of 1.74 Å (Table 1). The 4. DISCUSSION
quite large Debye–Waller value (6103 Å2) suggests that
this shell is a convolution of As–O sub-shells split at differ- The complete mineral assemblage of the travertine se-
ent distances. This is consistent with the two different local quence, to which the F1b sample belongs, was studied in
environments corresponding to the two As oxidation states detail by Di Benedetto et al. (2006), by complementary
suggested by XANES analysis. analyses including optical microscopy, X-ray diffraction,
In contrast to F1b and F1bM, the mixed oxidation state and SEM-EDS. Beside calcite and variable amounts of
sample (F1bNM) spectrum has an intense signal from coor- quartz, these authors found traces of phyllosilicates, Ca-sul-
dination shells higher than the first. According to the bulk phates, iron-hydroxides, and Ti-oxides. SEM-EDS semi-
chemical analyses (Di Benedetto et al. 2006), which detected quantitative analyses pointed out the presence of small
Ca, Al, Si, Mg, As and Fe as major elements present in iron-oxide fragments with high arsenic contents (up to
these same samples, different As-backscatterer couples about 6% by weight). The chemical composition of traver-
(As–Fe, As–(Si/Al/Mg), As–Ca, and As–As) were used in tine (major CaO, with variable amounts of SiO2 and Al2O3;
different combinations to fit higher shell contributions. Di Benedetto et al., 2006) is consistent with this mineral
Mg, Al and Si backscatterers, which are hardly distinguish- assemblage. Travertine is rather heterogeneous at the scale
able by EXAFS, were ruled out after first trials. The most of the single outcrop. The large XAS spectral difference be-
stable and reliable configuration resulted the one with tween F1b–F1bM, and F1bNM, sub-samples of the same
a split As–Ca coordination shell (RAs–Ca(1) = 3.07, bulk sample, proves this high heterogeneity also at the scale
CN = 1.5, and CN = 2.1, RAs–Ca(2) = 3.36 Å), and a longer of the single sample.
contribution from an As–As coordination at RAs–As = Samples dominated by As(V) (F1b and F1bM) have a
3.93 Å (CN = 4), Table 1. A sketch of the local environ- single oxygen first coordination shell, with four almost
ment of As refined by EXAFS analysis is reported in equivalent As–O bonds at 1.68 Å. Next neighbor features,
Fig. 6d and e. if present, are negligibly weak. This finding suggests that
As, when occurring as arsenate, is not incorporated in
3.3. DFT results any crystalline structure. In particular, the similarity of
F1b and F1bM spectra to that of As in ferrihydrite suggests
DFT simulations on the calcite unit cell with arsenite that As(V) could be adsorbed on the surface of poorly crys-
replacing carbonate reveal a distorted AsO3 pyramid with talline iron-oxyhydroxides, which are well known as strong
three different As–O distances (Table 2). The As–O1 bond adsorbers for As(V) oxyanions (Saalfield and Bostick, 2009,
(1.845 Å) is considerably longer than the As–O2 and As– and reference therein). Refined structural parameters of
O3 bonds (1.738 and 1.733 Å, respectively). In average F1b and F1bM, in particular the observed low Debye–Wal-
the As–O distance is about 1.772 Å, which is longer than ler factor, are evidence for equally compensated charge,
the value obtained by EXAFS (about 1.74 Å). Second shell suggesting the absorption of protonated As oxyanions
features reveal two different As–Ca distances at 3.042 and rather than (AsO4)3 (Sherman and Randall, 2003).
3.316 Å, which are very close to the same distances mea- In the case of F1bNM sample, the first oxygen shell can
sured by EXAFS (3.07 ± 0.01 and 3.36 ± 0.01), (Table 1). be interpreted as a convolution of As–O sub-shells deriving
The calcite cell parameters calculated by DFT differ from different local environments for As(V) and As(III).
from those of the pure calcite, the largest difference being The first shell CN (3.3) is in agreement with the value aris-
for the length of c: 18.239 Å (DFT) versus 17.061 Å ing from the average of arsenite and arsenate CNs weighted
(Markgraf and Reeder, 1985). Within the limits of validity for the As(III)/As(V) ratio (3:1) determined by linear
of the chosen model, this large difference suggests that square fitting of XANES spectra discussed above. The va-
incorporation of the pyramidal AsO3 group in place of lue for the As–O bond length (1.74 Å) is also in accordance
the triangular CO3 has a strong effect, especially along the with an admixture of different As local structures, being
c axis. intermediate (if weighted for the arsenite/arsenate 3:1 ratio)

Table 2
Comparison between calculated (DFT), experimental (EXAFS), and crystallographic (ICSD) structural parameters (atomic distances and
lattice cell parameters). EXAFS values refer to the F1bNM sample, DFT values to a calcite cell with a single HAsO32 unit in place of a
CO32 one (see text), and crystallographic values refer to the pure calcite cell (Markgraf and Reeder, 1985). The numbers within parenthesis
represent the uncertainty on the last digit.
Atomic distances (Å) Lattice parameters (Å)
Path DFT EXAFS Calcite cell + HAsO32 (DFT) Crystallographic values (ICSD)
As–O1 1.845 }1.772
As–O2 1.738 1.74(1) a = 4.975 a = 4.988(1)
As–O3 1.733 b = 5.027 b = 4.988(1)
As–Ca1 3.042 3.07(1) c = 18.239 c = 17.061(1)
As–Ca2 3.316 3.36(1)
C–O1 1.287 – a = 87.3° a = 90°
C–O2 1.288 – b = 90.0° b = 90°
C–O3 1.286 – c = 118.9° c = 120°
3020 F. Bardelli et al. / Geochimica et Cosmochimica Acta 75 (2011) 3011–3023

between the As–O distance in the arsenate group (1.68 Å in [0 0 0 1] direction out of the oxygen plane (Fig. 6a) (in this
aqueous solution, Ramı́rez-Solı́s et al., 2004; Fig. 6c) and way, As reaches a favorable pyramidal configuration, min-
the average value obtained from simulations on the arsenite imizing the steric effects of the lone pair). The value ob-
group substituted in the calcite lattice (1.77 Å, see Table 2). tained from DFT simulations for the As displacement
The occurrence of higher shell contributions in F1bNM along the c-axis (0.80 Å) is in good agreement with the va-
suggests that As is here hosted in a more ordered environ- lue measured by Cheng et al. (1999) in arsenite groups ad-
ment with respect to F1b and F1bM samples. It is interest- sorbed onto calcite surface. On the other hand, the
ing to note that in F1b and F1bM, characterized by increment in the value of the c lattice parameter reported
negligible contributions from next neighbor coordination by DFT simulations (Table 2) is highly overestimated. This
shells, arsenic occurs as As(V) only. Conversely in is due to the fact that simulations were performed substitut-
F1bNM, where arsenic is mainly As(III), next neighbor ing one HAsO32 group for one CO3 in a single calcite cell,
contributions are more intense and evident (peak in the resulting in a As concentration much higher than the real
range 2.5–3.5 Å in the Fourier transform). This fact would one, and therefore, in an unrealistic elongation of the c lat-
suggest that the structural contributions due to the next tice parameter. We do expect that in As-bearing calcites
neighbor coordination shells mainly originate from the there is a slight, but measurable, effect of elongation of
As(III) species. The total coordination number of the As– the c parameter. In natural calcites, evidence of this effect
Ca shells, being the sum of coordination numbers of As– could be complicated by other lattice effects due to cation
Ca sub-shells at 3.07 and 3.36 Å, is CN 4(1). This value impurities (such as Mg, Fe, Mn).
is lower than that expected for As fully substituting for C Although As5+O4 , CO3 substitution was claimed to
in the calcite lattice (CN = 6). However, it must be stressed be observed in synthetic calcite samples (Alexandratos
here that only a fraction of As is incorporated into the cal- et al., 2007), the occurrence of the arsenate anion in the
cite lattice, i.e., As(III), while As(V), which is probably ad- calcite lattice is unlikely, or in any case minor. The slightly
sorbed on host structures (Fe/Mn-oxyhydroxides or calcite larger size of the tetrahedral As(V)O4 groups, with respect
itself), has negligible second shell contributions (as in F1b to the pyramidal As(III)O3 group, would favor the
and F1bM). Since the amplitude of the EXAFS signal is isomorphic As(III)O3 , CO3 substitution instead to the
normalized to the total As amount in the sample, the appar- As(V)O3 , CO3 one. Accordingly, our findings suggest a
ent coordination numbers of the two As species are scenario in which As(III) mainly substitutes C in the CO3
weighted by the As(III) and As(V) fractions. Therefore, group, while As(V) is present as an adsorbed oxyanion
since the As(III)/As(V) ratio is close to 3:1, the fraction onto Fe-oxyhydroxides.
of As(III) is about 3/4 and the expected As–Ca CN is It is noteworthy to mention that Alexandratos et al.
6  3/4 = 4.5, which is higher but consistent with our finding (2007), who claimed that arsenate group (AsO43) replaces
within error. Concerning the As–As contribution around carbonate in calcite, found two distinct As–Ca distances
3.9 Å, its CN is expected to depend on the dilution level (3.4 and 3.6 Å), which are longer than those found in
and distribution of As ions in the host structure: random this study. This finding strengthens the idea that we are
distribution and low concentrations of As leads to a reduc- dealing with a different mechanism of As uptake when com-
tion of the As–As CN. However, the observed CN is pared to Alexandratos et al. (2007).
around 4, which corresponds to an actual CN of about The hypothesis that As is bound in the calcite lattice is
5.3 if we take into account the As(III) fraction as done also supported by the conclusion of the ESEEM study by
above (4  4/3 = 5.3), that is definitively too large for ran- Di Benedetto et al. (2006). The As–Ca distance suggested
domly distributed/highly diluted As(III). The observed by these authors, 3.2 Å, was inferred by the distance be-
CN, rescaled by the fraction of As species which we believe tween As and Mn substituting for Ca in the calcite lattice,
substituted in the calcite lattice (i.e., As(III)), is close to the assuming a similar deformation of the calcite lattice in the
maximum number of As nearest neighbors in a fully As- case of co-uptake by As and Mn. The distance value pro-
substituted calcite structure (CN = 6). This finding support posed by Di Benedetto et al. (2006) is in full agreement with
the hypothesis of an As(III) clustering in the calcite struc- the average of the distances observed by EXAFS, 3.07–
ture, in accordance with spectroscopic results provided by 3.36 Å (notice that modeling of ESEEM data provides only
Romanelli et al. (2008). an average of the possible anisotropic location of interact-
The occurrence of a strong signal from As–Ca (at 3.07 ing nuclei in the shells around the paramagnetic center).
and 3.36 Å) and As–As (at 3.93 Å) coordination shells is The presence of arsenite groups in the calcite structure
a strong evidence that As(III) is bound to the calcite lattice. necessarily implies that there was an interaction between
Support to this hypothesis comes from the comparison be- arsenite and the calcite surface during the precipitation of
tween the structural information obtained by EXAFS anal- this mineral in the PRV travertine. Sø et al. (2008), in con-
ysis and the DFT simulations performed replacing a trast with the experimental results of Roman-Ross et al.
carbonate group with an arsenite one. The experimentally (2006), contend that this interaction is negligible unless at
observed coordination shells (As–O, As–Ca1 and As–Ca2) pH (9.5) and As concentrations (>50 lM) that are unli-
are reproduced with an excellent agreement by DFT simu- kely to occur in natural environments. We are not able to
lations (if we consider the discussion done above for the solve the contradiction between the laboratory data by Ro-
As–O distance). In particular, DFT simulations also con- man-Ross et al. (2006) and Sø et al. (2008), although we no-
firm the splitting of the As–Ca distances, which is a conse- tice that the first is a co-precipitation experiment and the
quence of the displacement of the As atom along the second an adsorption one. Moreover, in natural environ-
 Arsenic uptake by natural calcite: An XAS study 3021

ments additional factors can occur (e.g., mediation by Proposals for access to high sensitivity microbeam facilities
microorganisms, Oremland and Stolz, 2003; Oremland are under consideration.
et al., 2005 and reference therein), inducing a different
behavior with respect to laboratory conditions. We notice, 5. CONCLUSIONS
for example, that point of zero charge of calcite surface oc-
curs at pH values in the range 9–10 (Sadiq, 1997), meaning A characterization of the morphology and composition,
that the surface of calcite is positively charged at pH values oxidation state, short and long range lattice structure of As
lower than this threshold. It is expected that the OH con- in natural travertine samples was carried out combining
centration near to the calcite surface is higher than in the SEM-EDS, X-ray Absorption Spectroscopy (XAS), X-ray
bulk solution when the net surface charge of this mineral Diffraction (XRD), and theoretical ab initio simulations.
is positive. In fresh water, the counter-ions (mainly OH) SEM-EDS revealed a heterogeneous composition of the
layer thickness next to the calcite surface can extend up samples, which is confirmed by structural probe techniques:
to 100 Å in the solution (Appelo and Postma, 2007). in two samples, the XANES part of the spectra indicates
Some As reducing bacteria (driving the As(V)–As(III) reac- As(V) in a poorly ordered environment; in the third sample,
tion) are thought to be insensitive to As-rich conditions XANES reveals a coexistence of As(III) and As(V), roughly
only in high pH environments, where As(III) species are in a 3:1 ratio. This ratio reflects the relative abundance
charged and thus cannot enter the cell (Oremland and of As incorporated into calcite and onto Fe/Mn
Stolz, 2003). In fact, optimal growth conditions for As oxyhydroxides.
reducing bacteria are matched at pH 9.5 in continental EXAFS refinements suggest As(III) to occur in a local
lakes (Oremland et al., 2005). The local high-pH environ- environment compatible with the structure derived by the
ment next to positively charged mineral surface may led substitution of arsenite for carbonate in calcite lattice, as
to the dissociation of arsenious acid and its migration to- previously suggested by Di Benedetto et al. (2006) on the
ward the positive surface causing adsorption and coprecip- basis of an EPR-ESE study on same samples. EXAFS data
itation observed, under laboratory conditions, by Roman- are in excellent agreement with structural DFT simulations
Ross et al., 2006. In systems characterized by diffusive ion performed substituting a CO32 unit with a HAsO32
transport and by high specific surface of the precipitating group in the calcite cell. EXAFS analysis also suggests
mineral, such as in a lake where calcite precipitates in mi- the possible clustering of As anionic groups in limited vol-
cro-crystals, these processes are feasibly maximized. How- umes in calcite.
ever, it is also well known that precipitation rate can On the contrary, As(V) seems to be present in a poorly
strongly affect the incorporation of trace elements into min- crystalline phase, probably adsorbed on Fe-oxyhydroxides.
erals: for instance, high As(V) levels in garnet were ascribed Recent laboratory studies gave contrasting indications
to rapid precipitation (Charnock et al. 2007). Therefore, in on the interaction of arsenite species with calcite; the results
principle, both mechanisms have to be considered concur- of our study suggest that in natural system this interaction
ring in calcite formation. The question remains open to dif- may occur. The fact that only a fraction of the arsenic pres-
ferent interpretations and additional studies appear ent in these travertines can be assigned to the carbonate–
necessary. Nevertheless, we want to stress here that poten- arsenite substitution, suggests that the process of incorpo-
tial important contributions may result from the study of ration of As(III) in calcite can occur, but it is not known
natural systems like the PRV travertine. whether, in general, this process is effective as the adsorp-
It is worth noting that the carbonate replacement by tion of As by Fe-oxyhydroxides. We notice that the As
arsenite groups raises a question about charge balance. hosted in natural calcites is low if there is a competition
The simplest way would be uptake of the HAsO32 group with iron/manganese oxyhydroxides (see also Le Guern
instead of AsO33, as suggested by Roman-Ross et al. et al., 2003) but the effectiveness of mechanisms controlling
(2006). However, the presence of a protonated oxyanion the partition of As between these phases is still poorly
in this specific sample is contradicted by EPR evidence known. On the other hand, the evidence from laboratory
(Di Benedetto et al., 2006). Two possible alternatives can studies that As(III) concentrations in calcite may exceed
be considered: a concurrent uptake of a trivalent cation 2 g/kg (Roman-Ross et al., 2006), suggests that As incorpo-
(e.g., Rare Earth Elements) in the octahedral Ca site, or ration by this mineral may become important in conditions
coupled vacancy-replacement processes (i.e., two arsenite (e.g., reducing and/or alkaline environments) where adsorp-
groups and a vacancy for three carbonates). tion of As onto iron oxyhydroxides is not favored.
The observed heterogeneity of the samples would bene-
ficiate of the use of micro probe techniques, such Micro-X- ACKNOWLEDGMENTS
Ray Fluorescence (l-XRF), Micro-X-Ray Diffraction
(l-XRD) and Micro-X-ray Absorption Spectroscopy The authors want to express their warmest thanks to Andrea
(l-XAS), in order to further confirm the results presented Caneschi and Donella Rovai, University of Florence, and to Luisa
Poggi, Museo di Storia Naturale, University of Florence, for pro-
in this study. In particular, two dimensional fluorescence
viding some standards used in this study. FDB benefited of a re-
elemental mapping with microscopic resolution (l-XRF), search grant funded by the Regional Administration of Tuscany.
would reveal the association of As with other elements, Authors are indebted with Mario Paolieri and Mauro Rovezzi
while l-XRD and l-XAS would indicate the crystallo- for their support in the experimental and analytical work. This re-
graphic phases of As and its speciation in selected areas. search was funded by the PRIN 2008 (PC). We also acknowledge
3022 F. Bardelli et al. / Geochimica et Cosmochimica Acta 75 (2011) 3011–3023

the European Synchrotron Radiation Facility for provision of syn- Costagliola P., Benvenuti M. M., Benvenuti M. G., Di Benedetto
chrotron radiation facilities during experiments ME-1182 and EC- F. and Lattanzi P. (2010) Quaternary sediment geochemistry as
130, performed at BM08 (GILDA) beamline. a clue for tracing the source of toxic elements: a case study of
arsenic in the Pecora Valley (southern Tuscany, Italy). Chem.
Geol. 270, 80–89.
REFERENCES Di Benedetto F., Costagliola P., Benvenuti M., Lattanzi P.,
Romanelli M. and Tanelli G. (2006) Arsenic incorporation in
Alexandratos V. G., Elzinga E. J. and Reeder R. J. (2007) Arsenate natural calcite lattice Evidence from electron spin echo spec-
uptake by calcite: macroscopic and spectroscopic characteriza- troscopy. Earth Planet. Sci. Lett. 246, 458–465.
tion of adsorption and incorporation mechanisms. Geochim. Dixit S. and Hering J. G. (2003) Comparision of arsenic (V) and
Cosmochim. Acta 71, 4172–4187. arsenic (III) sorption on to iron oxide minerals: implication for
Ankudinov A. L., Ravel B., Rehr J. J. and Conradson S. D. (1998) arsenic mobility. Environ. Sci. Technol. 37, 4182–4189.
Real-space multiple-scattering calculation and interpretation of Dovesi R., Saunders V. R., Roetti C., Orlando R., Zicovich-Wilson
X-ray-absorption near-edge structure. Phys. Rev. B 58, 7565– C. M., Pascale F., Civalleri B., Doll K., Harrison N. M., Bush
7576. I. J., D’Arco P., Llunell M. (2006) CRYSTAL06 User’s
Appelo C. A. J. and Postma D. (2007) Geochemistry groundwater Manual, Torino (Italy).
and pollution. Balkema, p. 649. Fenter P. A, Rivers M. L., Sturchio N. C. and Sutton S. R. (2002)
Aurelio G., Fernández-Martı́nez A., Cuello G. J., Roman-Ross G., Applications of synchrotron radiation in low-temperature
Alliot I. and Charlet L. (2010) Structural study of selenium(IV) geochemistry and environmental science. Reviews in Mineralogy
substitutions in calcite. Chem. Geol. 270, 249–256. and Geochemistry, 49, Mineralogical Society, Washington.
Bardelli F., Meneghini C., Mobilio S., Ray S. and Sarma D. D. Fernández-Martı́nez A., Cuello G. J., Jonson M. R., Bardelli F.,
(2009) Local structure of Sr2FeMoxW1xO6 double perovskites Román-Ross G., Charlet L. and Turrillas X. (2008) Arsenate
across the composition driven metal to insulator transition. J. incorporation in gypsum probed by neutron, X-ray scattering
Phys.: Condens. Matter 21, 195502. and density functional theory modeling. J. Phys. Chem. A 112,
Becke A. D. (1993) Density functional thermochemistry: the role of 5159–5166.
exact exchange. J. Chem. Phys. 98, 5648–5652. Ferraris G., Jones D. W. and Sowden J. M. (1974) Hydrogen
Benvenuti M. G., Benvenuti M., Costagliola P. and Tanelli G. bonding in the crystalline state. A neutron and X-ray diffraction
(2009) The Quaternary in the mid-upper reach of the Pecora study of NaH2AsO4H2O, a crystal structure with unusual
River (southern Tuscany, Italy): implications for the paleohyd- pseudo-symmetry. Atti Accademia Scienze Torino 108, 507–527.
rographic evolution and surface sediments provenance. Boll. Fletcher R. (1970) A new approach to variable metric algorithms.
Soc. Geol. It. 128, 61–72. Comput. J. 13, 317–322.
Broyden C. G. (1970a) The convergence of a class of double-rank Goldberg S. and Glaubig R. (1988) Anion sorption on a calcar-
minimization Algorithms 1. “General Considerations”. IMA J. eous, montmorillonitic soil–arsenic. Soil Sci. Soc. Am. J. 52,
Appl. Math. 6, 76–90. 1297–1300.
Broyden C. G. (1970b) The convergence of a class of double-rank Goldfarb D. (1970) A family of variable-metric methods derived by
minimization Algorithms 1. “The New Algorithm”. IMA J. variational means. Math. Comp. 24, 23–26.
Appl. Math. 6, 222–231. Gopal R. and Calvo C. (1971) Crystal structure of Ca3(AsO4)2.
Causà M., Dovesi R. and Roetti C. (1991) Pseudopotential Canad. J. Chem. 49, 1036–1046.
Hertree-Fock Study of 17 III-V-Semiconductors. Phys. Rev. B Isaure M. P., Laboudigue A., Manceau A., Sarret G., Tiffreau C.,
43, 11937–11943. Trocellier P., Lamble G., Hazemann J. L. and Chateigner L.
Chakraborty S., Bardelli F. and Charlet L. (2010) Reactivities of (2002) Quantitative Zn speciation in a contaminated dredged
Fe(II) on calcite: selenium reduction. Environ. Sci. Technol. 44, sediment by mu-PIXE, mu-SXRF, EXAFS spectroscopy and
1288–1294. principal component analysis. Geochim. Cosmochim. Acta 66,
Charnock J. M., Polya D. A., Gault G. and Wogelius R. (2007) 1549–1567.
Direct EXAFS evidence for incorporation of As5+ in the James F. and Winkler M. (2004) Available from: <http://
tetrahedral site of natural andraditic garnet. Am. Mineral. 92, www.cern.ch/minuit>.
1856–1861. Jansen E., Kyek A., Schafer W. and Schwertmann U. (2002) The
Cheng L., Fenter P., Sturchio N. C., Zhong Z. and Bedzyk M. J. structure of six-line ferrihydrite. Appl. Phys. A 74, S1004–S1006.
(1999) X-ray standing wave study of arsenite incorporation at Jansen M. Z. (1979) Ueber eine neue Modifikation von As2O5.
the calcite surface. Geochim. Cosmochim. Acta 63, 3153–3157. Naturforsch. B 34, 10–13.
Costagliola P., Benvenuti M., Benvenuti M. G., Innocenti A., Jia Y., Xu L., Wang X. and Demopoulos G. P. (2007) Infrared
Mascaro I., Paolieri M., Rossato L. and Tanelli G. (2004) spectroscopic and X-ray diffraction characterization of the
Arsenic distribution in the Quaternary sediments of the median nature of adsorbed arsenate on ferrihydrite. Geochim. Cosmo-
valley of the Pecora stream (Grosseto, Italy). In Brownfields II chim. Acta 71(7), 1643–1654.
(eds. A. Donati, C. Rossi and C. A. Brebbia). WIT Press, Jönsson J. and Sherman D. M. (2008) Sorption of As(III) and
Southhampton UK, pp. 201–209. As(V) to siderite, green rust (fougerite) and magnetite: impli-
Costagliola P., Rimondi V., Benvenuti M., Chiarantini L., Di cations for arsenic release in anoxic groundwaters. Chem. Geol.
Benedetto F., Gasparon M., Lattanzi P. and Paolieri M. (2007) 255, 173–181.
Arsenic uptake by natural calcites: preliminary results from Kitahama K., Kiriyama R. and Baba Y. (1975) Refinement of the
sequential extraction of Travertines (Southern Tuscany, Italy). crystal structure of scorodite. Acta Crystallogr. B 31, 322–324.
IMWA Symposium 2007: May 27–31 (eds. F. Frau and R. Lee C., Yang W. and Parr R. G. (1988) Development of the Colle-
Cidu), Cagliari/Sardinia Italy, pp. 415–418. Salvetti correlation-energy formula into a functional of the
Costagliola P., Benvenuti M., Chiarantini L., Bianchi S., Di electron density. Phys. Rev. B 37, 785–789.
Benedetto F., Paolieri M. and Rossato L. (2008) Impact of Lee P. A., Citrin P. H., Eisenbergher P. and Kincaid B. M. (1981)
ancient metal smelting on arsenic pollution in the Pecora River Extended X-ray absorption fine structure – its strengths and
Valley, Southern Tuscany, Italy. Appl. Geochem. 23, 1241–1259. limitations as a structural tool. Rev. Mod. Phys. 53, 769–806.
 Arsenic uptake by natural calcite: An XAS study 3023

Le Guern C., Baranger P., Crouzet C., Bodénan F. and Conil P. Romanelli M., Benvenuti M., Costagliola P., Di Benedetto F.,
(2003) Arsenic trapping by iron oxyhydroxides and carbonates Lattanzi P. (2008) Electron Spin Echo invesitgation of Mn(II)-
at hydrothermal spring outlets. Appl. Geochem. 18, 1313–1323. As interaction in calcite, CaCO3: towards a quantitative model.
Lenoble V., Bouras O., Deluchat V., Serpaud B. and Bollinger J. C. Atti del 1 Meeting SIMP-AIC, Sestri Levante, 7–12 settembre
(2002) Arsenic adsorption onto pillard clays and iron oxides. J. 2008.
Colloid Interface Sci. 255, 52–58. Saalfield S. L. and Bostick B. C. (2009) Changes in iron, sulfur, and
Markgraf S. A. and Reeder R. J. (1985) High-temperature arsenic speciation associated with bacterial sulfate reduction in
structure refinements of calcite and magnesite. Am. Mineral. ferrihydrite-rich systems. Environ. Sci. Technol. 43, 8787–8793.
70, 590–600. Sadiq M. (1997) Arsenic chemistry in soils: an overview of
Matschullat J. (2000) Arsenic in the geosphere – a review. Sci. Total thermodynamic predictions and field observations. Water, Air
Environ. 249, 297–312. Soil Contam. 93, 117–136.
Menary J. W. (1958) The crystal structure of sodium polymeta- Shanno D. (1970) Conditioning of quasi-Newton methods for
arsenite (NaAsO2). Acta Crystallogr. 11, 742–743. function minimization. Math. Comp. 24, 647–656.
Monesi C., Meneghini C., Bardelli F., Benfatto M., Mobilio S., Sherman D. M. and Randall S. R. (2003) Surface complexation of
Manju U. and Sarma D. D. (2005) Local structure in LaMnO3 arsenic (V) to iron(III) hydroxides: structural mechanism from
and CaMnO3 perovskites: a quantitative structural refinement ab initio molecular geometries and EXAFS spectroscopy.
of Mn-K edge XANES data. Phys. Rev. B 74, 174104. Geochim. Cosmochim. Acta 67, 4223–4230.
Nriagu J. (1994) Arsenic in the Environment. Part I: Cycling and Schwertmann U., Staniek H. and Becher H.-H. (2004) Long-term
Characterization. Wiley Interscience Publication, New York. in vitro transformation of 2-line ferrihydrite to goethite/
Oremland R. S. and Stolz J. F. (2003) The ecology of Arsenic. hematite at 4, 10, 15 and 25 °C. Clay Mineral 39(4), 433–438.
Science 300, 939–944. Smedley P. L. and Kinninburg D. G. (2002) A review of the source,
Oremland R. S., Kulp T. R., Blum J. S., Hoeft S. E., Baesman S., behaviour and distribution of arsenic in natural waters. Appl.
Miller L. G. and Stolz J. F. (2005) A microbial arsenic cycle in a Geochem. 17, 517–568.
salt-saturated, extreme environment. Science 308, 1305–1308. Sø H. U., Postma D., Jakobsen R. and Larsen F. (2008) Sorption
Ouvrard S., de Donato P., Simonnot M. O., Begin S., Ghanbaja J., and desorption of arsenate and arsenite on calcite. Geochim.
Alnot M., Duval Y. B., Lhote S., Barres O. and Sardin M. Cosmochim. Acta 72, 5871–5884.
(2005) Natural manganese oxide: combined analytical approach Tang Y., Elzinga E. J., Lee Y. J. and Reeder R. J. (2007)
for solid characterization and arsenic retention. Geochim. Coprecipitation of chromate with calcite: batch experiments
Cosmochim. Acta 69, 2715–2724. and X-ray absorption spectroscopy. Geochim. Cosmochim. Acta
Pascarelli S., Boscherini F., D’Acapito F., Hrdy J., Meneghini C. 71, 1480–1493.
and Mobilio S. (1996) X-ray optics of a dynamical sagittal- Tufano K. J., Reyes C., Saltikov C. and Fendorf S. (2008)
focusing monochromator on the GILDA beamline at the Reductive processes controlling arsenic retention: revealing the
ESRF. J. Synch. Rad. 3, 147–155. relative importance of iron and arsenic reduction. Environ. Sci.
Pentecost A. (1995) The quaternary travertine deposits of Europe Technol. 42, 8283–8289.
and Asia Minor. Quat. Sci. Rev. 14, 1005–1028. Valenzano L., Torres F. J., Doll K., Pascale F., Zicovich-Wilson C.
Pertlik F. (1978) Structure refinement of cubic As2O3 (arsenolite) M. and Dovesi R. Z. (2006) Ab initio study of the vibrational
with single crystal data. Czech. J. Phys. B 28, 170–176. spectrum and related properties of crystalline compounds; the
Ramı́rez-Solı́s A., Mukopadhyay R., Rosen B. P. and Stemmler T. case of CaCO3 calcite. Z. Phys. Chem. 220, 893–912.
L. (2004) Experimental and theoretical characterization of Yokoyama Y., Mitsunobu S., Tanaka K., Itai T. and Takahashi Y.
arsenite in water: insights into the coordination environment of (2009) A study on the coprecipitation of arsenite and arsenate
As–O. Inorg. Chem. 43(9), 2954–2959. into calcite coupled with the determination of oxidation states
Ravel B. (2001) ATOMS: crystallography for the X-ray absorption of arsenic both in calcite and water. Chem. Lett. 38, 910–911.
spectroscopist. J. Synch. Rad. 8, 314–316. Wilke M., Farges F., Petit P. E., Brown, Jr., G. E. and Martin F.
Rehr J. J. and Albers R. C. (2000) Theoretical approaches to X-ray (2001) Oxidation state and coordination of Fe in minerals: an
absorption fine structure. Rev. Mod. Phys. 72, 621–654. Fe K-XANES spectroscopic study. Am. Miner. 86, 714–730.
Roman Ross G., Cuello G. J., Turrillas X., Fernandez Martinez A.
and Charlet L. (2006) Arsenite sorption and co-precipitation
Associate editor: Roy A. Wogelius
with calcite. Chem. Geol. 233, 328–336.