Chemical imaging of historical mortars by Raman microscopy

Construction and Building Materials 114 (2016) 506–516

Contents lists available at ScienceDirect

Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat

Chemical imaging of historical mortars by Raman microscopy Thomas Schmid a,⇑, Petra Dariz b a b

Federal Institute for Materials Research and Testing, Richard-Willstätter-Str. 11, 12489 Berlin, Germany Bern University of the Arts, Conservation-Restoration, Fellerstr. 11, 3027 Bern, Switzerland

h i g h l i g h t s

g r a p h i c a l a b s t r a c t


Chemical images with approx.

500 nm resolution were acquired by Raman microscopy.
Clinker remnants in 19th-century Roman and Portland cement stone were analysed.
Gypsum mortars from the Middle Ages and early 20th century were analysed as well.
Alite, tricalciumaluminate, wollastonite, gypsum, and anhydrite were identified.
Dedolomitisation of accessory mineral grains in gypsum mortar was visualised.

a r t i c l e

i n f o

Article history: Received 27 November 2015 Received in revised form 10 March 2016 Accepted 23 March 2016

Keywords: Raman microscopy Chemical imaging Cement clinker Stucco Gypsum Anhydrite Dedolomite

a b s t r a c t Raman microspectroscopic imaging was just recently introduced into the analysis of cement stone. Here, we demonstrate this approach on 19th-century Roman and Portland cement mortars and extend it to gypsum-based samples originating from a medieval stucco sculpture (high-burnt gypsum) and a stucco ornament prefabricated at the beginning of the 20th century (plaster of Paris). Furthermore, the distributions of dolomite and calcite were mapped in an accessory mineral grain with approx. 500 nm lateral resolution demonstrating the ability for studying alteration processes such as dedolomitisation. As we would like to make this approach accessible to other researchers, we discuss its present status, advantages, limitations and pitfalls. Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction Very generally speaking, mortars are made by mixing binder, aggregates and water. From the chemical point of view, binder and aggregate are umbrella terms for a chemically highly heterogeneous variety of materials, some of them, such as hydraulic binders, consisting of complex mixtures of inorganic phases. The ⇑ Corresponding author. E-mail address: [email protected] (T. Schmid). http://dx.doi.org/10.1016/j.conbuildmat.2016.03.153 0950-0618/Ó 2016 Elsevier Ltd. All rights reserved.

chemical analysis of such complex materials can only be adequately handled by employing an arsenal of analytical methods yielding complementary information, which in combination provides insight into, for example, the type of binder, composition and provenience of the raw materials as well as heating and cooling regime during the fabrication. Historical mortars are characterised by tremendously more heterogeneous compositions than the standardised building materials employed nowadays. The use of marlstone, a natural mixture of lime and clay, for the burning of Roman cement – the first highly hydraulic

T. Schmid, P. Dariz / Construction and Building Materials 114 (2016) 506–516

binder – intrinsically lead to varying chemical compositions of the end product, depending on the properties of the stone pit and the kiln conditions. Also the early Portland cements produced from well-defined synthetic mixtures of lime and clay in the course of the 19th century, differ from modern ordinary Portland cement (OPC), because of the lower (sub-optimal) and much more heterogeneously distributed calcination temperatures reached in the shaft kilns used at that time. Many phases found in 19th-century Portland cement occur just as intermediate reaction products in the current burning process carried out in rotary kilns, disseminated in Europe not before the turn of the century. From the chemical point of view or, more precisely, in terms of elemental composition, historical gypsum-based mortars are seemingly more simple and homogeneous, but a closer look reveals a plethora of dehydration and hydration stages, ranging from medieval high-burnt to 19thcentury binder produced at much lower temperatures. Correct determination of the type of mortar preserved in historic masonry is key for the manufacture of compatible restoration materials. So far, mainly light microscopy (including polarised light microscopy), scanning electron microscopy (SEM) – optionally combined with energy-dispersive X-ray spectroscopy (EDX) – and X-ray diffraction (XRD) were employed in the morphological and chemical analysis of historical building materials. Raman spectroscopy – an additional analytical approach – is based on inelastic scattering of light by molecules, crystals or amorphous phases and is typically performed with visible or near-infrared lasers as excitation sources. Inelastic scattering generates photons with frequencies (i.e., colours), which are slightly shifted from the laser frequency. The frequency differences (expressed in wavenumbers, unit: cm1; also termed ‘Raman shift’) equal the frequencies of molecular or crystal lattice vibrations [1]. The set of all Raman-active vibrations (i.e., the Raman spectrum) depends on the masses of the involved atoms, their distances and spatial arrangements, and is therefore a very characteristic ‘‘fingerprint” of a molecular or crystal structure. Therefore, chemical compounds (including their polymorphic forms) can be unambiguously identified by a fingerprint comparison of sample spectra with reference spectra in terms of band positions and shapes. Indeed, the ‘‘suspects” and their fingerprints have to be known in advance and thus, as described below, collection of a spectral database library of relevant phases is key for the application of this technique to building materials and part of our ongoing research. Extensive work in the determination of clinker minerals as well as hydration and carbonation products in cementitious materials was performed by Black et al.; a comprehensive – not yet complete – collection of reference spectra can be found in Ref. [2]. Examples of applications of Raman spectroscopy in the field of cultural heritage research, concerning architectural surfaces and wall paintings, are published in Refs. [3–7]. Raman microspectroscopic imaging (also termed ‘Raman microscopy’) is typically realised by step-wise movement of a sample through the laser focus employed for excitation [8]. In every step – typically ranging from several micrometres down to sub-micrometre size – the full Raman spectrum is collected, enabling the calculation of distribution maps of sample constituents in the form of colour-coded intensities of marker bands. Only few examples of Raman microscopic imaging in the analysis of building materials can be found in the literature so far, for example, the study of salts in concrete [9,10]. Examples for applications in cultural heritage research were given by Ropret et al. [11], Jallad et al. [12], Veneranda et al. [4] and Antunes et al. [13]. Recently, we introduced chemical imaging by Raman microscopy to the study of historic cement stone and proved this to be a suitable approach for the determination of the composition and for the visualisation of the distribution of the mineral phases [14,15]. The method, adept to overcome some of the

507

limitations of the microscopic techniques or the pointspectroscopic sample characterisation used so far, provides the spatial distribution of mineral (including polymorphic) and amorphous phases, as well as crystal orientations with a sublm spatial resolution. It therefore adds chemical information to light microscopy images, provides phase information that is complementary to the information on purely elemental composition provided by EDX and electron microprobe measurements and extends the phase information, which is also provided by XRD (performed as a bulk analysis method) with spatial resolution. In contrast to XRD also amorphous phases can be studied by Raman microscopy. For example, amorphous carbon was previously detected in granulated blast furnace slag, found as aggregate in 19th-century cement stone [14]. Whereas the spectra of diamond and graphite are dominated by the well-known sharp bands at 1330 cm1 (D band) and 1580 cm1 (G band), respectively, disordered and amorphous carbon gives rise to both, D and G bands appearing strongly broadened and overlapping, but still characteristic enough to be specifically assigned to that material [16]. Furthermore, mineral components in mortars present only in amounts below the detection limit of XRD (e.g., c-Ca2SiO4 in Roman cement [14]) can be detected by Raman microscopy because of the micrometre-sized measurement volume of this spatially-resolved analytical approach. In contrast, individual mineral grains and in some cases not even individual remnant clinker grains can be resolved with micro-XRD instruments available on the market with typical spatial resolutions of 50 lm up to some 100 s of micrometres [17–19]. Only by using synchrotron radiation, the resolution can be improved to the lower micrometre range [20,21]. Infrared microspectroscopy, closely related and complementary to Raman microscopy, was already successfully employed for the chemical imaging of historical lime mortars [22] and for the characterisation of compounds and degradation products in paint layers [23–26]. Due to the use of infrared radiation (instead of visible light), the lateral resolution is on the order of micrometres to tens of micrometres due to the optical diffraction limit. Moreover, with a measurement range typically limited to 400–4000 cm1 crystal lattice vibrations of inorganic phases (mortar binder, mineral aggregates, inorganic pigments, etc.) occurring at lower wavenumbers cannot be detected, whereas commercial Raman spectrometers provide wavenumber ranges typically starting at P100 cm1 and the use of special notch filters offers access to even lower vibrational frequencies. For these reasons, we expect that Raman microscopy in the future can valuably extend the arsenal of techniques applied in the analysis of both historic and contemporary building materials. In this communication we demonstrate its application to two types of historic cement binders and extend our approach to gypsum based mortars, i.e. medieval high-burnt gypsum and stucco from the 19th century, demonstrating the capability to map distributions of different hydration stages of calcium sulphate. Furthermore, slight changes in band positions enabling to distinguish dolomite and calcite provide the basis for an imaging study of dedolomitisation in an accessory mineral grain. A central idea of this communication is to make researchers from the field of construction and building materials aware of this relatively new analytical approach and its actual status and capabilities. Therefore, we show how Raman microscopy can be used to add chemical information to optical microscopy images by demonstrating the tracking of the same features by both approaches in a remnant Roman cement clinker grain including the mineral wollastonite, but also discuss typical limitations and pitfalls of this approach – such as interfering fluorescence and Raman signals from embedding resins – as well as strategies to overcome them.

508

T. Schmid, P. Dariz / Construction and Building Materials 114 (2016) 506–516

2. Materials and methods 2.1. Samples Samples were taken from the façade of the Merkurhaus (Horw/Lucerne, Switzerland), the tympanum of the former printing office Schläpfer (Weinfelden, Switzerland), the pieta sculpture of the Benedictine abbey Marienberg (Burgeis, Italy) and a prefabricated stucco ornament in the Auditorium maximum of ETH Zurich (Switzerland). In a specialised lab, they were embedded into resins and cut into thin-sectional samples, which were mounted onto conventional glass slides without coverslips on top. This is a well-known preparation method for light and electron microscopy samples, which are also compatible with Raman microscopy.

2.2. Raman microscopy The Raman measurements were performed using a LabRam HR 800 instrument (Horiba Jobin Yvon, Bensheim, Germany) coupled to an upright BX41 microscope (Olympus, Hamburg, Germany). The instrument can be used for sample observation through ocular lenses by transmitted light (or, in the case of opaque samples, reflected light) brightfield illumination (here, generally termed white-light microscopy). Alternatively, white-light microscopy images can be acquired using a charge-coupled device (CCD) camera. These images are shown together with the corresponding Raman results in Figs. 1b, 2a, and 3b. In case of the clinker remnant from Roman cement, we show an additional overview image with less magnification in Fig. 1a, which was acquired using an Eclipse E600 microscope (Nikon Instruments, Amsterdam, Netherlands). For excitation of Raman spectra of selected microscopic sample spots, a HeNe laser having a wavelength of k = 632.8 nm (red) was employed. The laser power reaching the sample can be precisely controlled by (software-controlled) insertion of neutral density filters into the beam path. In the measurements discussed here, 1 mW up to 10 mW of laser power was focused onto the samples by using a 50x/N. A. = 0.75 objective lens. As thermal damage by laser irradiation mainly depends on the actual light absorption coefficient and heat dissipation properties of a sample, the laser power should be chosen with care for each individual type of samples. According to the Rayleigh criterion [27], with a numerical aperture (N.A.) of 0.75, the lateral resolution of the microscope is approx. 520 nm at k = 632.8 nm. The diameter of the focal spot is approximately twice this value [28]. The same objective lens was used for both, excitation and collection. The collected light was dispersed by a grating with 300 grooves/mm and detected by a liquid-nitrogen cooled CCD detector (operating temperature: 126 °C) with 1024  256 pixels leading to a spectral resolution of approx. 2 cm1/pixel at 632.8 nm. The acquisition time per spectrum within this study typically ranged from 1 s up to 10 s. For better comparability, most spectra shown here were stacked by adding arbitrary offset values. Therefore no Raman intensity (y axis) scale is given in the figures. When necessary, spectra were additionally scaled by multiplication with a factor, which is mentioned next to the according traces. Moderate smoothing by calculating the moving average over 3 adjacent data points in all spectra was performed. No baseline correction was applied, except for the 3 spectra shown in Fig. 2c. Raman images (Raman maps) were acquired by step-wise movement of the sample through the laser focus by a software-controlled sample-scanning stage and acquisition of the full Raman spectrum at every pixel. Thus, a Ramanspectroscopic mapping experiment yields a data cube that is defined by the x and y coordinates of the scanned sample area and the wavenumber axis of the Raman spectra, in which each element contains the Raman intensity at a specific sample position and wavenumber. We used home-made (T.S.) Labview-based software (National Instruments, Austin, TX, USA) to extract distributions of intensities of selected marker bands, leading to spectroscopic images (also termed ‘‘chemical images”) of the sample reflecting the distributions of its chemical constituents. As will be shown below, fluorescence emission from the embedding medium often caused an irregularly shaped baseline (i.e., a part of a fluorescence band, which is spectrally much broader than Raman signals) with intensities varying over the mapped areas. In order to avoid cross-talk between Raman and fluorescence signal intensities, a background correction for each evaluated Raman band was performed. Therefore, in our Labview-based data analysis software, the end points of a linear baseline were defined on the left and right of every evaluated band and the baseline was subtracted before determining the peak intensity (here, simply as peak height, not integral). As in every pixel the full Raman spectrum is collected and a map can easily consist of hundreds to thousands of pixels, the acquisition time per pixel (or spectrum, respectively) in mapping experiments typically was kept as short as 1 s. Only for the dolomite map shown in Fig. 3b and the Portland cement clinker measurements shown in Fig. 2b, the acquisitions times were 2 s and 4 s, respectively. Measurement times for whole maps ranged from 10.5 min (Fig. 1c, 25  25 = 625 pixels, 1s acquisition time) up to 9.5 h (Fig. 3b, 113  151 = 17,063 pixels, 2 s acquisition time) in a software-controlled overnight experiment.

Fig. 1. Optical and chemical images of a preserved remnant of a 19th-century Roman cement clinker grain: (a, b) Optical microscopy images with the analysed area marked with red boxes, (c) Raman microscopy map showing the distributions of wollastonite and embedding medium, and (d) exemplary Raman spectra from the microspectroscopic map, which can be assigned to wollastonite (lower trace) and fluorescence emission from the embedding medium (upper trace). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

T. Schmid, P. Dariz / Construction and Building Materials 114 (2016) 506–516

509

Fig. 2. Analysis of a remnant clinker grain in a Portland cement stone sample from the 19th century: (a) Optical microscopy image, (b) Raman microscopy map, and (c) exemplary spectra from the map containing Raman signatures of alite (Ca3SiO5 or C3S, respectively), tricalciumaluminate (Ca3Al2O6 or C3A, respectively), and presumably a calcium aluminoferrite phase with so far unknown Raman spectrum (CxAyFz).

3. Results and discussion 3.1. Characterisation of not hydrated clinker remnants in a historical Roman cement mortar Roman cement (patented by James Parker, 1796) was produced by firing marlstones below the sintering temperature, i.e. between

Fig. 3. Dedolomitisation of an accessory mineral grain in a medieval gypsum-based mortar: (a) Optical microscopy image, (b) zoning of the grain visualised by Raman microscopy, (c) spectra with the highest intensities of the 1098-cm1 dolomite (lower trace) and the 1086-cm1 calcite band (upper trace) observed on the mapped area.

510

T. Schmid, P. Dariz / Construction and Building Materials 114 (2016) 506–516

800 and 1100 °C – this in contrast to ordinary Portland cement (patented by Joseph Aspdin in 1824; vitrification attained by his son William Aspdin around 1848), a mixture of limestone and clay heated to the point of liquefaction for sintering the raw materials into clinker. The grain size of 19th-century cements is relatively coarse due to the limited technological equipment for grinding, too coarse for an effective hydration reaction. Preserved remnants of clinker grains in the hydrated and aged mortar matrix allow the identification of the binder of historical samples and beyond that, the analytical discrimination of Roman and Portland cements by microscopic and spectroscopic techniques. Roman cement mortar was used for the manufacture of the cast building ornaments mounted at the façade of the Merkurhaus in Horw (Switzerland), constructed in 1892/93. Fig. 1 shows a clinker grain relict with wollastonite laths embedded in the hydrated binder matrix and the corresponding Raman mapping. The calcium inosilicate mineral (CaSiO3 or CS in cement chemist’s notation, respectively) was classified as a typical compound of overfired Roman cement phenograins in the course of the research project ROCEM (Roman Cement to Restore Built Heritage Effectively, 5th Framework Programme, European Commission); it shows no hydraulic activity [29,30]. The light microscopy images Figs. 1a and 1b exhibit the typical columnar habitus of wollastonite crystals, whose chemical identity is confirmed by the Raman map in Fig. 1c, in which the distribution of the background-corrected intensity of the most prominent Raman band of wollastonite at 987 cm1 is plotted in red. All red pixels contain a Raman spectrum, which as a whole can be undoubtedly assigned to wollastonite. Fig. 1d displays one representative spectrum showing the typical bands at 375 cm1, 581 cm1 and 987 cm1, which perfectly match reference spectra of wollastonite in both, band positions and band shapes (see the wollastonite spectrum in the RRUFF spectral database [31], RRUFF ID 120122 or in Ref. [32]). Only the comparison of the whole spectral fingerprint with a reference allows the unambiguous identification of mineral phases. Once a phase is identified, a marker band can be selected, which in the ideal case is strong and does not spectrally overlap with bands of other components of the same sample. A map of the intensity of the marker band corresponds to the distribution of this specific compound. Indeed, we propose to check for resemblance of the reference for all spectra in a map, which were assigned to a certain compound. Another test of the reliability of the data interpretation is to plot distribution maps of all typical bands of a compound. In this case, also maps of the 375-cm1 and 581-cm1 band intensities resulted in the same lath shapes (data not shown), confirming the assignment of the red pixels in Fig. 1c to the inosilicate wollastonite. The green pixels represent the spectral intensity at 2100 cm1, in this case without background correction. A typical spectrum from the green area in Fig. 1c is shown as the upper trace in Fig. 1d, revealing the rising edge of a – compared to the Raman signals – broad fluorescence band. As fluorescence occurs on a different spectroscopic scale, the band position is more correctly expressed in wavelength rather than wavenumbers. With the employed laser wavelength of 632.8 nm (equal to 0 cm1 Raman shift), the observed wavenumber of 2100 cm1 equals a fluorescence emission wavelength of approx. 725 nm. Fluorescence emission from the embedding medium used for sample preparation (Epo-Tek 301, Epoxy Technology Inc., John P. Kummer AG, Cham, Switzerland) is a typical spectral interference we encountered in measurements of polished cross-section and thin-section samples. Thus, often background correction is necessary in the evaluation of Raman band intensities to avoid cross-talk from fluorescence emission of the embedding resin. In some cases, strong fluorescence emission overwhelms the Raman bands and hampers their evaluation. The upper trace in Fig. 1d is a good example for this.

Obviously the porosity of the sample material around the wollastonite crystals enabled the embedding resin to strongly penetrate into that area, making Raman measurements impossible, even though we expect the presence of Raman-active clinker minerals (possibly iron-containing phases, according to their colour seen in Fig. 1a) in that area. Interestingly, the distribution of the fluorescence intensity in Fig. 1c shows a stripe pattern, which is also visible in the white-light images (Fig. 1a and b), underlining that Raman microscopic imaging adds chemical information to the optical image of a sample. This typical limitation of Raman spectroscopy can be overcome by, for example, trying other laser wavelengths for excitation. As explained above, Raman bands always occur close to the excitation laser wavelength, shifted by just some wavenumbers (i.e., the Raman shift). In contrast, fluorescence emission of a certain fluorophore (e.g., a certain embedding resin) always takes place at a characteristic emission wavelength. Therefore, switching to another laser wavelength shifts the Raman spectrum to a different wavelength range, offering the chance to measure the same Raman shifts with less interference from fluorescence. Fluorescence emission from organic compounds, such as embedding resins, can often be reduced by employing longerwavelength lasers (towards or in the near-infrared spectral range), but cement is known to emit fluorescence in the near-infrared spectral range hampering the collection of Raman spectra, which can be detected at 632.8 nm in the red range without interference from fluorescence [33]. More efficient would be the use of nonfluorescing mounting media. It is part of our ongoing research to study the fluorescence emissions and Raman spectra of typical mounting media used in biology, pathology [34] as well as for the preparation of polished cross-sections and thin-sections of construction materials. According to our current experience, synthetic resins used for embedding do not necessarily fluoresce in the visible range if their spectra are collected from fresh and pure samples, because yellowing and fluorescence emission are often linked to ageing and contamination. Thus, the search for an ‘ideal’ embedding medium has to include studies on ageing and long-term stability. 3.2. Characterisation of not hydrated clinker remnants in a historical Portland cement mortar Fig. 2 displays the Raman map of a remnant Portland cement clinker grain found in a sample taken from the tympanum of the former printing office Schläpfer in Weinfelden (Switzerland). As previously shown for not hydrated Roman cement clinker nodules in aged cement stone [14], Raman spectroscopy enables the analysis of their phase composition, representing a local snapshot of the burning and cooling history undergone by the feedstock in the shaft kiln. The identification of marker phases like the tricalcium silicate alite (Ca3SiO5 or C3S in cement chemist’s notation, respectively), a widely recognised indicator for Portland cement [35–37], allows the determination of the type of cement used in the mortar mixture. In another study, we introduced the calcium aluminates as additional marker phases detectable by Raman microscopy for a more precise determination of the different variants of 19thcentury cement, including the differentiation between ‘modern’ ordinary Portland cement and its predecessor meso Portland cement [15]. Tricalcium aluminate (Ca3Al2O6 or C3A, respectively) can be observed only in hydraulic binders burnt like nowadays in a rotary kiln, while the lower temperatures reached in 19thcentury shaft kilns yield only the calcium aluminates CaAl2O4 (CA) and mayenite (Ca12Al14O33 or C12A7, respectively), currently just occurring as intermediate products. In the Weinfelden sample, C3A was identified based on the detection of the two strongest Raman bands of this compound at 508 cm1 and 756 cm1 [2]. The 508-cm1 mode was used for mapping, because of strong

T. Schmid, P. Dariz / Construction and Building Materials 114 (2016) 506–516

spectral overlap of the 756-cm1 mode with Raman signals from the third compound detected in this sample area. The top trace in Fig. 2c reveals very strong, broad and overlapping Raman bands with peaks at 671 cm1 and 724 cm1 (for better comparability this very strong spectrum was scaled by a factor of 0.25). Furthermore, this Raman spectrum contains weak and overlapping contributions at 270 cm1 and 314 cm1. As there is no reference spectrum resembling this signature available in the literature or our own database library, we just tentatively assign it to a calcium aluminoferrite phase CxAyFz (or xCaOyAl2O3zFe2O3, respectively). This raises another issue. As mentioned above, for performing a fingerprint comparison, the suspects and their fingerprints (or spectra, respectively) need to be known in advance. Thus, building up of a spectral database of relevant phases found in construction materials, is the key for the complete phase analysis of such samples. We showed in another study that the calcium aluminoferrites C4AF, C6AF2 and C6A2F, often summarised under the term ‘brownmillerite’, which as a mineral name specifically terms the aluminoferrite C4AF (Ca4Al2Fe2O10 or Ca2AlFeO5, respectively), can be clearly distinguished based on their Raman spectra [15]. These spectra typically show a strong and broad band appearing at a wavenumber position ranging from approx. 730 cm1 to 760 cm1, which is specific for one of the mentioned stoichiometries, and weaker contributions of two bands in the range between 260 cm1 and 320 cm1. The latter just slightly shift depending on the stoichiometry and are typical for all calcium aluminoferrites we studied so far. (For previously published spectra of C4AF, C6AF2 and C6A2F, for example in Ref. [32], only the range around the strongest Raman band was shown, and its width and illdefined position did not allow to use them as reference spectra for the identification of unknown phases in samples of historical cement stone). The presence of both, a doublet in the range between 260 cm1 and 320 cm1 and the main peak at 724 cm1, close to the position of the most prominent band of C6AF2 at 732 cm1, lead us to the assignment of this unknown spectrum to a calcium aluminoferrite phase. The different band position might be explained by another stoichiometry of the phase, and the second peak at 671 cm1 might be due to the presence of further chemical elements in that compound. Extension of our own spectral database by Raman measurements of pure clinker phases, which are synthesised for this purpose and whose identity is independently confirmed by XRD, is an important part of our ongoing research [15]. We would like to point out that all the cement stone samples were taken close to the surface. Therefore, the not hydrated clinker relicts are embedded in a fully carbonated matrix. Nevertheless, as (single-point) spectra are already described in the literature, Raman microspectroscopy has also the potential for the identification of products of the hydration reaction (for Raman spectra of CS-H and C-A-H see Refs. [2,8,32,33]). 3.3. Characterisation of a medieval gypsum-based mortar In correlation with local deposits, burnt and powdered gypsum was used as binder for masonry and joint mortars or stuccowork in Central Europe in the Early and High Middle Ages [38,39]. The monastery Marienberg, a Benedictine abbey in South Tyrol (Italy), houses a stucco sculpture of a pieta dated around 1420. Thin sections of the gypsum-anhydrite mortar reveal the presence of accessory and impurity minerals like dolomite (CaMg(CO3)2), celestine (SrSO4) and quartz (SiO2) in the binder matrix [40]. A Raman mapping of a brownish aggregate of dolomite is shown in Fig. 3. The zoning of the grain (Fig. 3a) is due to the partial pseudomorphic replacement of dolomite by calcite (CaCO3), an alteration phenomena most probably driven by the dissolution of the anhydrite phases in the mixing water, which forces calcium carbonate

511

to precipitate because of the common-ion effect [41,42] (Dedolomitisation has also been explained by the infiltration of calcium-rich water or the oxidation of ferrous iron in dolomite; see, for example, Refs. [43,44]). The minerals calcite (and its polymorphs aragonite and vaterite, respectively) and dolomite are widely studied by Raman spectroscopy [33,45–47]. The main band in the Raman spectra allows the differentiation between calcium and magnesium carbonates; it occurs at 1086 cm1 for calcite, while the wavenumber position of the symmetric carbonate stretching mode is shifted to 1098 cm1 in the case of dolomite [48–50]. Due to the relative shift of only 12 cm1, the two bands overlap, but with the spectral resolution of approx. 2 cm1 applied here, they can clearly be distinguished (Fig. 3c) and their intensities independently mapped (Fig. 3b). For a correct interpretation of the Raman map, we would like to point out that, like in the other chemical images shown here, the colour scales are defined according to the maximum and minimum intensities of each individual Raman band. So, for example, the strongest green colour is correlated with the highest intensity of the band at 1086 cm1 independent of its relation to the (in this case typically higher) intensities of the 1098-cm1 bands detected on this sample area. As generally different chemical compounds show different Raman activities, the band intensities of different phases cannot be quantitatively compared with each other. Absolute quantification can only be achieved by performing calibrations for each sample component individually. Nevertheless, in this special case of two chemically very similar phases having their most prominent peaks (evaluated for calculating the Raman map in Fig. 3b) at very similar spectral positions, we can add some semiquantitative interpretation of this data. Independent of the impression Fig. 3b provides at first glance, dolomite (marked red) was detected all over the investigated sample area, even though with varying Raman intensities. Calcite (marked green) was found only co-localised with dolomite, or more precisely, appeared colocalised within the spatial resolution of the microscope of approx. 500 nm. The two traces shown in Fig. 3c represent the spectra with the highest intensities of the main dolomite band (lower trace) and calcite band (upper trace), respectively, showing that even at the pixel with the strongest calcite signature, the 1086-cm1 calcite band is just as strong as the 1098-cm1 mode of dolomite, whereas on many red pixels, spectra purely attributable to dolomite were acquired. We therefore interpret the detail of the aggregate grain chosen for the mapping experiment to predominantly consist of dolomite crystals with some parts additionally containing calcite or recalcified dolomite, respectively. On the geological scale, dissolution and replacement of dolomite crystals take place from the core outwards [51]. Fig. 4 displays a Raman mapping of a sample of the same sculpture, in this case acquired within the stucco matrix. Similar to the minerals calcite and dolomite discussed above, the different phases in the system CaSO4–H2O can be distinguished based on their slightly shifted most prominent Raman bands: the symmetric sulphate stretching modes of gypsum and of the high-temperature modifications of anhydrite occur at 1008 cm1 and 1017 cm1, respectively [13,52,53]. Dehydration of gypsum (CaSO42H2O) by heating, firstly yields bassanite (hemihydrate, CaSO40.5H2O), existing in three polymorphic forms, termed a, b, and b0 . Further heating above temperatures of 100 °C leads to the formation of anhydrite III (CaSO4), more precisely also termed water-free bassanite, which in analogy to its water-containing counterpart is known as a, b, and b0 modifications. Gypsum exposed to temperatures above 200 °C yields anhydrite II, which in turn is converted into anhydrite I above 1180 °C [54–57]. (Note that different nomenclatures especially for the various anhydrite phases can be found in the literature.) Within anhydrite II three sub-types having different rate constants in their reaction with the mixing water are

512

T. Schmid, P. Dariz / Construction and Building Materials 114 (2016) 506–516

Fig. 4. Raman map containing two anhydrite grains in the lower left part and one gypsum grain in the upper right part of the mapped area of a medieval gypsum-based mortar sample: (a) Intensity distribution of the most prominent Raman band of anhydrite at 1017 cm1, (b) intensity distribution of another anhydrite band at 1129 cm1, (c) intensity distribution of the main band of gypsum at 1008 cm1, (d) overlay of all intensity distributions, (e) Raman spectra revealing the highest intensities of the bands at 1017 cm1 (1), 1129 cm1 (2), and 1008 cm1 (3) collected at the pixels marked in (a), (b), and (c). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

T. Schmid, P. Dariz / Construction and Building Materials 114 (2016) 506–516

known to exist as a function of burning temperature. Ignoring the complexity of these calcium sulphate binders, anhydrite II and III are often simplistically summarised as ‘thermo anhydrite’ [58,59]. In the mapped sample area shown in Fig. 4 both, gypsum and anhydrite are detected, with the major part consisting of anhydrite. As can be seen in spectrum (3) (Fig. 4e), even in case of the strongest gypsum signal detected here, the band at 1008 cm1 only appears as a shoulder of the 1017-cm1 anhydrite peak. Nevertheless, it can be evaluated independently. For calculating the Raman maps, we determined the intensities of three bands: the 1008cm1 band of gypsum (green, Fig. 4c), the 1017-cm1 band of anhydrite (red, Fig. 1a), and a band at 1029 cm1 (blue, Fig. 4b). The latter is the second-strongest band of anhydrite. As previously shown for remnant crystals of the cement clinker mineral b-Ca2SiO4 (belite, b-C2S) in historical Roman cement stone, relative intensities of some bands of the same phase significantly change depending on the local crystal orientation [14]. This effect can be explained by the efficiency of the excitation of crystal vibrations, which depends on the orientation of that vibration relative to the polarisation direction of the exciting laser beam (i.e., the plane in space in which the electric part of the electromagnetic wave oscillates). Due to their different orientations, individual crystals in a material can be visualised in Raman imaging due to their different signal intensities [60]. This effect can be seen in the individual maps in Fig. 4a and b as well as in the overlay shown in Fig. 4d. The whole lower left part of the map predominantly consists of anhydrite, in which the left part shows a relatively high intensity of the 1029-cm1 mode (with the highest intensity at the pixel marked with the cross, where spectrum (2) was acquired) and therefore appearing blueish in the overlay (Fig. 4d), whereas the intensity of that band is lower in the right part, appearing purple in the overlay. Another interface identified on this map separates anhydrite (red and blue) from gypsum (green). Thus, three individual grains of anhydrite and gypsum, respectively, can be found on this map. According to the Raman spectra of calcium sulphate phases presented by Prieto-Taboada et al. [52], our anhydrite spectra can be assigned to anhydrite II, which can be spectroscopically distinguished from the anhydrite types III (water-free bassanite) and I. 3.4. Characterisation of a gypsum-based mortar dating from the 19th century The technological development of the grinding machinery is responsible for the small grain size of the gypsum-based binders

513

produced in the course of the 19th century. Composition, microstructure and porosity of binders and mortars, respectively, dating of this period differ significantly from the characteristics of medieval counterparts. The higher content of hemihydrate and the presence of different anhydrite phases in the former are due to the lower burning temperatures and the improved control of the burning conditions in the newly developed furnace systems. Remnant kiln run grains comprising finely crystalline anhydrite polymorphs or pseudomorphs of calcium sulphate dihydrate after fibrous ‘thermo anhydrite’ occur rather infrequently in the hydrated mortar mixture. Besides the mineralogical composition, the high porosity of the binder matrix formed by fibrous or acicular euhedral gypsum crystals allows to differentiate historicism gypsum or art nouveau gypsum plaster from older historical gypsum mortars, as repeated wet-dry cycles during ageing and long-term weathering respectively result in large, rounded gypsum crystals and a denser microstructure (dissolution and precipitation/ recrystallization of calcium sulphate dihydrate) [61–63]. Fig. 5 displays Raman spectra, which were recorded on four selected spots of a thin-sectional sample taken from a prefabricated ornamental strip in the Auditorium maximum in the main building of the Swiss Federal Institute of Technology (ETH) in Zurich (Switzerland). In contrast to the spectra acquired in mapping experiments shown above, longer acquisition times of 5–10 s were employed providing higher signal-to-noise ratios and enabling to find further spectroscopic differences between gypsum and anhydrite. Furthermore, a much wider spectral range from 100 cm1 to 3800 cm1 was acquired by software-controlled movement of the spectrometer grating between 3 individual acquisitions that subsequently were combined to a single spectrum. This timeconsuming acquisition mode is typically avoided in mapping experiments. The range around 3500 cm1 contains the O–H stretching vibrations of water, which allow to further study the system CaSO4–H2O. Fig. 5 shows two excerpts from the full Raman spectra: the lowwavenumber range from 350 cm1 to 1250 cm1 and the highwavenumber range between 2800 cm1 and 3550 cm1. For better comparability, in this case the spectra were not only stacked by adding offset values but also scaled to the same intensity of the 1008-cm1 band of gypsum. The selected ranges contain several features that are characteristic for gypsum or anhydrite, respectively. The low-wavenumber range provides clear evidence that spectra (1)–(3) can be assigned to gypsum having its symmetric sulphate stretch vibration at 1008 cm1, whereas spectrum (4) can be best explained with co-localised gypsum and anhydrite

Fig. 5. Raman spectra collected on selected spots on a stucco sample from the early 20th century. Spectra (1)–(3) can be assigned to gypsum, whereas on spot (4) a mixture of anhydrite and gypsum was present. All bands marked with asterisks can be assigned to interfering Raman bands from the embedding medium.

514

T. Schmid, P. Dariz / Construction and Building Materials 114 (2016) 506–516

phases showing overlapping modes at 1008 cm1 (gypsum) and 1017 cm1 (anhydrite) [13,52,53]. A closer look reveals further differences between gypsum and anhydrite spectra. Some bands are present in all spectra (e.g., at 417 cm1 and 674 cm1), whereas others are specific for gypsum (e.g., 495 cm1, 624 cm1) or anhydrite (e.g., 499 cm1 and 608 cm1). Also the second-strongest mode of anhydrite at 1029 cm1 mentioned above (see Fig. 4 and its discussion) is shifted in the gypsum spectrum and appears at 1032 cm1. The high-wavenumber range provides evidence for the presence or absence of water-containing phases. In Fig. 5, all spectra have bands at approx. 3400 cm1 and 3490 cm1 in strongly varying intensities. Spectrum (4) shows the lowest Raman intensities in this range (especially when considering the signal-to-noise ratio), which is consistent with the additional presence of anhydrite on this spot of the gypsum material. We expect that in further investigations of (historic) construction materials this spectral range will allow us to clearly distinguish between gypsum (also according to the literature having two broad bands at approx. 3400 cm1 and 3490 cm1 [64]), bassanite (according to the literature showing a band at 3553 cm1 [52] or two bands at 3554 cm1 and 3610 cm1 [64]), and anhydrite (for obvious reason without water bands [52]). The spectral features in the low-wavenumber range, additionally enable the differentiation of the anhydrite phases I, II and III. All anhydrites investigated in this study, resemble the Raman-spectroscopic properties of anhydrite II, as anhydrite III would show a significantly shifted main band at 1025 cm1, and in the anhydrite I spectrum – according to the literature – two additional bands at 170 cm1 and 989 cm1 are expected to show up, which were not present in our spectra so far [52]. Again, we would like to point out that crystal orientation in solid samples influences relative band intensities, which here can be seen for the bands of gypsum in the range between 400 cm1 and 700 cm1 in spectra (1)–(3). Also the different relative intensities of the water modes around 3500 cm1 are ascribed to this effect [65]. With these spectra we also would like to draw the reader’s attention to another spectroscopic interference, which is typical for such thin-sections of construction materials. Above, we mentioned fluorescence emission from the embedding medium present in porous parts of the sample. Even if the employed embedding resin does not fluoresce in the spectral range, in which Raman measurements are performed, the medium might be the source of a Raman spectrum. As fluorescence is intrinsically stronger than Raman scattering (depending on the sample, by up to several orders of magnitude), only traces of fluorescing embedding medium present on the measurement spot can have a significant influence on the result. Detection of the Raman spectrum of a nonfluorescing medium requires concentrations, which are similar to the concentrations of analytes and, for example, in large resinfilled pores Raman spectra purely assignable to the mounting medium can be acquired (data not shown). Indeed, care has to be taken with the interpretation of such spectra, especially because Raman bands of inorganic analytes and of organic mounting media can occur in the same spectral range. When just considering single bands from a spectrum, their origin from either inorganic or organic sample constituents cannot be evidenced. Again, according to the concept of fingerprint comparison described above, only groups of bands, the full spectrum, or in the best case several spectra together, which were acquired on different spots, should be interpreted. When, for example, comparing all gypsum spectra ((1)–(3) in Fig. 5) with the gypsum/anhydrite spectrum (4), the similarities and differences described above can be found. On the other hand, strong similarities between the gypsum spectrum (3) and the anhydrite/gypsum spectrum (4) are present, for example, in the

range between 1100 cm1 and 1200 cm1, seemingly contradicting the concept of spectroscopic fingerprints. These are bands, which can be assigned to the organic embedding medium. The bands at around 2900 cm1 (ACAH stretching modes involving singlebonded, sp3-hybridised carbon atoms) and, in this specific mounting medium, also at approx. 3100 cm1 (@CAH stretching vibrations with double-bonded, sp2-hybridised carbon atoms) are obvious marker bands of organic compounds or, more precisely, hydrocarbons in general. Other typical organic signatures are C–H bending vibrations, which typically lead to broad (consisting of several overlapping modes) Raman bands at around 1450 cm1 (not shown here). A look at the whole spectra makes obvious that the intensity of the C–H stretching bands is correlated with the intensities of other resin bands, all of them marked with asterisks, which are present in spectra (4), (3) and (1). Indeed, each embedding medium has a different Raman spectrum [34], and samples intrinsically containing organic materials (e.g., varnish) would also show bands in the range around 2900 cm1 and in the low-wavenumber range, potentially interfering with bands from inorganic analytes.

4. Conclusions Raman microscopy provides phase analysis of building materials with approx. 500 nm spatial resolution, enabling imaging of the spatial distributions of crystalline (including polymorphic) and amorphous phases as well as crystal orientations. By this approach, not only mortar components can be detected, for which other techniques are ‘blind’ (e.g., amorphous phases in XRD or minor constituents in concentrations below the detection limits of bulk analysis techniques), but the co-localisation and spatial arrangement of phases provides access to the study of their formation mechanisms as well as relationships between microstructure and material properties. The sub-micrometre lateral resolution obtained with commercial Raman microscopes operating at ambient conditions without special pre-treatment of conventional cross- and thin sections overcomes limitations in resolution of other microscopic techniques that yield chemical information, such as infrared microspectroscopy and laboratory-scale microXRD. As the heterogeneity of the calcination conditions in the different historical kiln types influenced the phase composition of the mineral binder (cement, gypsum) and with it the process of setting, the phase composition of not hydrated relicts of the binder in aged mortars reflects historical know-how. In this study, we successfully identified alite Ca3SiO5, tricalciumaluminate Ca3Al2O6 or wollastonite Ca2SiO4, as well as dehydration and hydration stages in the system CaSO4–H2O in cement or gypsum-based mortars, respectively. Furthermore, by combining the knowledge from written sources and modern analyses, a deeper understanding of the huge variety of historical building materials can be achieved, allowing the practitioner to choose appropriate restoration materials for the conservation of the architectural heritage. Beyond the analysis of preserved remnants of binder grains, Raman microscopy provides insight into reaction products of ageing and alteration phenomena, as successfully demonstrated for dedolomitisation observed in dolomitic aggregates, natural accessory mineral of the exploited gypsum deposit of a medieval gypsumbased mortar. As user-friendly Raman microscopy instruments become more and more commercially available, the technique becomes accessible to a wider field of users. Therefore, we discussed also some of the limitations and pitfalls of the approach and provided strategies to overcome them. We would like to point out again, that an unambiguous identification of phases can only be achieved by taking the

T. Schmid, P. Dariz / Construction and Building Materials 114 (2016) 506–516

concept of fingerprint comparison seriously and using at least groups of bands, in the ideal case whole spectra, for the comparison with reference data. Indeed, only components can be determined of which reference spectra are known, and therefore the build-up of a spectra collection of relevant phases found in building materials is part of our ongoing research [15]. Another issue to take care of are spectroscopic interferences from embedding media. Fluorescence can lead to a raised baseline (making baseline correction of analysed bands inevitable in many cases) or in extreme cases can overwhelm and therefore hamper the evaluation of Raman signals. Furthermore, high concentrations of nonfluorescent resins in porous parts of the mortar samples can lead to the detection of their Raman spectra, which interfere with the spectra of analytes. Care has to be taken with the interpretation of such data, and the typical organic Raman signatures – namely the C–H stretching and bending modes – should be known to correctly recognise these interferences. The study of the spectroscopic properties of embedding media and the search for ‘ideal’ resins for sample preparation are part of our ongoing research as well [34]. Even with the limitations of a not yet complete spectra database library and sub-optimal embedding media, Raman microscopy has been proven to be a valuable extension of the existing methodological arsenal employed for the analysis of both, historic and modern building materials.

[14]

[15]

[16] [17]

[18]

[19]

[20]

[21]

[22]

[23]

Acknowledgements [24]

The authors thank Dr. Christine Bläuer (Conservation Science Consulting Sárl, Fribourg, Switzerland) and Ueli Fritz (Bern University of the Arts, Bern, Switzerland) for providing samples from Merkurhaus, Horw (C.B.), and the former printing office Schläpfer, Weinfelden (U.F.). Financial support from Stiftung zur Förderung der Denkmalpflege (Zurich, Switzerland) is gratefully acknowledged. References [1] R.L. McCreery, Raman Spectroscopy for Chemical Analysis, Wiley-Interscience, 2000. [2] L. Black, Raman spectroscopy of cementitious materials, Spectrosc. Prop. Inorg. Organomet. Compd. 40 (2009) 72–127. [3] J. Perez-Rodriguez, M. Robador, M. Centeno, B. Siguenza, A. Duran, Wall paintings studied using Raman spectroscopy: a comparative study between various assays of cross sections and external layers, Spectrochim. Acta A 120 (2014) 602–609. [4] M. Veneranda, M. Irazola, A. Pitarch, M. Olivares, A. Iturregui, K. Castro, J. Madariaga, In-situ and laboratory Raman analysis in the field of cultural heritage: the case of a mural painting, J. Raman Spectrosc. 45 (2014) 228–237. [5] H. Edwards, D. Farwell, The conservational heritage of wall paintings and buildings: an FT-Raman spectroscopic study of prehistoric, Roman, mediaeval and Renaissance lime substrates and mortars, J. Raman Spectrosc. 39 (2008) 985–992. [6] I. Bell, R. Clark, P. Gibbs, Raman spectroscopic library of natural and synthetic pigments, Spectrochim. Acta A 53 (1997) 2159–2179. [7] L. Burgio, R. Clark, Library of FT-Raman spectra of pigments, minerals, pigment media and varnishes, and supplement to existing library of Raman spectra of pigments with visible excitation, Spectrochim. Acta A 57 (2001) 1491–1521. [8] L. Opilik, T. Schmid, R. Zenobi, Modern Raman imaging: vibrational spectroscopy on the micrometer and nanometer scales, Annu. Rev. Anal. Chem. 6 (2013) 379–398. [9] S. Sahu, D. Exline, M. Nelson, Identification of thaumasite in concrete by Raman chemical imaging, Cem. Concr. Compos. 24 (2002) 347–350. [10] K. Jallad, M. Santhanam, M. Cohen, D. Ben-Amotz, Chemical mapping of thaumasite formed in sulfate-attacked cement mortar using near-infrared Raman imaging microscopy, Cem. Concr. Res. 31 (2001) 953–958. [11] P. Ropret, C. Miliani, S. Centeno, Raman mapping in the scientific investigations of works of art, in: A. Zoubir (Ed.), Raman Imaging, Springer Verlag, Berlin Heidelberg, 2012, pp. 189–217. [12] K. Jallad, D. Ben-Amotz, Chemical imaging of iron oxides and oxyhydroxides using near-infrared Raman imaging microscopy, Mater. Sci. Technol. 17 (2001) 1479–1486. [13] V. Antunes, A. Candeias, M. Oliveira, S. Longelin, V. Serrão, A. Seruya, J. Coroado, L. Dias, J. Mirão, M. Carvalho, Characterization of gypsum and anhydrite ground layers in 15th and 16th centuries Portuguese paintings by

[25]

[26]

[27] [28]

[29]

[30]

[31] [32] [33]

[34]

[35] [36] [37]

[38] [39] [40] [41]

515

Raman spectroscopy and other techniques, J. Raman Spectrosc. 45 (2014) 1026–1033. T. Schmid, P. Dariz, Determination and imaging of binder remnants and aggregates in historic artificial stone by Raman microscopy, J. Raman Spectrosc. 44 (2013) 882–891. P. Dariz, J. Neubauer, F. Götz-Neunhoeffer, T. Schmid, Calcium aluminates in clinker remnants as marker minerals for various types of 19th-century cement studied by Raman microscopy, 2015 (submitted). A. Ferrari, J. Robertson, Interpretation of Raman spectra of disordered and amorphous carbon, Phys. Rev. B 61 (2000) 14095–14107. R. Flemming, Micro X-ray diffraction (lXRD): a versatile technique for characterization of Earth and planetary materials, Can. J. Earth Sci. 44 (2007) 1333–1346. J. Skibsted, C. Hall, Characterization of cement minerals, cements and their reaction products at the atomic and nano scale, Cem. Concr. Res. 38 (2008) 205–225. A. Crumbie, G. Walenta, T. Füllmann, Where is the iron? Clinker microanalysis with XRD Rietveld, optical microscopy/point counting, Bogue and SEM-EDS techniques, Cem. Concr. Res. 36 (2006) 1542–1547. M. Aranda, Recent studies of cements and concretes by synchrotron radiation crystallographic and cognate methods, Crystallogr. Rev. 22 (2016), http://dx. doi.org/10.1080/0889311X.2015.1070260. Z. Ristanovic, J. Hofmann, U. Deka, T. Schülli, M. Rohnke, A. Beale, B. Weckhuysen, Intergrowth structure and aluminium zoning of a zeolite ZSM5 crystal as resolved by synchrotron-based micro X-ray diffraction imaging, Angew. Chem. Int. Ed. 52 (2013) 13382–13386. A. Diekamp, R. Stalder, J. Konzett, P. Mirwald, Lime mortar with natural hydraulic components: characterisation of reaction rims with FTIR imaging in ATR-mode, in: J. Valek, J. Hughes, C. Groot (Eds.), Historic Mortars: Characterisation, Assessment and Repair, Springer, Netherlands, Dordrecht, 2012, pp. 105–113. E. Joseph, S. Prati, G. Sciutto, M. Ioele, P. Santopadre, R. Mazzeo, Performance evaluation of mapping and linear imaging FTIR microspectroscopy for the characterisation of paint cross sections, Anal. Bioanal. Chem. 396 (2010) 899– 910. M. Spring, C. Ricci, D. Peggie, S. Kazarian, ATR-FTIR imaging for the analysis of organic materials in paint cross sections: case studies on paint samples from the National Gallery, Anal. Bioanal. Chem. 392 (2008) 37–45. S. Zumbuehl, N. Scherrer, U. Eggenberger, Derivatization technique to increase the spectral selectivity of two-dimensional Fourier transform infrared focal plane array imaging: analysis of binder composition in aged oil and tempera paint, Appl. Spectrosc. 68 (2014) 458–465. G. Van der Snickt, K. Janssens, J. Dik, W. De Nolf, F. Vanmeert, J. Jaroszewicz, M. Cotte, G. Falkenberg, L. Van der Loeff, Combined use of synchrotron radiation based micro-X-ray fluorescence, micro-X-ray diffraction, micro-X-ray absorption near-edge, and micro-Fourier transform infrared spectroscopies for revealing an alternative degradation pathway of the pigment cadmium yellow in a painting by van Gogh, Anal. Chem. 84 (2012) 10221–10228. L. Rayleigh, Investigations in optics, with special reference to the spectroscope, Philos. Mag. 8 (1879) 261–274. J. Álvarez-García, V. Izquierdo-Roca, P. Pistor, T. Schmid, A. Pérez-Rodríguez, Raman spectroscopy on thin-film solar cells, in: D. Abou-Ras, T. Kirchartz, U. Rau (Eds.), Advanced Characterization Techniques for Thin Film Solar Cells, second ed., Wiley-VCH, Weinheim, 2016. J. Weber, N. Gadermayer, D. Hughes, R. Kozlowski, M. Stillhammerova, D. Ullrich, R. Vyskocilova, Roman cement mortars in Europe’s architectural heritage of the 19th century, J. ASTM Int. 4 (2007) 69–83. J. Weber, N. Gadermayr, R. Kozlowski, D. Mucha, D. Hughes, D. Jaglin, W. Schwarz, Microstructure and mineral composition of Roman cements produced at defined calcination conditions, Mater. Charact. 58 (2007) 1217– 1228. R.T. Downs, in: Program and Abstracts of the 19th General Meeting of the International Mineralogical Association in Kobe, Japan 2006, 2006. O03-13. M. Conjeaud, H. Boyer, Some possibilities of Raman microprobe in cement chemistry, Cem. Concr. Res. 10 (1980) 61–70. I.G. Richardson, J. Skibsted, L. Black, R.J. Kirkpatrick, Characterisation of cement hydrate phases by TEM, NMR and Raman spectroscopy, Adv. Cem. Res. 22 (2010) 233–248. T. Schmid, R. Jungnickel, B. Neuhaus, J. Riedel, J. Kneipp, C. Lüter, Raman spectroscopy as a tool for the collection management of microscope slides, Zool. Anz. (2015) (submitted). P. Hewlett (Ed.), Lea’s Chemistry of Cement and Concrete, Butterworth Heinemann, Oxford, 2007. H. Taylor, Cement Chemistry, Thomas Telford Publishing, London, 1997. K. Callebaut, J. Elsen, K. Van Balen, W. Viaene, Nineteenth century hydraulic restoration mortars in the Saint Michael’s Church (Leuven, Belgium). Natural hydraulic lime or cement?, Cem Concr. Res. 31 (2001) 397–403. M. Exner (Ed.), Stuck des frühen und hohen Mittelalters. Geschichte, Technologie, Konservierung, Karl M. Lipp Verlag, München, 1996. M. Hoernes (Ed.), Hoch- und spätmittelalterlicher Stuck. Material – Technik – Stil – Restaurierung, Verlag Schnell und Steiner, Regensburg, 2002. P. Dariz, Die Pietà im Kloster Marienberg. Eine Studie zu mittelalterlichem Kunststein, ZKK 19 (2006) 100–116. J. Bischoff, R. Julia, W. Shanks, R. Rosenbauer, Karstification without carbonic acid – bedrock dissolution by gypsum-driven dedolomitization, Geology 22 (1994) 995–998.

516

T. Schmid, P. Dariz / Construction and Building Materials 114 (2016) 506–516

[42] A. Palmer, M. Palmer, Sulfate-carbonate interactions in the development of karst, Northeastern Geol. Environ. Sci. 26 (2004) 93–106. [43] C. Ayora, C. Taberner, M.W. Saaltink, J. Carrera, The genesis of dedolomites: a discussion based on reactive transport modeling, J. Hydrol. 209 (1998) 346– 365. [44] V. Vandeginste, C. John, Influence of climate and dolomite composition on dedolomitization: insights from a multi-proxy study in the central Oman Mountains, J. Sediment. Res. 82 (2012) 177–195. [45] S. Martinez-Ramirez, S. Sanchez-Cortes, J. Garcia-Ramos, C. Domingo, C. Fortes, M. Blanco-Varela, Micro-Raman spectroscopy applied to depth profiles of carbonates formed in lime mortar, Cem. Concr. Res. 33 (2003) 2063–2068. [46] A. Dandeu, B. Humbert, C. Carteret, H. Muhr, E. Plasari, J.-M. Bossoutrot, Raman spectroscopy – a powerful tool for the quantitative determination of the composition of polymorph mixtures: application to CaCO3 polymorph mixtures, Cem. Eng. Technol. 29 (2006) 221–225. [47] S. Potgieter-Vermaak, J. Potgieter, R. Van Grieken, The application of Raman spectroscopy to investigate and characterize cement. Part I: a review, Cem. Concr. Res. 36 (2006) 656–670. [48] H. Edwards, S. Jorge Villar, J. Jehlicka, T. Munshi, FT-Raman spectroscopic study of calcium-rich and magnesium-rich carbonate minerals, Spectrochim. Acta A 61 (2005) 2273–2280. [49] S. Gunasekaran, G. Anbalagan, S. Pandi, Raman and infrared spectra of carbonates of calcite structure, J. Raman Spectrosc. 37 (2006) 892–899. [50] J. Sun, Z. Wu, H. Cheng, Z. Zhang, R. Frost, A Raman spectroscopic comparison of calcite and dolomite, Spectrochim. Acta A 117 (2014) 158–162. [51] El.-Sayed. Youssef, Dolomitization and dedolomitization of miocene carbonates at Gabal El-Safra, Sherm el Sheikh Area, Sinai, Egypt. Carbonate. Evaporite. 12 (1997) 185–191. [52] N. Prieto-Taboada, O. Gómez-Laserna, I. Martínez-Arkarazo, M.Á. Olazabal, J.M. Madariaga, Raman spectra of the different phases in the CaSO4–H2O system, Anal. Chem. 86 (2014) 10131–10137. [53] H. Chang, P. Huang, S. Hou, Application of thermo-Raman spectroscopy to study dehydration of CaSO42H2O and CaSO40,5H2O, Mater. Chem. Phys. 58 (1999) 12–19. [54] S. Seufert, C. Hesse, F. Goetz-Neunhoeffer, J. Neubauer, Quantitative determination of anhydrite III from dehydrated gypsum by XRD, Cem. Concr. Res. 39 (2009) 936–941.

[55] C. Berthold, V. Presser, N. Huber, K. Nickel, 1 + 1 = 3: Coupling l-XRD2 and DTA. New insights in temperature-dependent phase transitions. The gypsumbassanite-anhydrite system as an example, J. Therm. Anal. Calorim. 103 (2011) 917–923. [56] A. Christensen, M. Olesen, Y. Cerenius, T. Jensen, Formation and transformation of five different phases in the CaSO4–H2O system: crystal structure of the subhydrate b-CaSO40, 5H2O and soluble anhydrite CaSO4, Chem. Mater. 20 (2008) 2124–2132. [57] Bundesverband der Gipsindustrie, Gips-Datenbuch, Darmstadt, 2006. [58] H. Kulke, Gips, seine Entstehung, Ausbildung und Vorkommen. Historische Gips(Mörtel) – Verwendungsprovinzen besonders in Mitteleuropa, in: M. Hoernes (Ed.), Hoch- und spätmittelalterlicher Stuck. Material – Technik – Stil – Restaurierung., Verlag Schnell und Steiner, Regensburg, 2002, pp. 27–42. [59] F. Schlütter, M. Jakubek, H. Juling, Charakterisierung und Eigenschaften historischer Gipsmörtel aus unterschiedlichen Epochen und Anwendungsgebieten, in: Institut für Steinkonservierung e.V. (Ed.), Gips als Baugrund, Mörtel und Dekorationsmaterial, Self-Publishing, Mainz, 2012, pp. 49–59. [60] T. Schmid, N. Schäfer, S. Levcenco, T. Rissom, D. Abou-Ras, Orientationdistribution mapping of polycrystalline materials by Raman microspectroscopy, Sci. Rep. 5 (2015) 18410. [61] B. Middendorf, Physico-mechanical and microstructural characteristics of historic and restoration mortars based on gypsum: current knowledge and perspective, in: S. Siegesmund, T. Weiss, A. Vollbrecht (Eds.), Natural Stone. Weathering Phenomena, Conservation Strategies and Case Studies, Geological Society, London, 2002, pp. 165–176. [62] R. Lenz, R. Sobott, Beobachtungen zu Gefügen historischer Gipsmörtel, in: M. Auras, H.-W. Zier (Eds.), Gipsmörtel im historischen Mauerwerk und an Fassaden, Wissenschaftlich-Technische Arbeitsgemeinschaft für Bauwerkserhaltung und Denkmalpflege, München, 2008, pp. 23–34. [63] F. Schlütter, Mittelalterlicher Hochbrandgips, in: Brandenburgisches Landesamt für Denkmalpflege und Archäologisches Landesmuseum, 800 Jahre Kunststein vom Imitat zum Kunstgut, Wernersche Verlagsgesellschaft, Worms, 2012, pp. 27–39. [64] P. Prasad, A. Pradhan, T. Gowd, In situ micro-Raman investigation of dehydration mechanism in natural gypsum, Curr. Sci. 80 (2001) 1203–1207. [65] N. Krishnamurthy, V. Soots, Raman spectrum of gypsum, Can. J. Phys. 49 (1971) 885–896.