Journal of Cultural Heritage 32 (2018) 30–37

Available online at

ScienceDirect
www.sciencedirect.com

Original article

Laboratory study of the sulfation of carbonate stones through SWIR

hyperspectral investigation
Amelia Suzuki a , Silvia Vettori b,∗ , Silvia Giorgi a , Emiliano Carretti c ,
Francesco Di Benedetto a,d,∗ , Luigi Dei c , Marco Benvenuti a,d , Sandro Moretti a ,
Elena Pecchioni a,b , Pilario Costagliola a
a
Department of Earth Sciences, University of Florence, via La Pira 4, 50121 Florence, Italy
b
Institute for the Conservation and Valorization of Cultural Heritage (ICVBC), National Research Council, via Cozzi 53, 20125 Milan, Italy
c
Department of Chemistry, University of Florence, Florence, Italy
d
Institute for Geosciences and Georesources (IGG), National Research Council, Florence, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Stone-built Cultural Heritage is subjected to decay in urban environment over the centuries, due to sur-
Received 8 September 2017 face interaction and reaction with natural atmospheric agents and, particularly in the last centuries, air
Accepted 10 January 2018 pollutants. The Short wave Infrared (SWIR) characterisation of stone surface through portable instru-
Available online 9 February 2018
ments is attracting increasing interest in the field of Cultural Heritage. In this study, SWIR hyperspectral
investigation of carbonate rocks, undergoing acid attack under laboratory conditions was performed
Keywords: with the aim of providing useful quantitative information on the degree of sulfation of the surfaces of
IR reflectance
carbonate stone. Six marble and six travertine specimens were attacked by aqueous solutions of H2 SO4 at
Hyperspectral device
®
ASD Fieldspec 3
variable acid concentrations leading to the formation of gypsum. The reacted surfaces of stones were then
Stone damage investigated by a portable SWIR spectroradiometer. The resulting spectra were thus modelled through a
Gypsum full profile approach, in order to obtain a reliable and efficient fitting procedure. Thus, the SWIR charac-
Marble terisation of sulfated carbonate surfaces seems to be a promising, ready-to-use technique for monitoring
Travertine the conservation state of carbonate stone monuments (e.g. facades, statues). This method could provide
valuable support both for restoration practices and for continuous monitoring of stone alteration over
time, when assessing the best strategy of intervention and conservation against sulfation processes of
historical buildings.
© 2018 Elsevier Masson SAS. All rights reserved.

1. Research Aims 2. Introduction

This work is focus on the capability of the hyperspectral tech-
nique to give information on the degree of sulfation of the surfaces
of carbonate stone. This technique, recently introduced in the field
of Cultural Heritage conservation, is completely non-invasive and
allows to gain spectral information in both the visible (VIS) and near
infrared (NIR) regions using a portable spectroradiometer (ASD
®
Fieldspec 3). The study is undertaken through the artificial alter-
ation of carbonate stone specimen under laboratory conditions.
∗ Corresponding authors. Department of Earth Sciences, University of Florence,
via La Pira 4, 50121 Florence, Italy.
E-mail addresses: vettori@icvbc.cnr.it (S. Vettori), francesco.dibenedetto@unifi.it
(F. Di Benedetto).
The impact of atmospheric agents and air pollution on Cultural
Heritage stone-built materials is a serious concern, as it potentially
threatens important parts of our history and culture. Atmospheric
pollution, frost and salt weathering are traditionally considered the
major causes of building stone decay [1]. Main damages include sur-
face corrosion, soiling and bio-degradation, with consequent loss
of details, blackening and formation of crusts on stone surfaces.
In recent years, major changes in both the sources and amounts
of emissions of air pollution have deeply modified both the rate
and the extent of damage of historical buildings. The proficiency
of a combined study of meteorological parameters and air pollut-
ant distribution to unravel the role played by climate changes in
the weathering of stone buildings and monuments is attracting an
increasing consensus among scientists [2,3]. Among the main com-
ponents of air pollution, a reduction of the sulfur species (mainly
SO2 ) and, at the same time, an increasing role of nitrogen and

https://doi.org/10.1016/j.culher.2018.01.006
1296-2074/© 2018 Elsevier Masson SAS. All rights reserved.
 A. Suzuki et al. / Journal of Cultural Heritage 32 (2018) 30–37 31

carbonaceous species (NOx , nitrate, organic gases, elemental and Table 1

List of the investigated samples.
organic particles) [4–7] have been recently observed.
In this study, we have examined one of the most diffused Rock type Label Sulfation procedure SWIR SEM-EDS
degradation processes affecting both natural and artificial carbon- Marble M No acid attack X
atic materials exposed to the urban atmosphere: the formation of M01 1 h in H2 SO4 0.1 M X
sulfate-based deposits (i.e. “black crusts”) [8]. In particular, the vol- M03 1 h in H2 SO4 0.3 M X
ume increase due to the formation of gypsum from calcite [9] is the M1 1 h in H2 SO4 1 M X
M2 1 h in H2 SO4 2 M X
major cause of the development of cracks in the carbonate substra-
M4 1 h in H2 SO4 4 M X
tum of black crusts, often leading to ultimate degradation of stone M5 1 h in H2 SO4 5 M X
materials. M1SC 1 h in H2 SO4 1 M X
The decreased content of acidic pollutants in urban atmosphere M2SC 1 h in H2 SO4 2 M X
M5SC 1 h in H2 SO4 5 M X
during the late 20th century is held responsible for the observed
Travertine T01 1 h in H2 SO4 0.1 M X
decreasing rate of formation of degradation crusts in the last years. T03 1 h in H2 SO4 0.3 M X
According to Bonazza et al. [2], it can be envisaged that in the T1 1 h in H2 SO4 1 M X
next future black crusts will be substituted by other forms of stone T2 1 h in H2 SO4 2 M X
alteration driven by pollutants other than SOx . In this scenario, T4 1 h in H2 SO4 4 M X
T5 1 h in H2 SO4 5 M X
black crusts themselves could paradoxically become part of our
T1SC 1 h in H2 SO4 1 M X
Cultural Heritage as witnesses of past atmospheric conditions of T2SC 1 h in H2 SO4 2 M X
our cities, although presently sulfation is still among the most rele- T5SC 1 h in H2 SO4 5 M X
vant alteration processes of carbonatic stone buildings in the urban Gypsum gy1 No acid attack X
environment [10]. Accordingly, the monitoring of stone alteration
represents a key factor in the knowledge and prediction of the
trend affecting the building stone in the XXI cent. urban framework ornamental purposes in the architecture of central Italy, since the
[2,11]. The continuous monitoring of the conservation state of out- Roman age.
door surfaces, including the qualitative and quantitative evaluation Six specimens of each lithotype (M01, M03, M1, M2, M4, M5
of the surface degradation, is then considered a good practice to and T01, T03, T1, T2, T4, T5), obtained after cutting small paral-
understand stone damages and for planning and optimize conser- lelepipeds (about 5 × 5 × 1.5 cm in size) from a large stone sample,
vative actions aimed to preserve historical buildings [12–14]. were used for the acid attack (at different acid concentrations: see
Continuous monitoring requires a non-destructive analytical below) and subsequent SWIR investigations (Table 1). In addition,
approach and, possibly, a simple, low-cost and effective tool to three specimens of marble (labelled as M1SC, M2SC, M5SC) and
study the decay processes affecting works of art. three of travertine (labelled as T1SC, T2SC, T5SC) (Table 1), smaller
Many non-destructive and non-invasive techniques like ther- in size (cubes of side 2.5 cm) were investigated by SEM-EDS after
mography [15], P-wave velocity measurements [16], Laser Imaging acid attack carried out with aqueous solutions of H2 SO4 at different
Detection and Ranging (LIDAR) [17] and Fiber Optic Reflectance concentrations.
Spectroscopy (FORS) [18] have been successfully applied in the The calibration of the SWIR spectroradiometer for this study
last 20 years to monitor the degradation of works of art exposed requires the use of suitable materials as reference standards for
outdoors. Traditional invasive or micro-invasive monitoring tech- the polycrystalline calcite and gypsum phases. A sample of gypsum
niques like X-ray diffraction (XRD), Fourier-Transform Infrared (gy1), obtained by grinding and sieving to a particle size < 15 ␮m a
spectroscopy (FTIR), Scanning Electron Microscopy (SEM) and ion single crystal of gypsum, and of a calcitic marble specimen (M, from
chromatography still represent the most frequently used meth- the Gioia quarry, Apuane Alps, Italy) were thus selected as internal
ods due to their accuracy, although the analyses are expensive and standards for gypsum and calcite, respectively.
usually time-consuming. As a further drawback, an accurate and The sulfation of the specimens was carried out through acid
representative information about the general distribution of the attack under ambient conditions (T = 20◦ C). Aqueous solutions of
degradation products and of the general conservation status of a H2 SO4 were prepared at different molar concentrations (from 0.1
work of art, through XRD, FTIR and SEM, implies a collection of a to 5 M, Tab. 1) and each sample was submerged half-height in one
large number of samples in order to obtain a statistically represen- of these solutions for one hour (Fig. S2). The specimen was then
tative physico-chemical characterization of the material substrate dried for 3 days at room temperature (around 20◦ C); then, it was
and an accurate monitoring of its conservation status. maintained under vacuum at 60◦ C for 3 days and put in a dry box
Hyperspectral technique is an analytical method recently containing anhydrous calcium chloride until a constant weight was
introduced in the field of Cultural Heritage conservation. Namely, reached. Then the total amount of gypsum formed on each sample
hyperspectral analysis has been used to monitor the presence of was calculated as follows:
both sulfates and oxalates on carbonatic stone materials of histor- W
ical, artistical and architectonical interest [19,20]. This technique a n=
MWgy − MWcc
quick discrimination of different components (in rocks, soils, etc.)
on the basis of their different spectroscopic signatures [21–23]. Where:
n (in mol) is the total amount of gypsum formed on the surface
of each sample, W is the difference of sample weight before and
after the H2 SO4 attack; MWgy and MWcc are the molecular weights
3. Experimental procedures of gypsum and calcite, respectively.
The SEM-EDS analysis was carried out on carbon-coated sam-
3.1. Materials and methods ples in order to verify the presence of gypsum, to measure the
thickness of the sulfated layer, and to evaluate the change in mor-
Two different carbonate stones were used for this study: Carrara phology, grain size and distribution of the gypsum crystals as a
marble samples, from the Gioia quarry (Apuane Alps, north- function of H2 SO4 concentration. Besides conventional samples,
ern Tuscany) and travertine from northern Latium. These two exhibiting the sulfated surface, cross-sections were also realised. A
stones were chosen because they are among the most used for Scanning Electron Microscope ZEISS EVO MA 15, operating at 20 kV
32 A. Suzuki et al. / Journal of Cultural Heritage 32 (2018) 30–37

Table 2
Main bands detectable for calcite and gypsum in the SWIR region (1200–2500 nm).

Mineral ␭ (nm) a
Type Group Assignment Ref.

Calcite 2550 1 Combination and/or overtone Carbonate ␯1 + 2␯3 [24,25]

Calcite 2350 2 Overtone Carbonate 3␯3 [24,25]
Calcite 2260 3 Combination and/or overtone Carbonate – [24,25]
Calcite 2160 4 Combination and/or overtone Carbonate ␯1 + 2␯3 + ␯4/3␯1 + 2␯4 [24,25]
Calcite 2000 5 Combination and/or overtone Carbonate 2␯1 + 2␯3 [24,25]
Calcite 1900 6 Combination and/or overtone Carbonate ␯1 + 3␯3 [24,25]
Calcite 1756 7 Combination and/or overtone Carbonate – [24,25]
Gypsum 1751 1 Combination Hydroxyl/water – [26]
1780 2
Gypsum 1944 3 Combination Water – [26]
1970 4
Gypsum 2170 5 Combinations and/or overtones Sulfate/hydroxyl/water 33 S–O or OH/H2 O [26]
Gypsum 2220 6 Combinations and/or overtones Sulfate/hydroxyl/water 33 S–O or OH/H2 O [26]
Gypsum 2280 7 Combinations and/or overtones Sulfate/hydroxyl/water 33 S–O or OH/H2 O [26]
Gypsum 2430 8 Overtone Sulfate 33 [26]
2480 9
Gypsum 2550 10 Combination Hydroxyl/water OH/H2 O + /ıOH/H2 O [26]
a
Label of the considered band.

with a tungsten filament and equipped with an Oxford INCA EDS
X-ray microanalysis, was used to analyse the six samples.
3.2. SWIR spectroscopy
A field portable high-resolution spectroradiometer, ASD-
®
FieldSpec 3, has been used to collect reflectance measurements.
The technical features of this spectroradiometer have been thor-
oughly described by Camaiti et al. [20]. Briefly, the FieldSpec
spectroradiometer is a compact instrument designed to acquire
VIS/NIR (350–1050 nm wavelength domain) punctual reflectance
spectra with a rapid scan: data collection times are of the order
of ∼ 0.1 s for each spectrum. The SWIR portion of the spectrum is
obtained by Gaffey [25], who successfully refined the SWIR region
of calcite spectra through a spectral decomposition in several
Gaussian contributions. With respect to that study, our fitting pro-
cedure differs because no preliminary normalization of the spectra
was carried out. This procedure was chosen to minimize any pos-
sible numerical alteration of the original spectral features (namely,
positions, intensities and widths of the bands).
The spectra of the standards (i.e. the marble M and gypsum gy1,
Table 1) were fitted by a model described by the following equation
[25]:

y () = a0 + a1 + a2 2 + a3 3 + FI () (1)
i=1,n

acquired with two scanning spectrometers. The first spectrome-
ter (SWIR1) operates in the range 900–1850 nm; the second one
(SWIR2) covers the region between 1700 and 2500 nm. The control-
ling software automatically accounts for the overlap in wavelength
intervals. The sampling interval for each SWIR region is about
2 nm, and the spectral resolution varies between 10 nm and 12 nm,
depending on the scan angle [20,24]. Field spectrometry typically
requires the ambient solar illumination. In our case, however, we
®
used the ASD-FieldSpec 3 spectroradiometer’s contact reflectance
probe (spot size of about 10 mm) with the internal light source
(halogen lamp, 6.5 W; Fig. S3).
®
The ASD-FieldSpec 3 measures the reflectance of a given target
by comparing it with a reference material with known reflectance
properties. We used as a reference the Spectralon (Labsphere,
Inc.), a polytetrafluoroethylene (PTFE) thermoplastic resin show-
ing approximately 100% reflectance across the entire spectrum. In
this study, white reference (Spectralon) has been acquired every
two sample spectra and, in order to minimise the spectral noise,
we registered 3 spectra (70 scans for each spectrum) for each sam-
ple, averaging them. The standard deviation of the 3 spectra on the
same specimens is estimated to be around 10−4 . Aiming at obtain-
ing reflectance spectra corresponding to the same sampling area,
we prepared transparency films masks to be correctly positioned
on the study area using a suitable number reference points; the
masks were provided by holes having a diameter equal to the con-
tact probe spot size (Fig. S4). The measurements were carried out
on each sample before and after the acid attack.
3.3. SWIR fitting procedure
The interpretation of the SWIR spectra was based on a full-
profile approach, i.e. simultaneously fitting bands and background
by an opportune model. This procedure is based on the results
where y() are the experimental SWIR data, a0 , a1 , a2 and a3 are
the polynomial coefficients of a cubic function (used to model the
background), and Fi () are the Gaussian contributions, expressed
by
 2
−xc,i
−2
Ai wi
Fi () = √ e (2)
wi 
⁄2
with Ai , wi , xc,i representing the intensity, width and cen-
tre of the i-th contribution, respectively. Spectra were fitted
in the region between 1630 and 2500 nm. This range was
chosen to avoid/minimize spectral interference at wavelength
value < 1630 nm. Best fit was achieved by a non linear least squares
refinement, carried out under the Levenberg-Marquardt strategy,
through the use of the Microcal Origin 9.0 software. The only free
parameters were: a0 , a1 , a2 , a3 , wi and Ai (i = 1,n). The number (n)
and position (xc,i ) of all contributions were chosen on the basis
of the bands described in the literature [25–27] in the considered
spectral region. In the 1630–2500 nm range, seven different bands
characteristic of calcite, listed in Table 2, occured between 1500
and 2500 nm. They arised from combination and/or overtones of
the fundamental vibrations of the carbonate group [25,26]. More
complex is the situation for gypsum: in this phase, both OH/H2 O
and sulfate groups have fundamental vibrations, from the combi-
nation of which a band in the SWIR region can appear. Accordingly,
10 bands are assigned to gypsum (9 of which in the selected region,
Table 2) [27]. As a consequence, calcite and gypsum spectral con-
tributions are quite well defined, largely due to the differences in
the main absorption bands at 2350 and 1950 nm, respectively.
Spectra of the sulfated specimens were fitted by a second model,
where the calcite and gypsum contributions were evaluated by
weighing the best fit simulated contributions of the reference com-
pounds. This approach is justified by the additive properties of the
 A. Suzuki et al. / Journal of Cultural Heritage 32 (2018) 30–37 33

Table 3
Results of the gravimetric analysis.

Sample W (g) Gypsum mol Gypsum wt %

M01 −0.03 −4.72E-04 −0.06

M03 0.01 1.29E-04 0.02
M1 0.05 6.92E-04 0.14
M2 0.06 8.35E-04 0.17
M4 0.15 2.13E-03 0.29
M5 0.16 2.16E-03 0.43
T01 −0.04 −5.88E-04 −0.09
T03 0.03 3.78E-04 0.05
T1 0.12 1.71E-03 0.33
T2 0.24 3.30E-03 0.66
T4 0.36 5.06E-03 0.79
T5 0.49 6.73E-03 1.28

Fig. 1. Amount of formed gypsum on the sulfated carbonatic specimens: marble (in
SWIR spectroscopic signatures of calcite and gypsum [19,20]. The black) and travertine (in red).
background is still described by a cubic polynomial function. The
overall adopted function is:

y () = a0 + a1  + a2 2 + a3 3 + kG GY () + kG CC () (3)

where y() are the experimental SWIR data, a0 , a1 , a2 and a3 are

the polynomial coefficients of the background function, kG and kC
the weights through which the two spectral components of gyp-
sum, GY(), and calcite, CC(), are evaluated. Indeed, GY() and
CC() are defined as:

 
GY () =

Fi () (4)

i=1,9 best fit

 

CC () = Fi () (5)

i=1,7 best fit

Also in this case, the best fit was achieved through a non-linear

least squares refinement, carried out under the Levenberg-
Marquardt strategy, using of the Microcal Origin 9.0 software. The
refined parameters were: a0 , a1 , a2 , a3 , kG and kC .

4. Results

4.1. Gravimetry, capillary water absorption, SEM-EDS

Table 3 shows that the amount of gypsum formed onto the

surface of the treated carbonate specimens, calculated assuming
that the stone samples were totally made by calcium carbonate,
increases with the molarity of the H2 SO4 solution. In particular,
the wt% of gypsum crystallized onto the surface of the two carbon-
atic matrixes monotonically increases upon increasing the molar
concentration of the H2 SO4 solution (Fig. 1). After acidification, the
gypsum contents of the travertine samples are always higher than
those of marble. The negative values observed for the samples after
very mild treatment (i.e. T01 and M01) are within the uncertainty
range of the procedure.
SEM micrographs of the surface of sulfated samples are shown
in Fig. 2. Secondary electron (SE) images and semiquantitavive EDS
analyses indicates that the surfaces of all sulfated samples (inde-
pendent from the acid concentrations of reacting solutions) are
characterised by a continuous and homogeneous layer of acicular
crystals, about 5–10 ␮m long (Fig. 2a and b), showing a composition
fully compatible with the presence of a Ca-sulfate phase.
The thickness of the sulfated layer could be estimated in cross-
Fig. 2. a and b: SEM micrographs (secondary electrons) of the MC1S and TC1S
sections perpendicular to the surface. As shown in Fig. 2c, this layer, samples, respectively, registered at 1800 × magnification, where acicular crystals
evidenced by the sulfur X-ray EDS map, shows variable thickness, of gypsum are visible; c: semiquantitative (for sulphur) S X-ray map of the section
on average around 50 ␮m. This variability (Fig. 2c) may be ascribed normal to the attacked surface (on the top edge) of the sample MC2S; magenta
to local textural features of the rock fabric, such as the presence regions reveal high S contents; scale bar is 200 ␮m.
34 A. Suzuki et al. / Journal of Cultural Heritage 32 (2018) 30–37

Table 4
Linear combination coefficients for calcite (kC ) and gypsum (kG ), with uncertainty
in brackets; least squares agreement factors (R2 ) are also reported.

Sample kC kG R2

M01 0.93 (1) 0.00 (1) 0.97960

M03 0.85 (1) 0.12 (1) 0.98859
M1 0.87 (1) 0.46 (1) 0.99082
M2 0.73 (1) 0.59 (1) 0.98918
M4 0.51 (1) 0.69 (1) 0.99218
M5 0.52 (1) 0.82 (1) 0.99168
T01 0.58 (1) 0.11 (1) 0.97232
T03 0.60 (1) 0.18 (1) 0.97737
T1 0.61 (1) 0.68 (1) 0.92281
T2 0.52 (1) 0.78 (1) 0.98382
T4 0.31 (1) 0.83 (1) 0.99739
T5 0.36 (1) 0.84 (1) 0.98846

experimental and calculated spectra is evident. A representative

SWIR spectrum of a sulfated sample is shown in Fig. 3c. Despite of
the rigidity of the functional model applied to fit the SWIR spec-
tra (only two parameters are kept free during the refinement to
account for the band intensity, in the whole range: see eq. (3)), the
best-fit curves reproduce the experimental data with a relevant
accuracy (mean agreement factor R2 = 0.981, Table 4). The output
parameters obtained from the best fit of the linear combination
coefficients, kG and kC , are also listed in Table 4. The agreement
factors confirm the statistical reliability of the performed fits.

5. Discussion

5.1. Analytical consideration about the SWIR protocol

Under the assumption that the moles of calcite and of gypsum

Fig. 3. a: exemplar experimental spectra of the investigated samples: reference cal-
cite (M), aged surface of the M5 sample and reference gypsum powders (gy1); b:
comparison of the experimental (in black) and best fit calculated spectra (in red) for
the pristine surface of M1(cc) and of the gy1 gypsum powders; c: comparison of the
experimental (in black) and best fit calculated spectra (in red) for the aged surface
of M1.
of pores or holes. For instance, in travertine, close to the pores S-
contents have been revealed at depth higher than 50 ␮m.
4.2. SWIR spectroscopy
The spectra of the reference M calcite and gy1 gypsum, together
with that of a sulfated specimen (M5) are shown in Fig. 3a. As indi-
cated above, calcite and gypsum spectral contributions are well
defined. However, the comparison between the spectra of the two
reference compounds in Fig. 3a confirms that calcite and gypsum
SWIR bands are partially overlapped. The large values of width
are due to both crystal grain size [28]. The best-fit curves closely
reproducing the SWIR experimental spectra of calcite and gyp-
sum are shown Fig. 3b, where the good agreement between the
bring additive contributions to the SWIR spectrum, the trends of
the kC and kG coefficients with the increase of gypsum, produced
by acidification, can be considered proxies of the amount of calcite
and gypsum, respectively, in the studied sample.
In Fig. 4a kC is plotted against the amount of gypsum, as derived
from the gravimetric analysis (Table 3). This parameter shows only
moderate variations with increasing the gypsum content: maxi-
mum variation is −0.4 for marble, and −0.3 for travertine. The fact
that kC never attains a null value, even for the samples that under-
went the strongest attack, points to the occurrence in the samples
of a layered structure, i.e. a thin layer of gypsum covers the car-
bonate surface, in agreement with the SEM investigations. SEM
images indicate, in fact, that the gypsum layer is homogeneously
distributed over the whole sample surface. Thus, the residual pres-
ence of carbonate in the SWIR spectra shows that the sample depth
investigated by the SWIR radiation is deeper than the thickness of
the gypsum layer.
When the trends of kC versus the gypsum moles for marbles and
travertines are compared, significant differences, especially at low
amounts of gypsum formed, are observed. Different trends for mar-
ble and travertine are in agreement with the different porosities of
the two lithotypes [29] (see also Supplementary Materials). In par-
ticular, for gypsum contents below 1 mmol, the marble samples
present kC values up to 0.9, whereas in travertines the maximum
value is slightly above 0.6. The uncertainty of the method has been
extrapolated from these values. One has to recall, in fact, that the
procedure described in paragraph 3.3 provides a quantitative esti-
mate of the calcite contribution by comparing the SWIR pattern of
an unknown sample with that of a marble reference. Similarly, the
gypsum pattern in an unknown sample has been compared with
that of the gy1 powders (eq. (3), (4) and (5)). Thus, differences in
the kC values for samples where almost no gypsum is present are
only due to differences in the rock fabric between unknown and
reference samples. From a physical point of view, the difference
 A. Suzuki et al. / Journal of Cultural Heritage 32 (2018) 30–37
Fig. 4. Data of (a) kC and (b) kG versus the moles of the gypsum formed after the
attack, for travertine (full circles) and marble (open circles) specimens. Error bars
for kC and kG .
between kC for marble and travertine samples relies again in the
different porosities of the two lithotypes, which in turn reflect in
different volumes sampled by the SWIR radiation (i.e. the pene-
tration depth changes) and, ultimately, in the different amounts of
calcite contained in these volumes.
Concerning marble, the fact that the highest kC value (0.93,
M01 sample) is below the unity points to slight differences in the
matrix, the effect of which to the SWIR spectrum can be consid-
ered unavoidable. Indeed, we assume hereafter that the uncertainty
in the determination of kC is ± 0.1. The uncertainty of kG rela-
tive to the whole procedure can be estimated in a similar way.
In the left side of the Fig. 4b, the discrepancy between gravimet-
ric and SWIR evaluation of the gypsum content point again to a
± 0.1 units.
5.2. SWIR parameter trends versus increasing gypsum content
and interpretation model
The plot showing the trend of kG versus the amount of gyp-
sum (Fig. 4b) is characterised by a steep increase of the kG value
at relatively low gypsum contents, followed by an asymptotic
approach to a plateau for values close to 0.9. The origin of this
plateau is related to the presence of the gypsum layer (about 50
microns), evidenced by SEM analyses, covering the altered sam-
ples. We attribute the plateau to the occurrence of a passivation
layer of gypsum over the carbonate surface, which prevents the
continuous increase of the film thickness. This process is in turn
related to the increase of viscosity of the acidic solution with
increasing of the H2 SO4 concentration. With increasing acid con-
centration, under the experimental conditions of the attack (the
degradation of the carbonate surface is induced without stirring
the solution, so that reaction is controlled only by the diffu-
35
Fig. 5. Schematic representation of the interaction between the SWIR radiation and
the layered samples. The region investigated by the SWIR radiation during the mea-
surement is schematized by the dotted cylinder, where the grey region represents
the gypsum layer, and the white one that of calcite. hTOT = depth investigated by
SWIR; hG = the thickness of the gypsum layer.
sion of the reactants), the inlet of the fresh solution towards the
sample surface decreases. Indeed, the alteration process results
limited.
To have a clearer picture of the numerous factors concurring
to the final profile of the SWIR reflectance spectra, we can assume
a simplified model (Fig. 5), based on our experimental evidence.
Our results indicate that the sulfation process of a carbonate stone
(marble or travertine) surface results in a layered structure. Sup-
posing that the thickness of the gypsum layer formed on the sample
surface is uniform, and that the apparent density of the calcite sub-
strate and of the gypsum layer are homogeneous, the geometry of
the system can be described by two relevant parameters (besides
carbonate and sulfate apparent density): the depth investigated by
SWIR radiation (hTOT ), and the thickness of the gypsum layer (hG ).
The hG and hTOT values are in turn depending on numerous relevant
factors:
• the physico-chemical conditions of the acid attack, i.e. the molar-
ity of the H2 SO4 solutions and their viscosity, that apparently
control the amount of gypsum formed, and then its thickness,
hG ;
• the specific optical absorption coefficients, affecting hTOT , which
in turn depends on the composition of the sample surface as well
as on sample texture (porosity, crystal grain size, impurities, . . .)
[28].
We stress that the porosity of the material substrate is a key fac-
tor deeply influencing the SWIR spectra: the higher the differences
in porosity between reference compounds and specimen, the larger
the difference of the amount of the investigated phase (both calcite
and gypsum), detected by SWIR with respect to the amount of the
same phase in the reference sample.
From the above discussion, it appears that the quantitative
analysis of sulfation by using SWIR spectra is heavily influenced
by matrix effects. An accurate choice of representative stan-
dard materials appears crucial to this aim. In the present case,
where standards tailored to the specific features of the sam-
ples were available (i.e. M and gy1, Table 1), a straightforward
application for diagnostic purposes is achieved. In fact, by choos-
ing kG as the most relevant variable, our experimental results
indicate that the degree of sulfation of a carbonate surface can
be monitored by kG values falling in one out of three possible
ranges:
36 A. Suzuki et al. / Journal of Cultural Heritage 32 (2018) 30–37
• 0 < kG < 0.25–degree of sulfation: low;
• 0.25 < kG < 0.50–degree of sulfation: moderate;
• kG > 0.5–degree of sulfation: high.
Since the estimate of the gypsum abundance at the carbonate
surface is greatly dependent upon the differences in grain size, tex-
ture, fabric, porosity existing between the reference and sample
matrix, it can be concluded that our SWIR based protocol gives
its best results when these differences are minimised. Fully quan-
titative determinations will be thus available after having set up
calibration procedures appropriate to the investigation subject (e.g.
outdoor/indoor, flat/modelled). Conversely, the semiquantitative
diagnostic tool here proposed can be considered as already mature
for case studies as:
• surface cleaning, where the progresses or the effectiveness of a
cleaning procedure could be evaluated using our approach by
acquiring spectra at different cleaning stages;
• monitoring of the alteration rate with time on stones exposed
outdoor to atmospheric reactants (as e.g. the case of the façades
of historical buildings).
6. Conclusions
This study provides a clue to an easy implemented protocol
to evaluate the state of sulfation of a carbonate stone, and to
establish continuous monitoring of open air stone artworksin a
non-destructive and non-invasive way [20].
The main results of the present study point that SWIR spec-
troscopy is capable to detect the presence of a gypsum layer, one
of the main products that cause the degradation and the alteration
of carbonate stone, with specific reference to marble and traver-
tine. However, the quantitative determination of gypsum content
by a portable SWIR device appears a task not generally allowed.
This is mainly due to the differences in the penetration depths of
the SWIR radiation in a layered sample, where also the rock’s tex-
ture plays a relevant role. The proposed approach is based on the
full profile analysis of the SWIR data and on the use of the lin-
ear combination kG coefficient as a proxy (within defined ranges)
of the state of sulfation of the surface. The development of this
calibration procedure opens very interesting perspectives in the
conservation of ancient monuments. Specific applications, in fact,
can be hypothesized for monitoring the state of conservation of
outdoor monuments through the investigation of the alteration of
their carbonate surfaces over time, due to different exposures to
atmospheric agents (e.g. rain, wind, solar radiation) and to pollut-
ants (e.g. vehicular traffic). A monitoring campaign, in fact, based
on the proposed procedure can give better results when applied
in temporal sequence on the same surface spot. Thus, a monitor-
ing campaign conducted on a given number of strategic positioned
check points may record how the gypsum layer may evolve trough
time. Moreover, SWIR monitoring could be applied:
• during the assessment and control of the proper cleaning meth-
ods needed to remove the sulphated layers, when the most
effective cleaning tests can be compared and control steps mon-
itored;
• to determine whether or not a protective layer should be applied
to the carbonate surface analysing the increase of sulfation with
time on treated and untreated areas, and if necessary, which type
of protective would be most effective;
• if there is a detailed record of conservation methods used in the
past, a comparison can be made between the areas left as they
were, those cleaned, and those cleaned and covered with a pro-
tective layer in order to identify which procedure guarantees the
smaller increase of sulfation on the surfaces investigated.
Such evaluations and comparisons are expected to be very useful
when assessing the best strategy of intervention and conservation
against sulfation processes on our historical buildings.
Acknowledgements
The Centro di Servizi di Microscopia Elettronica e Microanalisi
(MEMA), University of Florence, is acknowledged for kindly grant-
ing the use of the Scanning Electron Microscopy, as well as Mario
Paolieri and Maurizio Ulivi are acknowledged for their assistance
in the same investigations. The Italian CNR is acknowledged for
support. FDB also benefited for departmental funding (ex 60%).
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at https://doi.org.10.1016/j.culher.2018.01.006.
References
[1] D.B. Honeyborne, Weathering and decay of Masonry, in: J. Ashurst, F.G.
Dimes (Eds.), Conservation of building and decorative stones, Butterworth-
Heinemann, London, UK, 1990, pp. 153–184.
[2] A. Bonazza, P. Messina, C. Sabbioni, C.M. Grossi, P. Brimblecombe, Mapping the
impact of climate change on surface recession of carbonate buildings in Europe,
Sci. Total Environ. 407 (2009) 2039–2050.
[3] C.M. Grossi, P. Brimblecombe, B. Menéndez, D. Benavente, I. Harris, M. Déqué,
Climatology of salt transitions and implications for stone weathering, Sci. Total
Environ. 409 (2011) 2577–2585.
[4] M. Schaap, M. Van Loon, H.M. Ten Brink, F.J. Dentener, H.P.J. Builtjes, The
nitrate aerosol field over Europe: simulations with an atmospheric chemistry-
transport model of intermediate complexity, Atmos. Chem. Phys. Discuss. 3
(2003) 5919–5976.
[5] H. Cachier, R. Sarda-Este‘ve, K. Oikonomou, J. Sciare, A. Bonazza, C. Sabbioni,
M. Greco, C. Saiz-Jimenez, A. Hermosin, J. Reyes, Aerosol characterisation and
sources in different European urban atmospheres: Paris, Seville, Florence and
Milan, in: C. Saiz-Jimenez (Ed.), Air Pollution and Cultural Heritage, Lisse,
Balkema, The Netherlands, 2004, pp. 3–14.
[6] A. Bonazza, C. Sabbioni, N. Ghedini, Quantitative data on carbon fractions in
interpretation of black crusts and soiling on European built heritage, Atmos.
Environ. 39 (14) (2005) 2607–2618.
[7] C.M. Grossi, A. Bonazza, P. Brimblecombe, I. Harris, C. Sabbioni, Predicting
twenty-first century recession of architectural limestone in European cities,
Environ. Geol. 56 (2–3) (2008) 455–461.
[8] G.S. Tyndall, A.R. Ravishankara, Atmospheric oxidation of reduced sulfur
species, Int. J. Chem. Kinet. 23 (1991) 483–527.
[9] J.R. Smyth, T.C. McCormick, Crystallographic data for minerals, in: T.J. Ahrens
(Ed.), Mineral physics & crystallography: a handbook of physical constants,
American Geophysical Union, Washington, USA, 1995, pp. 1–17.
[10] V. Comite, M. Álvarez de Buergo, D. Barca, C.M. Belfiore, A. Bonazza, M.F. La
Russa, A. Pezzino, L. Randazzo, S.A. Ruffolo, Damage monitoring on carbonate
stones: Field exposure tests contributing to pollution impact evaluation in two
Italian sites, Constr. Build. Mater. 152 (2017) 907–922.
[11] C. Cardell-Fernández, G. Vleugels, K. Torfs, R. Van Grieken, The processes dom-
inating Ca dissolution of limestone when exposed to ambient atmospheric
conditions as determined by comparing dissolution models, Environ. Geol. 43
(2002) 160–171.
[12] E. Marengo, E. Robotti, M.C. Liparota, M.C. Gennaro, Monitoring of pigmented
and wooden surfaces in accelerated ageing processes by FT-Raman spec-
troscopy and multivariate control charts, Talanta. 63 (2004) 987–1002.
[13] E. Angelini, S. Grassini, D. Mombello, A. Neri, M. Parvis, An imaging approach
for a contactless monitoring of the conservation state of metallic works of art,
Appl. Phys. A. 100 (2010) 919–925.
[14] A. Bernardi, F. Becherini, A. Bonazza, B. Krupinska, L. Pockelè, R. Van Grieken, S.
De Grandi, I. Ozga, A.J. Veiga Rico, O. Garcia Mercero, A. Vivarelli, A methodol-
ogy to monitor the pollution impact on historic buildings surfaces: the TeACH
project, in: M. Ioannides, al. et (Eds.), Progress in cultural heritage preservation,
Springer, Lymassol, Cyprus, 2012, pp. 765–775.
[15] E. Grinzato, P.G. Bison, S. Marinetti, Monitoring of ancient buildings by the
thermal method, J. Cultur. Heritage. 3 (1) (2002) 21–29.
[16] S. Kahraman, T. Yeken, Determination of physical properties of carbonate rocks
from P-wave velocity, Bull. Eng. Geol. Environ. 67 (2008) 277–281.
[17] V. Raimondi, G. Cecchi, L. Pantani, R. Chiari, Fluorescence lidar monitoring of
historic buildings, Appl. Opt. 37 (1998) 1089–1098.
 A. Suzuki et al. / Journal of Cultural Heritage 32 (2018) 30–37
[18] M. Bacci, A. Casini, M. Picollo, B. Radicati, L. Stefani, Integrated non-invasive
technologies for the diagnosis and conservation of the cultural heritage, J. Neu-
tron Res. 14 (1) (2006) 11–16.
[19] S. Vettori, M. Benvenuti, M. Camaiti, L. Chiarantini, P. Costagliola, S. Moretti,
E. Pecchioni, Assessment of the deterioration status of historical buildings by
hyperspectral imaging techniques, in: P. Tiano, C. Pardini (Eds.), In situ moni-
toring of monumental surfaces, Edifir, Firenze, Italy, 2008, pp. 55–64.
[20] M. Camaiti, S. Vettori, M. Benvenuti, L. Chiarantini, P. Costagliola, F. Di
Benedetto, S. Moretti, F. Paba, E. Pecchioni, Hyperspectral sensor for gypsum
detection on monumental buildings, J. Geophys. Eng. 8 (2011) 126–131.
[21] G. Capobianco, S. Serranti, F. Prestileo, S. Serranti, G. Bonifazi, Hyperspectral
imaging-based approach for the in-situ characterization of ancient Roman wall
paintings, Periodico di Mineralogia. 84 (3A) (2015) 407–418.
[22] R.N. Clark, Reflectance Spectra, in: T.J. Ahrens (Ed.), Rock physics & phase rela-
tions: a handbook of physical constants, USA, American Geophysical Union,
Washington, 1995, pp. 178–188.
[23] E. Ben-Dor, K. Patin, A. Banin, A. Karnieli, Mapping of several soil properties
using DAIS-7915 hyperspectral scanner data – a case study over clayey soils in
Israel, Int. J. Remote Sens. 23 (2002) 1043–1062.
37
[24] ASD, Technical guide, 4th Edition., Analytical Spectral Devices Inc., Boulder, CO,
USA, 2007.
[25] S.J. Gaffey, Spectral reflectance of-carbonate minerals in the visible and near
infrared (0.35-2.55 microns): calcite, aragonite, and dolomite, Am. Mineral. 71
(1986) 151–162.
[26] G.R. Hunt, J.W. Salisbury, Visible and near-infrared spectra of minerals and
rocks: II. Carbonates, Modern Geology 2 (1971) 23–30.
[27] E.A. Cloutis, F.C. Hawthorne, S.A. Mertzman, K. Krenn, M.A. Craig, D. Marcino,
M. Methot, J. Strong, J.F. Mustard, D.L. Blaney, J.F. Bell III, F. Vilas, Detection and
discrimination of sulfate minerals using reflectance spectroscopy, Icarus 184
(2006) 121–157.
[28] J.L. Bishop, M.D. Lane, M.D. Dyar, S.J. King, A.J. Brown, G.A. Swayze, Spectral
properties of Ca-sulfates: gypsum, bassanite, and anhydrite, Am. Mineral. 99
(2014) 2105–2115.
[29] S. Siegesmund, H. Durrast, Physical and Mechanical properties of rocks, in: S.
Siegesmund, R. Snethlage (Eds.), Stone in Architecture. Properties, durability,
4th Edition, Springer, Berlin, Germany, 2011, pp. 97–225.