heritage

Article Raman Microspectroscopic Imaging of Binder Remnants in Historical Mortars Reveals Processing Conditions

1,2, , 2,3, Thomas Schmid * † and Petra Dariz † 1 BAM Federal Institute for Materials Research and Testing, Richard-Willstätter-Str. 11, 12489 Berlin, Germany 2 School of Analytical Sciences Adlershof (SALSA), Humboldt-Universität zu Berlin, Unter den Linden 6, 10099 Berlin, Germany; [email protected] 3 Bern University of Applied Sciences, Bern University of the Arts, Conservation-Restoration, Fellerstr. 11, 3027 Bern, Switzerland; [email protected] * Correspondence: [email protected] or [email protected] These authors contributed equally to this work. †


Received: 26 April 2019; Accepted: 12 June 2019; Published: 14 June 2019


Abstract: Binder remnants in historical mortars represent a record of the connection between the raw materials that enter the kiln, the process parameters, and the end product of the calcination. Raman microspectroscopy combines high structural sensitivity with micrometre to sub-micrometre spatial resolution and compatibility with conventional thin-sectional samples in an almost unique fashion, making it an interesting complementary extension of the existing methodological arsenal for mortar analysis. Raman spectra are vibrational fingerprints of crystalline and amorphous compounds, and contain marker bands that are specific for minerals and their polymorphic forms. Relative intensities of bands that are related to the same crystalline species change according to crystal orientations, and band shifts can be caused by the incorporation of foreign ions into crystal lattices, as well as stoichiometric changes within solid solution series. Finally, variations in crystallinity affect band widths. These effects are demonstrated based on the analysis of three historical mortar samples: micrometric distribution maps of phases and polymorphs, crystal orientations, and compositional variations of solid solution series of unreacted clinker grains in the Portland cement mortars of two 19th century castings, and the crystallinities of thermal anhydrite clusters in a high-fired medieval gypsum mortar as a measure for the applied burning temperature were successfully acquired.

Keywords: cement clinker remnants; polymorphic transitions; striations; remelted belite; high-fired gypsum; thermal anhydrite; pyrometamorphism; spectroscopic imaging; Raman microscopy

1. Introduction Morphology and phase assemblage of unreacted binder grains in the hydrated and/or carbonated binder matrix of historical mortars reflect the composition, comminution, as well as the firing and cooling history (temperature of firing, residence time, kiln atmosphere, etc.) of the kiln raw feed. Thus, detailed analyses of historical mortars are particularly essential in developing a better understanding of the parameters of the historical production of mineral binders, which is a prerequisite for tailoring compatible restoration materials. Microscopically resolved measurements with high structural and chemical specificity are needed to pinpoint and analyse such indicative mineral clusters within the mortar samples. Raman measurements on polished thin sections in the imaging mode, complemented by scanning electron and (polarised) light microscopies, have demonstrated fulfilling the demands for analysis of architectural heritage; the utility of Raman microscopy particularly with regard to the

Heritage 2019, 2, 1662–1683; doi:10.3390/heritage2020102 www.mdpi.com/journal/heritage Heritage 2019, 2 1663

Heritage 2019, 2 FOR PEER REVIEW 2 assessment of the original composition of aged mortars and the differentiation of the alteration phases 45 ismicroscopy discussed inparticularly Refs. [1–9 ].with regard to the assessment of the original composition of aged mortars 46 and theRaman differentiation spectroscopy of the yields alteration structural phases fingerprints is discussed that in Refs. enable [1–9]. the identification of organic 47 moleculesRaman and spectroscopy inorganic phases. yields Based structural on inelastic fingerprin light scattering—alsots that enable termedthe identification Raman scattering—the of organic 48 analyticalmolecules and technique inorganic provides phases. peaks Based (termedon inelastic bands) light thatscattering—also represent Raman-active termed Raman molecular scattering— or 1 49 crystal-latticethe analytical vibrations, technique whichprovides appear peaks at specific(termed vibrational bands) that frequencies represent or Rama wavenumbersn-active molecular (unit cm− or), 50 respectivelycrystal-lattice [10 vibrations,]. The samples which are appear excited byat anspecific intense vibrational and monochromatic frequencies source or wavenumbers of electromagnetic (unit 51 radiation,cm−1), respectively typically a[10]. continuous The samples visible laserare excited beam. Theby lightan intense that is scatteredand monochromatic by the sample source contains of 52 photonselectromagnetic with frequencies radiation, di typicallyffering from a continuous the excitation. visible This laser diff beam.erence The is called light that the Ramanis scattered shift by and the it 53 equalssample the contains frequency photons of a molecularwith frequencies or crystal-lattice differing from vibration, the excitation. which is This excited difference by this processis called (see the 54 FigureRaman1 ).shift Being and dependent it equals the on symmetry,frequency atomicof a molecular masses, or bond crystal-lattice strengths, lengths,vibration, and which angles, is theexcited set 55 ofby its this vibrational process (see frequencies Figure 1). (termed Being dependent the vibrational on symmetry, spectrum) atomic is highly masses, specific bond for strengths, a certain molecule lengths, 56 orand crystal. angles, Thus, the set Raman of its spectroscopyvibrational frequencies is related to (t infraredermed the (IR) vibrational spectroscopy spectrum) in terms is ofhighly information specific 57 contentfor a certain [11]. molecule or crystal. Thus, Raman spectroscopy is related to infrared (IR) spectroscopy 58

59 60 FigureFigure 1.1. Basic concept of elastic (Rayleigh) and inelastic (Raman) scattering of lightlight byby moleculesmolecules oror ν 61 crystals.crystals. AA monochromatic monochromatic laser laser beam beam with with frequency frequencyν0 excites 0 excites a characteristic a characteristic vibration vibration of a chemical of a ν 62 compoundchemical compound having the having frequency the frequencyνvib. The lightvib. The can light be either can be Rayleigh either Rayleigh scattered scattered resulting resulting in light ofin ν ν ν 63 thelight same of the frequency same frequencyν0 or Raman 0 or scatteredRaman scattered yielding yielding either Stokes either (ν Stokes0 νvib ()0 or − Anti-Stokesvib) or Anti-Stokes shifted ν ν − 64 (shiftedν0 + νvib ( )0 photons.+ vib) photons.

65 TheThe useuse of of visible visible light light instead instead of IRof radiation IR radiation is advantageous is advantageous for several for several reasons. reasons. While, inWhile, studies in 66 ofstudies biological of biological specimen, specimen, the weak the interference weak interferen of waterce bands of water with bands analyte with spectra analyte is often spectra mentioned, is often 67 thementioned, higher spatial the higher resolution spatial is an re importantsolution is point an inimportant microstructural point in materials microstructural investigations materials [11]. 68 Theinvestigations lateral resolution [11]. The of lateral a microscope resolution depends of a microscope on the wavelength depends of on light the wavelength and the numerical of light aperture and the 69 (NA)numerical of the aperture objective (NA) lens of used the [objective12]. The latterlens used is a measure[12]. The for latter the is focusing a measure angle, for the with focusing a high NAangle, or 70 angle,with a respectively,high NA or relatingangle, respectively, to a small focus relating spot, to and a small thus highfocus resolution spot, and (forthus a high detailed resolution explanation, (for a 71 seedetailed the Supplementary explanation, see Information the Supplementary for Ref. [13 Informatio]). The focusn for spot Ref. also [13]). shrinks The focus if a shorter spot also wavelength shrinks if 72 ofa shorter light is employed.wavelength Therefore, of light is theemployed. lateral resolution Therefore, of the IR microspectroscopylateral resolution of is IR in themicrospectroscopy order of several 73 tois in 10s the of orderµm, whileof several sub-micrometre to 10s of µm, resolution while sub- canmicrometre be achieved resolution with Ramancan be achieved microscopes with that Raman are 74 operatedmicroscopes in the that range are operated of visible in lightthe range [11]. of Due visible to this light spectral [11]. Due range, to this both spectral optics range, and samples both optics that 75 areand optimised samples that for classical are optimised light microscopy for classical are light typically microscopy compatible are withtypically this microscopic compatible technique.with this 76 Thus,microscopic polished technique. cross sections Thus, polished and thin cross sections sections of materials and thin sections on glass of slides materials can on be glass analysed slides with can 77 Ramanbe analysed microspectroscopy. with Raman microspectroscopy. 78 ThisThis last point point we we have have to to discuss discuss in inlittle little more more detail detail with with respect respect to typical to typical limitations limitations and 79 anddrawbacks drawbacks of Raman of Raman (micro)spectroscopy. (micro)spectroscopy. Raman Raman scattering scattering is istypically typically one one of of the weakestweakest 80 spectroscopicspectroscopic eeffectsffects excited excited by by an an intense intense laser laser beam. beam If. aIf sample a sample can can emit emit fluorescence fluorescence radiation, radiation, it can it 81 can easily be orders of magnitude stronger than Raman scattering. Thus, resins containing fluorescent 82 dyes are typically not compatible with this technique. Some embedding media can emit fluorescence, 83 even without purposely added fluorophores, for example, due to material ageing, and even traces of Heritage 2019, 2 1664

easily be orders of magnitude stronger than Raman scattering. Thus, resins containing fluorescent dyes are typically not compatible with this technique. Some embedding media can emit fluorescence, even without purposely added fluorophores, for example, due to material ageing, and even traces of

such fluorescingHeritage 2019, 2 FOR compounds PEER REVIEW can significantly interfere with Raman spectra due to the weakness3 of the latter. Furthermore, potentially interfering Raman bands of embedding resins and glues that are 84employed such fluorescing for attaching compounds cover slips can shouldsignificantly be taken interfere into with account Raman during spectra sample due to preparation. the weakness of 85 Figurethe latter.2 shows Furthermore, a schematic potentiall of oney interfering possible RamanRaman bands microscopy of embe set-up.dding resins Laser and light glues is focused that are onto 86the sampleemployed surface for attaching by a microscope cover slips objectiveshould be taken lens, whichinto account is also during employed sample forpreparation. the collection of the 87 Figure 2 shows a schematic of one possible Raman microscopy set-up. Laser light is focused onto backscattered light. Notch or edge filters filter out Raman-shifted photons and they are finally fed 88 the sample surface by a microscope objective lens, which is also employed for the collection of the into a grating spectrometer enabling the simultaneous detection of the wavenumbers in a selected 89 backscattered light. Notch or edge filters filter out Raman-shifted photons and they are finally fed 90spectral into rangea grating by spectrometer use of a charge-coupled enabling the simultaneous device (CCD) detection detector. of the Raman wavenumbers mapping in a experimentsselected 91can bespectral realised range by by step-wise use of a charge-coupled movement of device the sample (CCD) detector. through Raman the laser mapping focus, experiments while acquiring can a 92Raman be spectrumrealised by instep-wise every pixel movement of an image.of the sample The typical through parameters the laser focus, to be while optimised acquiring in a such Raman Raman 93microspectroscopic spectrum in every measurements pixel of an areimage. the laserThe typi wavelength,cal parameters laser to power, be optimised focusing in element such Raman (objective 94lens), microspectroscopic spectrometer grating, measurements the acquisition are the laser time, wave andlength, pixel size laser and power, number. focusing All element of these (objective parameters 95have tolens), be spectrometer carefully selected grating, to the reduce acquisition interference time, and from pixel fluorescence size and number. emission, All of avoid these sample parameters damage, 96and achievehave to reasonablebe carefully experimentselected to reduce times, interference even for acquiringfrom fluorescence Raman maps,emission, easily avoid consisting sample of 97 damage, and achieve reasonable experiment times, even for acquiring Raman maps, easily consisting thousands of spectra. Having performed a correct Raman mapping experiment is only one challenge, 98 of thousands of spectra. Having performed a correct Raman mapping experiment is only one as the data evaluation is typically even more demanding. Consisting of x and y coordinates, as well 99 challenge, as the data evaluation is typically even more demanding. Consisting of x and y 100as thecoordinates, wavenumber as well and as intensity the wavenumber axes of each and spectrum,intensity axes the of Raman each spectrum, imaging the data Raman sets areimaging typically 101four-dimensional data sets are andtypically they needfour-dimensional to be transferred and they into need two- to or be three-dimensional transferred into two- representations or three- of 102selected dimensional spectroscopic representations features byof appropriateselected spectroscopi data evaluationc features (seeby appropriat Figure3)[e 14data]. evaluation (see 103 Figure 3) [14].

104 105 FigureFigure 2. Schematic 2. Schematic of anof an upright upright Raman Raman microscope. microscope. The The Raman-scattered Raman-scattered light light collected collected and andfiltered filtered 106 by a microscopeby a microscope objective objective and and edge edge/notch/notch filter,filter, respectively,respectively, is is coupled coupled into into a grating a grating spectrometer. spectrometer. 107 Typically, without scanning of the spectrometer grating, the individual wavenumbers of a spectrum Typically, without scanning of the spectrometer grating, the individual wavenumbers of a spectrum 108 are simultaneously detected by a charge-coupled device (CCD) camera. Raman maps are acquired by are simultaneously detected by a charge-coupled device (CCD) camera. Raman maps are acquired by 109 raster-scanning of the sample through the laser focus and detection of a spectrum in every pixel. raster-scanning of the sample through the laser focus and detection of a spectrum in every pixel.

Heritage 2019, 2 FOR PEER REVIEW 4 Heritage 2019, 2 1665

110

111 FigureFigure 3. 3.Data Data cube cube typically typically yielded yielded byby aa RamanRaman microspectroscopicmicrospectroscopic imagingimaging experiment. On each each x 112 xand and y y coordinate of aa two-dimensionaltwo-dimensional area,area, aa RamanRaman spectrumspectrum consistingconsisting of of wavenumbers wavenumbers and and 113 intensitiesintensities (here (here shown shown as as di differentfferent colours) colours) is acquired.is acquired. The The simplest simplest possibility possibility to analyseto analyse the the data data is 114 theis the evaluation evaluation of Ramanof Raman intensities intensities at aat selected a selected wavenumber wavenumber (e.g., (e.g., resembling resembling a markera marker band band of of a a 115 certaincertain compound), compound), here here represented represented by by horizontal horizontal slices slices of of the the cube. cube For. For example, example, the the colours colours on on the the 116 toptop face face reveal reveal the the intensity intensity distribution distribution at at the the highest highest wavenumber wavenumber in in each each acquired acquired spectrum. spectrum.

117 InIn this this contribution, contribution, we we demonstrate demonstrate how how the the intensities intensities of of marker marker bands bands (for (for both, both, phases phases and and 118 polymorphs),polymorphs), relative relative band band intensities, intensities, band band shifts, shifts, and even and changes even changes in band shapesin band can shapes be evaluated can be 119 toevaluated yield specific to yield material specific information material information and contrast an ford contrast Raman microspectroscopicfor Raman microspectroscopic imaging. imaging.

120 2.2. Materials Materials and and Methods Methods

2.1. Samples 121 2.1. Samples The following observations concern prefabricated concrete elements and a cast lion sculpture, 122 The following observations concern prefabricated concrete elements and a cast lion sculpture, both produced for the industrial exhibition in Zurich (Switzerland) in 1894. The sample of a high-fired 123 both produced for the industrial exhibition in Zurich (Switzerland) in 1894. The sample of a high- gypsum mortar originates from the Carolingian Chapel of the Holy Cross of the Benedictine convent 124 fired gypsum mortar originates from the Carolingian Chapel of the Holy Cross of the Benedictine of St. John in Müstair (Switzerland). Both types of mortar were prepared for analysis in a specialised 125 convent of St. John in Müstair (Switzerland). Both types of mortar were prepared for analysis in a laboratory by embedding in epoxy resin under vacuum and manufacturing polished thin sections 126 specialised laboratory by embedding in epoxy resin under vacuum and manufacturing polished thin (30 µm thickness), which were studied under polarising light microscopes (Leica Microsystems 127 sections (30 µm thickness), which were studied under polarising light microscopes (Leica 128 LaborluxMicrosystems that was Laborlux equipped that with was ProgResequipped SpeedXT3 with ProgRes and Zeiss SpeedXT3 AxioScope.A1 and Zeiss MAT AxioScope.A1 equipped withMAT 129 AxioCamequipped MRc) with inAxioCam transmitted MRc) polarised in transmitted light. polarised light. 2.2. Raman Microspectroscopy 130 2.2. Raman Microspectroscopy The Raman spectra and maps were acquired by using a LabRam HR 800 instrument (Horiba Jobin 131 The Raman spectra and maps were acquired by using a LabRam HR 800 instrument (Horiba Yvon) coupled to a BX41 microscope (Olympus). For both, excitation and collection of the scattered light, 132 Jobin Yvon) coupled to a BX41 microscope (Olympus). For both, excitation and collection of the a 50x/NA = 0.75 objective lens was employed. The system is equipped with a diode-pumped solid-state 133 scattered light, a 50x/NA = 0.75 objective lens was employed. The1 system is equipped with a diode- (DPSS) laser that has a wavelength of 532 nm and a 1800 mm− grating offering a resolution of the 134 pumped solid-state (DPSS) laser that has a wavelength of 532 nm and a 1800 mm−1 grating offering a spectra that were acquired with a liquid N2-cooled ( 130 ◦C operating temperature) charge-coupled 135 resolution of the spectra that were acquired with a −liquid N2-cooled (−130 °C operating temperature) device (CCD) camera (Symphony CCD, Horiba Jobin Yvon) of approx. 0.47 cm 1 per CCD pixel at − −1 136 charge-coupled1 device (CCD) camera1 (Symphon1y CCD, Horiba Jobin Yvon) of approx. 0.47 cm per 750 cm− Raman shift and 0.45 cm− at 1100 cm− . The spectrometer entrance slit was 100 µm wide 137 CCD pixel at 750 cm−1 Raman shift and 0.45 cm−1 at 1100 cm−1. The spectrometer entrance slit was 100 and the confocal pinhole was in the fully open position (1000 µm). Prior to each Raman mapping 138 µm wide and the confocal pinhole was in the fully open position (1000 µm). Prior to each Raman experiment, the laser was allowed to stabilise for at least 2 h and the spectrometer was recalibrated 139 mapping experiment, the laser was allowed to stabilise1 for at least 2 h and the spectrometer was against the most prominent mode of silicon at 520.7 cm− . 140 recalibrated against the most prominent mode of silicon at 520.7 cm−1. The Raman maps (see Sections 3.1–3.5) were acquired under the described conditions with a laser 141 The Raman maps (see Sections 3.1 to 3.5) were acquired under the described conditions with a power at the sample of 7.5 mW (full power attenuated to 25% while using a neutral density filter). 142 laser power at the sample of 7.5 mW (full power attenuated to 25% while using a neutral density

Heritage 2019, 2 1666

1 The spectra of 19th-century Portland cement mortars were centred at 750 cm− , offering a spectral 1 1 range of approx. 500 cm− to 990 cm− in one acquisition, which includes the most prominent modes of the main calcium silicate, aluminoferrite, and aluminate clinker phases. In the measurement of a 1 high-fired medieval gypsum mortar, the spectra that were centred at 1100 cm− ranged from approx. 1 1 860 cm− to 1330 cm− , covering the most prominent features in the spectra of gypsum and anhydrite. Software-controlled (instrument software LabSpec 6, Horiba Jobin Yvon S.A.S., Longjumeau, France) step-wise movement of the sample stage and the collection of a Raman spectrum in each pixel were used to acquire the maps. The step size was 0.9 µm 0.9 µm for the cement mortars and 1 µm 1 µm × × in the case of the high-fired gypsum mortar. The lateral spatial resolution at 532 nm using a NA of 0.75 is approx. 430 nm (according to the Rayleigh criterion, calculated according to the Supporting Information for Ref. [13]) and the laser spot diameter approx. 700 nm, while the depth resolution with open confocal pinhole is estimated to amount to approx. 22.5 µm (recalculated from experimental values that were obtained with the same instrument using an NA of 0.9 and a laser wavelength of λ0 = 632.8 nm [13] by assuming that the dimension of the point spread function in depth direction 2 is proportional to λ0/(NA )). In mapping experiments, the individual spectra were acquired within 2 2 s, i.e., two accumulations of 2 s acquisition time each, by using the cosmic ray rejection option of × the software. Table1 summarises the mapping parameters.

Table 1. Measurement parameters of the Raman mapping experiments.

Sample Figures Pixel Number Area Time Baluster 1 6a, 9a, 10b, 12, 14a–b 80 100 = 8000 72 µm 90 µm 8 h 53 min × × Lion 1 5, 6b, 7d, 8, 9b, 11, 13, 14c–d 80 100 = 8000 72 µm 90 µm 8 h 53 min × × Chapel of the Holy Cross 2 16b–d 75 44 = 3300 75 µm 44 µm 3 h 40 min × × 1 19th-century cement mortar, 2 high-fired medieval gypsum mortar.

All of the mapping data were evaluated and converted into three-dimensional images (with the third dimension expressed in false colours) by using own LabView-based software (National Instruments, Austin, TX, USA).

2.3. Scanning Electron Microscopy Scanning electron micrographs and energy dispersive X-ray (EDX) maps were acquired by employing a table-top environmental scanning electron microscope (ESEM) Jeol JCM-6000 Neoscope.

3. Results and Discussion

3.1. Intensities of Marker Bands Provide Access to Distributions of Phases in Clinker Remnants in Early Cement Mortars Light microscopical examination of cement clinker is routinely used to monitor the operating kiln conditions and to evaluate the clinker quality since the pioneer work of Henri Le Chatelier and Alfred Elis Törnebohm in the late 19th century [15–21]. Beyond that, the phase assemblage and microstructure of unhydrated clinker grains in early cement mortars (see Figure4) evidence the historical process mastering [22–27]. Differences and anomalies in crystal size, morphology, abundance, and distribution when compared to present-day clinker phases are in consequence of (the combination of) factors, like deficient homogenisation of the feedstock, slow heating and slow cooling rate, long residence time, etc., as early highly hydraulic binders were produced in an intermittently operated shaft kiln; the rotary kiln, as patented by Frederick Ransome in 1885, was introduced in continental Europe only at the beginning of the 20th century [22,28]. Heritage 2019, 2 1667 Heritage 2019, 2 FOR PEER REVIEW 6

181

182 FigureFigure 4. 4.Polarised Polarised light light micrographs micrographs ofof PortlandPortland cement clinker clinker remnants remnants in in the the binder binder matrix matrix of of 183 samplessamples from from a prefabricateda prefabricated baluster baluster ((aa,,bb)) andand aa cast lion lion sculpture sculpture (c (c,d,d),), both both manufactured manufactured for for the the 184 industrialindustrial exhibition exhibition in in Zurich Zurich (Switzerland) (Switzerland) inin 1894.1894. Rectangles Rectangles highlight highlight the the areas areas on on which which the the 185 Raman maps discussed in Sections 3.1 to 3.4 were acquired. Raman maps discussed in Sections 3.1–3.4 were acquired.

186 The potential of the complementation of petrographic examination of the binder in early cement 187 mortarThe potentialsamples by of theRaman complementation microscopy was of demonstrat petrographiced examinationfor the first time of the in binder2013 [2]. in The early Raman cement 188 mortarimaging samples studies by build Raman on microscopythe decennia was of previous demonstrated research for on the Raman first time spectra in 2013 of cement-related [2]. The Raman 189 imagingcompounds studies [29–41], build which on the is the decennia main source of previous for gathering research reference on Raman Raman spectra spectra of that cement-related are needed 190 compoundsto elucidate [29 the–41 ],chemical which is identity the main of sourcecrystallit fores gathering found in reference Raman mapping Raman spectra experiments. that are Figure needed 5 to 191 elucidatevisualises the this chemical for the identity exampl ofe of crystallites beta dicalcium found silicate in Raman (β-C mapping2S or β-belite, experiments. respectively). Figure In5 this visualises part 192 thisof for the the study, example cement of betachemist’s dicalcium notation silicate is used (β-C for2S orconvenience,β-belite, respectively). according to Inwhich this partC ≙ ofCaO, the S study, ≙ 193 cementSiO2, chemist’sA ≙ Al2O3 notation, and F ≙ isFe used2O3. The for convenience,reference spectrum according in this to case which was C gathered=ˆ CaO, Sby=ˆ measuringSiO2,A =ˆ ownAl2O 3, β 194 andsynthesised F =ˆ Fe2O3 .-C The2S referencewith the same spectrum instrument, in this caseand it was agrees gathered well with by measuring the literature own data synthesised [29]. Spectralβ-C 2S 195 withlibraries the same represent instrument, another and source it agrees of reference well with data the, but literature so far only data some [29]. Spectralcontain the libraries spectroscopic represent 196 anotherdata of source inorganic of referencephases, and data, indeed, but they so far are only not specialised some contain to cement the spectroscopic clinker compounds. data of The inorganic rruff 197 phases,spectral and database indeed,they is areprincipally not specialised employed to cement in the clinker community, compounds. being The freely rruff spectralaccessible database at 198 is principallyhttp://rruff.info, employed which in comprises the community, photos, being elemen freelytal accessibleanalysis data, at http: Raman//rruff .infospectra,, which and comprises X-ray 199 diffractograms of natural minerals [42]. Thus, only cement-related phases are included, which have photos, elemental analysis data, Raman spectra, and X-ray diffractograms of natural minerals [42]. 200 a natural counterpart. In some cases, care has to be taken with spectral library data, because the Thus, only cement-related phases are included, which have a natural counterpart. In some cases, 201 natural mineral (for example, larnite in the case of β-C2S) crystallises in a different crystal class than care has to be taken with spectral library data, because the natural mineral (for example, larnite in 202 the typical clinker phase, thus being characterised by differing spectra and diffractograms (see the case of β-C S) crystallises in a different crystal class than the typical clinker phase, thus being 203 Section 3.2). The2 rruff library spectra can be downloaded and—like own reference data—directly 204 characterisedmatched with by ditheff eringsample spectra data. andIn accordance diffractograms with (seethe Sectionidea of 3.2a fingerprint). The rruff comparison,library spectra several can be downloaded and—like own reference data—directly matched with the sample data. In accordance 205 Raman bands in the sample spectrum shown in Figure 5 (labelled ‘β-C2S’) match the wavenumbers 206 withthat the are idea found of ain fingerprintthe reference comparison, spectrum (‘β several-C2S (ref.)’, Raman in this bands case inthe the spectrum sample of spectrum own synthesised shown in 207 Figureβ-C25S (labelled[38]). In general, ‘ β-C2S’) subtle match differences the wavenumbers between the that Raman are found spectra in theof synthetic reference clinker spectrum minerals (‘β-C 2S (ref.)’, in this case the spectrum of own synthesised β-C2S[38]). In general, subtle differences between

Heritage 2019, 2 1668

theHeritage Raman 2019, spectra 2 FOR PEER of synthetic REVIEW clinker minerals and of the corresponding phases occurring in historical7 or current commercial cement clinker have to be considered, which are due to the incorporation of 208 foreignand of the ions. corresponding phases occurring in historical or current commercial cement clinker have 209 to be considered, which are due to the incorporation of foreign ions.

210

211 FigureFigure 5. SpectraSpectra from from the the Raman Raman map map of a ofbinder a binder remnant remnant in the incement the cement stone of stone a cast oflion a sculpture cast lion β 212 sculpture(Zurich, Switzerland, (Zurich, Switzerland, 1894) with 1894) marker with bands marker for bands beta fordicalcium beta dicalcium silicate ( silicate-C2S, red), (β-C 2calciumS, red), 213 calciumaluminoferrite aluminoferrite solid solution solid solutionseries (C seriesxAyFz, (Cgreen),xAyF ztricalcium, green), tricalcium aluminate aluminate (C3A, white), (C3 A,and white), tricalcium and β 214 tricalciumsilicate (C3 silicateS, blue). (C The3S, blue).same Thecolour same coding colour applies coding to applies the Raman to the intensity Raman intensity maps inmaps Figure in 6. Figure The 6.- 215 TheC2S spectrumβ-C2S spectrum from the from sample the is sample compared is compared with a refe withrence a spectrum reference (labelled spectrum ‘ref.’) (labelled [42]. The ‘ref.’) asterisk [42]. 216 Themarks asterisk a band marks related a band to the related organic to embedding the organic embeddingmedium. medium.

217 FurtherFurther phasesphases that that are are found found in clinkerin clinker remnants remnan ofts the of late the 19th late century 19th century Portland Portland cement mortarscement 218 mortars under study are tricalcium silicate C3S, tricalcium aluminate C3A, as well as calcium under study are tricalcium silicate C3S, tricalcium aluminate C3A, as well as calcium aluminoferrites 219 aluminoferrites CxAyFz, which are identified based on comparison with literature data (see Ref. [29] CxAyFz, which are identified based on comparison with literature data (see Ref. [29] and references 220 and references cited therein). Tricalcium silicate or impure alite, respectively, crystallises between cited therein). Tricalcium silicate or impure alite, respectively, crystallises between about 1250 ◦C and 221 about 1250 °C and 1450 °C. Three triclinic, three monoclinic, and one rhombohedral form are known, 1450 ◦C. Three triclinic, three monoclinic, and one rhombohedral form are known, with an increase in 222 symmetrywith an increase with increasing in symmetry temperature with increasing [43]. The te Ramanmperature spectrum [43]. The that isRaman shown spectrum in Figure that5 documents is shown 223 thein Figure presence 5 documents of a monoclinic the presence polymorph of a inmonoclinic a residual polymorph clinker grain, in a thus residual of a highclinker temperature grain, thus form of a 224 mosthigh temperature probably stabilised form most by the probably dissolved stabilised impurities by the given dissolved the slow impurities cooling rategiven characterising the slow cooling the 225 rate characterising the technological level at the end of the 19th century. (Note that1 the band at 639 technological level at the end of the 19th century. (Note that the band at 639 cm− marked with an −1 226 asteriskcm marked is assigned with an to theasterisk embedding is assigned medium to the (epoxy embedding resin) usedmedium for sample (epoxy preparationresin) used [for44– sample46].) 227 preparation [44–46].)

Heritage 2019, 2 1669 Heritage 2019, 2 FOR PEER REVIEW 8

228

229 FigureFigure 6. Raman 6. Raman intensity intensity maps maps revealing revealing the the distributions distributions of clinkerof clinker phases phases in in binder binder remnants remnants found 230 in thefound cement in the stone cement of a stone baluster of a ( abaluster) and a ( casta) and lion a cast sculpture lion sculpture (b), both (b produced), both produced for the for industrial the 231 industrial exhibition in Zurich (Switzerland) in 1894. exhibition in Zurich (Switzerland) in 1894. 232 Marker bands for each phase can be identified once all the phases in a Raman map have been 233 Markeridentified bands by comparison for each with phase reference can be spectra. identified These once have all to the be phasesstrong (but in anot Raman necessarily map be have the been 234identified strongest by comparisonsignal in a spectrum) with reference and ideally, spectra. they sh Theseould havenot interfere to be strong with the (but marker not necessarilyband of other be the 235strongest sample signal constituents. in a spectrum) In case of and belite, ideally, the most they intense should band not at interfere 858 cm−1 was with chosen the marker (highlighted band red of other 236 in Figure 5), while for alite, a secondary band at 879 cm−1 was selected (Figure1 5, blue), because there sample constituents. In case of belite, the most intense band at 858 cm− was chosen (highlighted 237 was too much overlap between the main alite band and the spectroscopic1 signature of belite. In the red in Figure5), while for alite, a secondary band at 879 cm − was selected (Figure5, blue), because 238 case of the interstitial calcium aluminoferrite phases, the Raman intensity was integrated over a broad there was too much overlap between the main alite band and the spectroscopic signature of belite. 239 range (688 cm−1–783 cm−1), covering the varying compositions within the solid solution CxAyFz 240In thecausing case of shifts the interstitial of its most calcium prominent aluminoferrite band. (The phases, evaluation the Raman of band intensity shifts was for integrated elucidating over a 1 1 241broad stoichiometric range (688 cm changes− –783 is cm demonstrated− ), covering in theSection varying 3.4.) Tricalcium compositions aluminate within reveals the solid its most solution intense CxA yFz 242causing Raman shifts bands of its at most 360 cm prominent−1, 508 cm− band.1, and 756 (The cm evaluation−1 [38]. Theof latter band is shiftsseen in for some elucidating parts of the stoichiometric chosen 243changes detail is demonstratedas a sharp band in that Section is superimposed 3.4.) Tricalcium to one aluminate of the broad reveals CxAyF itsz bands. most intense(The information Raman bands in at 1 1 1 244360 cmband− , 508shapes cm is− discussed, and 756 below cm− [in38 Sections]. The latter 3.4 and is seen3.5.) Baseline-corrected in some parts of theintensities chosen of detail the marker as a sharp 245 bandbands that isyield superimposed phase distribution to one maps, of the as broadshown Cinx FiguresAyFz bands. 6a and (The6b. The information well-differentiated in band matrix shapes is 246discussed is correlated below inwith Sections slow cooling 3.4 and in 3.5 the.) Baseline-correctedclinker manufacture intensities as a coarse of structure the marker of aluminate bands yield and phase 247 ferrite indicates annealing or a gradual decline in temperature during the solidification of the melt distribution maps, as shown in Figure6a,b. The well-di fferentiated matrix is correlated with slow 248 [15–17]. Likewise, the ragged crystals of dicalcium silicate or belite, respectively, with lamellar cooling in the clinker manufacture as a coarse structure of aluminate and ferrite indicates annealing or 249 extensions into the interstitial matrix that are shown in Figure 6a, suggest very slow in-kiln cooling 250a gradual[15–17]. decline This is ina typical temperature feature of during highlythe hydraulic solidification binders of of the the late melt 19th [15 century,–17]. Likewise, as cement clinker the ragged 251crystals produced of dicalcium batchwise silicate in a orshaft belite, kiln respectively,is inevitably only with discharged lamellar extensions after natural into cooling the interstitial to ambient matrix 252that aretemperature. shown in Figure6a, suggest very slow in-kiln cooling [ 15–17]. This is a typical feature of 253highly hydraulicA comparison binders with of scanning the late 19thelectron century, microsco as cementpy–energy clinker dispersive produced X-raybatchwise spectroscopy in (SEM– a shaft kiln 254is inevitablyEDX) maps only of discharged the same sample after naturalexcerpt coolingconfirms to the ambient correct temperature.assignment of Raman signatures to 255 Acalcium comparison aluminoferrite with and scanning calcium aluminate electron microscopy–energyphases. Figure 7 reveals dispersivethe elemental X-ray distributions spectroscopy of 256(SEM–EDX) aluminium maps (green) of the and same iron sample(red) in comparison excerpt confirms with a Raman the correct map specifically assignment showing of Raman the intensity signatures to 257 distributions of the marker bands that are assigned to C3A (green) and CxAyFz (red), in which the first calcium aluminoferrite and calcium aluminate phases. Figure7 reveals the elemental distributions of aluminium (green) and iron (red) in comparison with a Raman map specifically showing the intensity distributions of the marker bands that are assigned to C A (green) and CxAyFz (red), in which the first 3 corresponds to structures containing Al, while the latter correctly relates to areas emitting EDX signals of both Fe and Al (Figure7c; yellow, as mixed colour of red and green). Heritage 2019, 2 FOR PEER REVIEW 9 Heritage 2019, 2 1670 258 corresponds to structures containing Al, while the latter correctly relates to areas emitting EDX 259

260

261 Figure 7. Comparison of scanning electron microscopy–energy dispersive X-ray spectroscopy (SEM– Figure 7. Comparison of scanning electron microscopy–energy dispersive X-ray spectroscopy 262 EDX) maps showing the elementary distributions of aluminium (a), iron (b) and an overlay of both (SEM–EDX) maps showing the elementary distributions of aluminium (a), iron (b) and an overlay 263 (c) with the according Raman map of C3A and CxAyFz confirming the correct assignments of Raman 264 of bothspectra (c) with to calcium the according aluminate Raman (without map Fe) of and C3 caAlcium and CaluminoferritexAyFz confirming (with Fe) the phases. correct The assignments data 265 of Ramancontained spectra in ( tod) is calcium a subset aluminateof the map shown (without in Figure Fe) and6b. calcium aluminoferrite (with Fe) phases. The data contained in (d) is a subset of the map shown in Figure6b. 266 3.2. Raman Spectra Enable the Discrimination of Polymorphs of Cement Clinker Phases

2673.2. RamanLike Spectra C3S, Enabledicalcium the silicate Discrimination occurs in different of Polymorphs polymorphs: of Cement The five Clinker known Phases crystalline forms are 268 designated α, α’, β, and γ, with the α’ polymorph existing in the two very similar forms α’H and α’L 269 Like[43]. C If3 S,the dicalcium cement clinker silicate is slowly occurs cooled in from different elevated polymorphs: temperatures, The the α five’L–β knownphase transition crystalline and forms 270are designatedthus the decreaseα, α’, β ,in and crystalγ, with symmetry the α’ from polymorph orthorhombic existing to monoclinic in the two frequently very similar leads forms to theα ’h and 271α’l [43].formation If the cement of polysynthetic clinker is twinning. slowly cooled H. Insley from proposed elevated a cl temperatures,assification of C the2S orα ’belitel–β phase grains transitionin 272and thuscement the decreaseclinker that in is crystal based symmetryon the characteristics from orthorhombic of their internal to monoclinicmicrostructure: frequently Type I belite leads is to the 273 defined as rounded grains showing two or more sets of intersecting lamellae (α and β), whereas type formation of polysynthetic twinning. H. Insley proposed a classification of C2S or belite grains in 274 II belite forms irregular grains with one set of distinct parallel striations due to the martensitic cement clinker that is based on the characteristics of their internal microstructure: Type I belite is 275 transformation. The nonlamellar type III belite, which occurs as overgrowth on type I or as individual 276defined crystals, as rounded characterises grains the showing Portland twocement or clinker more setsproduced of intersecting in a shaft kiln lamellae [15,47–49]. (α Theand Ramanβ), whereas 277type IImap belite in formsFigure 6b irregular depicts a grainscross-hatched with one belite set grain of distinctof the type parallel I that consists striations exclusively due to of the lamellae martensitic 278transformation. of the β-phase The intergrown nonlamellar with type exsolved III belite, calcium which aluminoferrite occurs as overgrowthsolid solution; on the type former I or α as-phase individual 279crystals, was characterises completely inverted. the Portland cement clinker produced in a shaft kiln [15,47–49]. The Raman map 280 in Figure6bAs depicts the processing a cross-hatched parameters belite affect the grain phase of assemblage the type I thatand microstructure consists exclusively of cement of clinker, lamellae of 281 it is possible to deduce the temperatures of belite formation and cooling rates on the basis of the β-phase intergrown with exsolved calcium aluminoferrite solid solution; the former α-phase was completely inverted. As the processing parameters affect the phase assemblage and microstructure of cement clinker, it is possible to deduce the temperatures of belite formation and cooling rates on the basis of polymorphic phase transitions. In Raman imaging experiments, the γ form of C2S can be clearly distinguished from 1 the β polymorph due to its characteristic Raman spectrum [50]. The main band of γ-C2S at 813 cm− is particularly distinct from the spectroscopic signature of β-C2S, and with the high spectral resolution HeritageHeritage2019, 20192 , 2 FOR PEER REVIEW 10 1671

282 polymorphic phase transitions. In Raman imaging experiments, the γ form of C2S can be clearly 283 distinguished from the β polymorph due to its characteristic Raman spectrum [50]. The main band that wasHeritage employed 2019, 2 FOR in PEER this REVIEW study, the presence of the γ polymorph can be additionally confirmed10 based γ −1 β 284 of -C2S at 813 cm is particularly1 distinct from the spectroscopic signature of -C2S, and with the 1 on its secondary band at 837 cm− typically appearing as a shoulder of the β-C2S band at 845 cm− 285 high spectral resolution that was employed in this study, the presenceγ of the γ polymorph can be 282(see Figurepolymorphic8), while phase a distincttransitions. peak In atRaman 837 cmimaging1 is obtainedexperiments, at athe few form relatively of C2S purecan beγ -CclearlyS grains. 283286 additionally confirmed basedβ on its secondary −band at 837 cm−1 typically appearing as a shoulder2 of Ramandistinguished intensity maps from ofthe the polymorph most prominent due to its bands characteristic of both polymorphs Raman spectrum reveal [50]. their The di mainfferent band spatial β 2 −1 −1 284287 ofthe γ -C-C2S Sat band 813 cm at −1845 is particularlycm (see Figure distinct 8), fromwhile the a distinctspectroscopic peak atsignature 837 cm of is β -Cobtained2S, and withat a fewthe distributions (Figureγ 9a,b). 285288 highrelatively spectral pure resolution -C2S grains. that Raman was employ intensityed mapsin this of study, the most the prominentpresence of bands the γ of polymorph both polymorphs can be 286289 additionally confirmed based on its secondary band at 837 cm−1 typically appearing as a shoulder of 287 the β-C2S band at 845 cm−1 (see Figure 8), while a distinct peak at 837 cm−1 is obtained at a few 288 relatively pure γ-C2S grains. Raman intensity maps of the most prominent bands of both polymorphs 289

290

291 Figure 8. Raman spectra of C2S, which include spectral signatures of both, the β (labelled red) and γ Figure 8. β γ 292 polymorphRaman (green) spectra of ofthis C clinker2S, which phase. include The inset spectral reveals signatures a deconvolution of both, of the spectrum(labelled (1) into red) five and 290293 polymorphLorentzian (green) peak ofprofiles, this clinker while trace phase. (2) shows The inset the exception reveals a of deconvolution a spectrum of relatively of spectrum pure (1)γ-C2 intoS. five Lorentzian peak profiles, while trace (2) shows the exception of a spectrum of relatively pure γ-C2S. 291 Figure 8. Raman spectra of C2S, which include spectral signatures of both, the β (labelled red) and γ 294 The crystallisation of the γ polymorph at the microstructural level is directly related to the 292 polymorph (green) of this clinker phase. The inset reveals a deconvolution of spectrum (1) into five 295 macroscopic (and hydraulic) properties of the resulting clinker, as changes in density due to the 293 TheLorentzian crystallisation peak profiles, of the whileγ polymorph trace (2) shows at th thee exception microstructural of a spectrum level of relatively is directly pure γ-C related2S. to the 296 transformation of monoclinic β-C2S into orthorhombic γ-C2S during cooling leads to disintegration macroscopic (and hydraulic) properties of the resulting clinker, as changes in density due to the 294297 or ‘dusting’The crystallisation [43]. The related of the crack γ polymorph formation atis faintlythe micr visibleostructural in the Ramanlevel is map directly in Figure related 9a toand the is transformation of monoclinic β-C2S intoγ orthorhombic γ-C2S during cooling leads to disintegration 295298 macroscopicconfirmed by (and an hydraulic)overlay of properthe -Cties2S ofRaman the resulting intensity clinker, distribution as changes onto ina densityscanning due electron to the 299or ‘dusting’ [43]. The related crack formation is faintly visible in the Raman map in Figure9a and is 296 transformationmicrograph of the of monoclinicsame clinker β -Cremnant2S into (Figureorthorhombic 10). γ-C2S during cooling leads to disintegration confirmed by an overlay of the γ-C S Raman intensity distribution onto a scanning electron micrograph 297 or ‘dusting’ [43]. The related crack2 formation is faintly visible in the Raman map in Figure 9a and is of the same clinker remnant (Figure 10). 298 confirmed by an overlay of the γ-C2S Raman intensity distribution onto a scanning electron 299

300

301 Figure 9. Raman intensity maps showing the distributions of two different polymorphic forms of 302 dicalcium silicate based on their characteristic marker bands in Portland cement clinker remnants 300303 found in the binder matrix of a prefabricated concrete element (a) and a cast lion sculpture (b). 301 Figure 9. Raman intensity maps showing the distributions of two different polymorphic forms of Figure 9. Raman intensity maps showing the distributions of two different polymorphic forms of 302 dicalcium silicate based on their characteristic marker bands in Portland cement clinker remnants 303 dicalciumfound silicatein the binder based matrix on their of a prefabricated characteristic concrete marker element bands (a in) and Portland a cast lion cement sculpture clinker (b). remnants found in the binder matrix of a prefabricated concrete element (a) and a cast lion sculpture (b). Heritage 2019, 2 1672 Heritage 2019, 2 FOR PEER REVIEW 11

304

305 Figure 10. Scanning electron micrograph acquired by employing backscattered electron (BSE) Figure 10. Scanning electron micrograph acquired by employing backscattered electron (BSE) detection 306 detection without (a) and with superimposed Raman distribution map of γ-C2S (b) demonstrating the without (a) and with superimposed Raman distribution map of γ-C S(b) demonstrating the effect of 307 effect of conversion from the β into the γ polymorph onto crack formation.2 conversion from the β into the γ polymorph onto crack formation. 308 3.3. Changes in Relative Band Intensities Visualise Different Crystal Orientations 3.3. Changes in Relative Band Intensities Visualise Different Crystal Orientations 309 We now come to physical material properties after having demonstrated that Raman spectra are 310 Wecharacteristically now come to affected physical by material chemical propertiesidentity and after crystal having structure demonstrated of clinker phases. that Raman The effect spectra of are 311characteristically crystal orientations affected onto by Raman chemical spectra identity is known and since crystal the early structure years ofof clinkerRaman spectroscopy phases. The [51]. effect of 312crystal Indeed, orientations depending onto on Ramantechnical spectra developments, is known such since as the the confocal early years microscope of Raman and laser, spectroscopy it took a [51]. 313 few decades until the application of this effect with microprobe setups [52], and some more years to Indeed, depending on technical developments, such as the confocal microscope and laser, it took a 314 realise the first orientation-distribution maps of polycrystalline materials surfaces [53]. A recent study few decades until the application of this effect with microprobe setups [52], and some more years 315 of combined electron backscatter diffraction (EBSD) and Raman microspectroscopic imaging 316to realisedemonstrated the first orientation-distributionthe ability of the latter to map maps the of sample polycrystalline surfaces with materials high pixel surfaces numbers [53 ].and A to recent 317study obtain of combined quantitative electron orientation backscatter data of each diffraction individual (EBSD) crystallite and [13]. Raman microspectroscopic imaging 318demonstrated We have the to ability make of clear the again latter that to map the Raman the sample spectrum surfaces of a crystalline with high material pixel numbers consists of and several to obtain 319quantitative bands representing orientation different data of eachvibrational individual motions crystallite and their [13 characteristic]. frequencies in order to 320 Weexplain have the to effect. make Each clear vibration again that has the certain Raman symme spectrumtry (with of respect a crystalline to the symmetry material consistsof the crystal) of several 321bands and representing main direction diff erentrelative vibrational to the crystal motions axes. The and laser their employed characteristic for excitation frequencies is typically in order linearly to explain 322the effpolarised;ect. Each thus, vibration the excitation has certain also has symmetry a certain (withdirection respect in space. to the Theoretically, symmetry the of theRaman crystal) scattering and main 323 intensity depends on the dipole moment that is induced in the molecular orbitals of an oscillating direction relative to the crystal axes. The laser employed for excitation is typically linearly polarised; 324 specimen and, as a result, on the change of polarisability during the vibrational motion [54]. thus, the excitation also has a certain direction in space. Theoretically, the Raman scattering intensity 325 According to a more practical explanation the resulting light intensity of each band in a spectrum 326depends depends on the of dipolethe relative moment orientations that is induced of laser inpolarisation the molecular direction, orbitals excited of an crystal oscillating vibration, specimen and and, 327as a result,polarisation on the direction change of of the polarisability scattered light, during meanin theg that vibrational crystal vibrations motion having [54]. their According main motion to a more 328practical parallel explanation to the given the laser resulting polarisation light direction intensity are of eachexcited band very in efficiently, a spectrum while depends the same of mode the relative is 329orientations less excited of laser if the polarisation crystal orientation direction, direction excited is changed. crystal vibration, Changes in and relative polarisation band intensities direction are of the 330scattered observed light, in meaning the case thatof varying crystal crystal vibrations orientations, having theiras each main band motion in the parallel spectrum to of the a givensingle laser 331polarisation compound direction reflects a are different excited vibrational very effi ciently,motion. while the same mode is less excited if the crystal β 332orientation Figure direction 11 shows is changed. the Raman Changes spectra of in -C relative2S acquired band within intensities the mapping are observed analysis inof thetype case I of 333 belite grains and normalised to the intensity of the most prominent peak. In contrast to the evaluation varying crystal orientations, as each band in the spectrum of a single compound reflects a different 334 of marker bands for different phases (Figure 6b) or polymorphs (Figure 9b), the intensities of three vibrational motion. 335 bands are analysed here, all of which assigned to the same polymorphic phase, i.e., monoclinic belite Figure 11 shows the Raman spectra of β-C S acquired within the mapping analysis of type I belite 336 or β-C2S, respectively. Because during the mapping2 experiment the polarisation direction of the laser 337grains was and kept normalised constant, relative to the intensities intensity of of β the-C2S mostbands prominent only change peak. due to Indifferent contrast crystal to the orientations. evaluation of marker bands for different phases (Figure6b) or polymorphs (Figure9b), the intensities of three bands are analysed here, all of which assigned to the same polymorphic phase, i.e., monoclinic belite or β-C2S, respectively. Because during the mapping experiment the polarisation direction of the laser was kept constant, relative intensities of β-C2S bands only change due to different crystal orientations.

Heritage 2019, 2 1673 Heritage 2019, 2 FOR PEER REVIEW 12

338 339 Figure 11. Spectra from the Raman map of the Portland cement mortar of a cast lion sculpture (Zurich, Figure 11. Spectra from the Raman map of the Portland cement mortar of a cast lion sculpture (Zurich, 340 Switzerland, 1894) showing the effect of different crystal orientations onto the relative intensities of Switzerland, 1894) showing the effect of different crystal orientations onto the relative intensities of the 341 the five labelled bands of β-C2S. five labelled bands of β-C2S. 342 While the images based on intensities of the individual bands (Figures 12a, 12b, 12c, and Figures While the images based on intensities of the individual bands (Figures 12a–c and 13a–c) 343 13a, 13b, 13c) demonstrate the independent behaviour of three β-C2S vibrational modes, the overlays β 344 demonstratein Figures 12d the and independent 13d provide behaviour deep insight of three into -Cthe2S complex vibrational microstructures modes, the overlays of the belite in Figures grains 12 ind 345 andthese 13 samplesd provide of deep 19th insight century into Portland the complex cement microstructures mortars. The of overlay the belite contains grains inthe these mixed samples colours of 346 19thyellow century (red–green), Portland purple cement (blue–red), mortars. and The turquoise overlay contains (green–blue); the mixed crystals colours that yellowappear (red–green),in the same purple (blue–red), and turquoise (green–blue); crystals that appear in the same (mixed) colour have 347 (mixed) colour have the same orientation. This enables better differentiation of discrete β-C2S crystals, β 348 theelucidating same orientation. the microstr Thisuctural enables protrusions better diff oferentiation β-belite into of discrete the calcium-C2S aluminoferrite crystals, elucidating matrix the as microstructural protrusions of β-belite into the calcium aluminoferrite matrix as extensions of crystals 349 extensions of crystals having different orientations than the surrounding C2S (Figure 12d, baluster), 350 havingas well diasff erentthe striations orientations as composed than the surrounding of cross lamellae C2S (Figure arranged 12d, in baluster), angles of as approx. well as 45° the striationsand 135°, 351 asrespectively, composed ofand cross different lamellae layers arranged (Figure in 13d, angles cast of lion approx. sculpture). 45◦ and 135◦, respectively, and different 352 layersIt (Figureseems misleading13d, cast lion that, sculpture). in Figure 13d, the almost horizontal lamella on top and the close to 353 verticalIt seems one on misleading the left have that, almost in Figure the same 13d, red the co almostlour. A horizontal closer look lamella exposes on this top as and coincidence. the close toA 354 verticallook at onethe onnumbers the left haveand, almostfor exampl the samee, the red intensity colour. Aratio closer of look the exposes858-cm− this1 and as coincidence.the 975-cm−1 Abands look 1 1 355 atreveals the numbers small but and, significant for example, differences the intensity between ratio the of thealmost 858-cm horizontal− and the and 975-cm vertical− bands lamellae reveals and small but significant differences between the almost horizontal and vertical lamellae and demonstrates 356 demonstrates the influence of the β-C2S crystals that are arranged below. As estimated in Section 2.2, β 357 thethe influencedepth resolution of the -Cof2 Sthis crystals experiment that are is arranged approx. below.22.5 µm As and estimated light absorption in Section 2.2in ,the the almost depth µ 358 resolutiontransparent of belite this experiment grains is not is expected approx. 22.5to significantlym and light affect absorption it. Thus, in the the colours almost in transparent the overlay belite map 359 grainsare also is mixed not expected between to the significantly superimposed affect layers, it. Thus, mo thestly colours leading in to the these overlay seemingly map arecontradictious also mixed 360 betweenred colourings. the superimposed layers, mostly leading to these seemingly contradictious red colourings.

Heritage 2019, 2 1674 Heritage 2019, 2 FOR PEER REVIEW 13 Heritage 2019, 2 FOR PEER REVIEW 13

361

361 β −1 −1 −1 362 Figure 12. Raman intensity maps of the -C2S bands at 858 cm (a),1 845 cm (b),1 874 cm (c), 1and an Figure 12. Raman intensity maps of the β-C2S bands at 858 cm− (a), 845 cm− (b), 874 cm− (c), and 363362 overlayFigure 12. of theseRaman intensities intensity plotted maps of in thered, β green,-C2S bands and blue, at 858 respectively cm−1 (a), 845 (d). cm This−1 (b), relative 874 cm intensity−1 (c), and map an an overlay of these intensities plotted in red, green, and blue, respectively (d). This relative intensity 364363 revealsoverlay differentof these intensities crystal orientatio plotted inns red,in a green, binder an remnantd blue, respectively in the matrix (d). of This a prefabricated relative intensity concrete map map reveals different crystal orientations in a binder remnant in the matrix of a prefabricated concrete 365364 elementreveals different(Zurich, Switzerland,crystal orientatio 1894).ns in a binder remnant in the matrix of a prefabricated concrete element (Zurich, Switzerland, 1894). 365 element (Zurich, Switzerland, 1894).

366 366 367 Figure 13. Raman intensity maps of the β-C2S bands at 858 cm−1 (a), 845 cm−1 (b), 874 cm−1 (c), and an 368 overlay of these intensities plotted in red,β green, and blue, respectively−1 (d). This−1 relative intensity−1 map 367 Figure 13. Raman intensity maps of the -C2S bands at 858 cm (1a), 845 cm (b1), 874 cm (c), 1and an Figure 13. Raman intensity maps of the β-C2S bands at 858 cm− (a), 845 cm− (b), 874 cm− (c), and 369368 revealsoverlay differentof these intensities crystal orientations plotted in red,in a green,residual and ce mentblue, respectivelyclinker grain ( din). theThis binder relative matrix intensity of a mapcast an overlay of these intensities plotted in red, green, and blue, respectively (d). This relative intensity 370369 lionreveals sculpture different (Zurich, crystal Switzerland, orientations 1894). in a residual cement clinker grain in the binder matrix of a cast map reveals different crystal orientations in a residual cement clinker grain in the binder matrix of a 370 lion sculpture (Zurich, Switzerland, 1894). 371 cast lion sculpture (Zurich, Switzerland, 1894).

371

Heritage 2019, 2 1675

3.4. Band Shifts Elucidate Stoichiometric Changes and Members of Solid Solution Series In Figure6, all the calcium aluminoferrite phases are summarised under the general formula CxAyFz or (CaO)x(Al2O3)y(Fe2O3)z, respectively. In studies that are based on classical petrographic analysis, these interstitial phases are often subsumed as brownmillerite or Ca2AlFeO5, neglecting other members of the solid solution series C4AxF1–x or Ca2(AlxFe1–x)O5, respectively, which are difficult to discriminate by polarised light or electron microscopies. We recently demonstrated, based on pure synthesised phases, that the different stoichiometries of calcium aluminoferrites can be distinguished due to composition-induced shifts of their Raman bands, which are most pronounced in the case of 1 1 the typically most intense and broad peak in the range of 731 cm− (C6AF2) via 742 cm− (C4AF) to 1 756 cm− (C6A2F) [37,38]. In simple terms, this can be understood as an effect of the atomic masses that are involved in the crystal vibrations: The replacement of iron with the lighter aluminium leads to an increase of the vibrational frequencies and wavenumbers, respectively. Also distinctive are the 1 1 1 1 low-frequency modes of these phases at 256 cm− , 310 cm− (both C6AF2); 259 cm− , 314 cm− (C4AF); 1 1 and, 263 cm− , 318 cm− (C6A2F). The intensities of the different CxAyFz bands cannot be individually evaluated within Raman mapping data comprising different members of this solid solution series due to their gradual shifts and large widths. Additionally, peak fitting failed because of the large width and overlap with neighbouring 1 1 bands, as, for example, the main belite signature from approx. 840 cm− to 880 cm− and another broad 1 band at approx. 650 cm− , which is often found to be associated with interstitial calcium aluminoferrite phases in cement clinkers and whose nature is not yet fully understood [2]. Thus, the only way to visualise different stoichiometries of these phases is to plot the wavenumber positions of the most prominent calcium aluminoferrite bands as a function of the lateral coordinates. Figure 14a,c show sample spectra as compared with the reference spectra of pure synthesised phases (labelled ‘ref.’), and Figure 14b,d are the results of these band shift plots. Please note that no peak fitting was performed, and thus these Raman band shift maps are composed of wavenumber positions of the highest features in each of the spectral profiles, and thus random fluctuations can be induced by noise superimposed to these broad bands (therefore, having relatively flat peak maxima). Changes in the baseline and the neighbouring overlapping bands might also contribute. These effects are the main reason for fluctuations within the blue colour range in Figure 14b,d, which is mainly assigned to C6AF2, but might also include sub-stoichiometric compositions. In contrast, it is intriguing that the maps contain features having the exact composition of brownmillerite, i.e., C4AF, which are well distinct from surrounding phases by their clearly defined boundaries. The interstitial matrix in Figure 14b consists of C6AF2 (with some C4AF crystallites) in the bottom half of the map and of C6A2F (also including C4AF) close to the top border, being characterised by a 1 band maximum at 756 cm− . A close look at the spectra (examples in Figure 14c) from the Raman map 1 in Figure 14d exposes that the band positions at 756 cm− in this case should be assigned to tricalcium aluminate or C3A, respectively. Sharing the same wavenumber position of their most prominent bands, C6AF2 and C3A, can be distinguished due to the significantly differing widths of these modes. The confirmation of these assignments is possible based on the appearance of other significant bands 1 1 in the lower-frequency range, e.g., the secondary C3A modes at 508 cm− and 360 cm− , and Figure7, as discussed above in Section 3.1, underlines this interpretation. The full widths at half maximum 1 (FWHMs) resulting from Gauss fits of the reference spectra shown in Figure 14a,c are approx. 45 cm− 1 for C6AF2 and approx. 12 cm− in the case of C3A. This demonstrates that information on material properties might also be included in band shapes, which is the main topic of the following Section 3.5. Heritage 2019, 2 1676

Heritage 2019, 2 FOR PEER REVIEW 15

418 419 Figure 14. Raman band shift maps revealing the wavenumbers of the peak maxima within a range of Figure 14. Raman band shift maps revealing the wavenumbers of the peak maxima within a range of 420 715 cm−1 to 762 cm−1 enabling the discrimination of individual members of the calcium aluminoferrite 1 1 421 715 cmsolid− tosolution 762 cm series− enabling in binder the remnants discrimination in the cement of individual stone of a membersbaluster (a of) and the a calcium cast lionaluminoferrite sculpture 422 solid( solutionb). In the series latter inexample, binder the remnants relatively in sharp the cement band at stone 756 cm of− a1 is baluster assigned (a ,tob )C and3A instead a cast lionof C6 sculptureA2F, 1 423 (c,d).which In the is latter confirmed example, by Figure the relatively 7. sharp band at 756 cm− is assigned to C3A instead of C6A2F, which is confirmed by Figure7. 424 From the band shift map in Figure 14d, we not only can deduce that the type I belite grain with 425 multidirectional lamellae is embedded in a matrix that mainly consists of C6AF2. Apparently, the From the band shift map in Figure 14d, we not only can deduce that the type I belite grain 426 exsolution of impurities from the parent belite following the α–α’ polymorphic phase transformation, with multidirectional lamellae is embedded in a matrix that mainly consists of C6AF2. Apparently, 427 the so-called remelting reaction [34,55–56], led to the crystallisation of C4AF on the boundaries of the 428the exsolutionlamellae during of impurities slow cooling. from the Different parent interpretations belite following regarding the α–α ’the polymorphic crystallinity phase of this transformation, solidified 429the so-calledexsolved remeltingliquid do exist: reaction While [34 several,55,56 ],studies led to as thesign crystallisation the relatively broad of C4 AFband on at the approx. boundaries 750 cm− of1 the 430lamellae to crystalline during slow C4AF cooling. [32,34,37,38], Diff erenta broad interpretations band shifted to regarding approx. 730 the cm crystallinity−1 is interpreted of this as being solidified 1 431exsolved indicative liquid for do amourphous exist: While calcium several aluminoferrite studies assign, as the this relatively band was broad also band observed at approx. after laser 750cm − 1 432to crystallineirradiation C of4AF crystalline [32,34,37 C,438AF], at a elevated broad band power shifted [34]. When to approx. looking 730 at the cm data− is that interpreted are presented as being 433indicative here, this for amourphousband might also calcium be assigned aluminoferrite, to C6AF2 and as thismight band be the was result also of observed a segregation after laserreaction, irradiation as 434 iron-containing phases can be easily heated and chemically transformed by the action of high laser of crystalline C4AF at elevated power [34]. When looking at the data that are presented here, this band 435 powers in Raman microspectroscopy setups (see Ref. [37] (p. 11) and references cited therein). might also be assigned to C AF and might be the result of a segregation reaction, as iron-containing 436 Comparison with reference6 data2 of crystalline calcium aluminoferrite phases in Figures 14a and 14c phases can be easily heated and chemically transformed by the action of high laser powers in Raman 437 does not show significant increases of band widths of the sample spectra, indicating their 438microspectroscopy crystallisation upon setups the cooling (see Ref. of the [37 exsolved] (p. 11) liquid. and references cited therein). Comparison with 439reference data of crystalline calcium aluminoferrite phases in Figure 14a,c does not show significant increases of band widths of the sample spectra, indicating their crystallisation upon the cooling of the exsolved liquid.

Heritage 2019, 2 1677

3.5. Band Widths are Measures for Crystallinities and Provide Access to the Thermal History of Medieval Gypsum Mortars High-fired medieval gypsum mortars comprise not only calcium sulphate phases dehydrated to different degrees, but also primary accessory minerals of the gypsum deposit as well as phases newly formed through pyrometamorphic reactions. The observation of high-temperature, low-pressure mineral transformations, and the correlation of coexisting phases occurring in unhydrated binder relicts preserved in the aged gypsum matrix to the mineralogy of the raw material and the burning regime constitute the only source to the historical technological know-how because of the absence of medieval textbooks. So far, the anhydrite morphology has been used to roughly estimate the process parameters, as the optical and electron microscopy of modern high-fired gypsum yielded the observation that the proportion of granular anhydrite, considered as hardly soluble, rises with increasing the firing temperatures at the expense of columnar anhydrite, which preferably dissolves and hydrates [57–60]. Recent studies have demonstrated that Raman microspectroscopy can trace the burning histories of individual anhydrite grains in medieval gypsum mortars by combining high structural sensitivity with spatial resolution in an almost unique fashion. Laboratory-scale burning experiments of gypsum powders have shown that anhydrite II dominates the product mixture in the temperature range from approx. 300 ◦C to 1180 ◦C (i.e., the transition temperature to the metastable anhydrite I), and it is the only phase at burning temperatures 500 C. In the range of 400 C to 900 C significant narrowing of ≥ ◦ ◦ ◦ the anhydrite II Raman bands as well as increase in crystallite sizes, the reduction of crystal stress and strain (in powder X-ray diffraction studies), and the decrease of specific surface area (according to the Brunauer-Emmett-Teller method) were observed. At higher burning temperatures, further but much smaller changes of Raman band widths are determined, which are in the range of the values of natural anhydrite [61–63]. The narrowing of Raman bands can be understood as a consequence of the healing of crystal lattice defects and the according increase of crystallinity. Raman band widths (typically expressed as 1 FWHMs) of the most prominent anhydrite II mode at 1017 cm− in the room-temperature spectra of burning products follow a decreasing sigmoidal trend in the range of 400 ◦C to 1100 ◦C. This trend can be applied as a calibration curve enabling the measurement of burning temperatures, which individual anhydrite crystallites have experienced, with an uncertainty of approx. 50 K [62,63]. ± In the polished thin-sectional samples of high-fired medieval gypsum mortars, anhydrite grains can be found by polarised light or electron microscopies, either in remnant clusters of firing products or spread as individual crystallites within the gypsum matrix. Raman maps of selected areas reveal phase compositions and their distributions (according to Section 3.1) along with Raman band width distributions of anhydrite II, which can be evaluated via comparison with reference data from the burning experiments to yield the approximate burning temperatures. Figure 15 displays the electron and light micrographs of an assemblage of firing products in the hydrated binder matrix of a high-fired medieval gypsum mortar. Experienced scientists can identify such clusters by both techniques. In electron microscopy with backscattered electron (BSE) detection, image contrast depends on the atomic numbers of the involved chemical elements. In the shown example, the black areas represent the organic embedding resin (mainly consisting of the elements C, H, and O), while the brightest features are celestine crystals (SrSO4, containing the high-atomic-number element strontium). Surrounded by gypsum (CaSO 2H O), anhydrite (CaSO ) grains appear to be 4· 2 4 slightly brighter due to their higher average atomic number. In polarised light micrographs that are captured with crossed Nicols, anhydrite can be distinguished from gypsum due to different (relief and) birefringence. Heritage 2019, 2 1678 Heritage 2019, 2 FOR PEER REVIEW 17

486 487 Figure 15. High-fired gypsum mortar originating from the Chapel of the Holy Cross of the Figure 15. High-fired gypsum mortar originating from the Chapel of the Holy Cross of the Benedictine 488 Benedictine convent of St. John in Müstair (Switzerland): scanning-electron (a) and polarised light convent of St. John in Müstair (Switzerland): scanning-electron (a) and polarised light micrographs 489 micrographs acquired with parallel (b) and crossed Nicols (c). The rectangles highlight the area acquired with parallel (b) and crossed Nicols (c). The rectangles highlight the area corresponding to 490 corresponding to the Raman map presented in Figure 16. the Raman map presented in Figure 16.

491 The Raman spectrum of anhydrite II includes bands at 124 cm−1, 133 cm−1, 152 cm−1, 169 cm−1, 233 1 1 1 1 492 cm−1,The and Raman 269 cm spectrum−1, corresponding of anhydrite to vibrational II includes crystal-latti bands at 124ce cmmodes,− , 133 while cm− the, 152 vibrations cm− , 169 of cm the− , 1 1 493 233sulphate cm− ,ion and are 269 reflected cm− , corresponding by bands at 417 to vibrationalcm−1, 500 cm crystal-lattice−1, 609 cm−1, 628 modes, cm−1 while, 676 cm the−1, vibrations 1000 cm−1, of 1017 the 1 1 1 1 1 1 494 sulphatecm−1, 1110 ion cm are−1, 1129 reflected cm−1, byand bands 1160 atcm 417−1. Figure cm− , 50016a provides cm− , 609 a cm close− , look 628 cm at −the, 676most cm intense− , 1000 Raman cm− , 1 1 1 1 495 1017band cm of −anhydrite, 1110 cm II− at, 11291017 cmcm−−1,. andThe 1160different cm− band. Figure shapes 16a that provides were aobtained close look at atroom the mosttemperature intense 496 are characteristic of the burning temperatures that were employed during the synthesis of anhydrite 497 from gypsum powder. Band widths, expressed as FWHMs, were determined by fitting Lorentz

Heritage 2019, 2 1679

1 Raman band of anhydrite II at 1017 cm− . The different band shapes that were obtained at room temperature are characteristic of the burning temperatures that were employed during the synthesis of anhydrite from gypsum powder. Band widths, expressed as FWHMs, were determined by fitting LorentzHeritage functions 2019, 2 FOR to PEER the peak REVIEW profiles of 64 individual Raman measurements on different crystallites18 of each temperature step. The FWHMs that are listed in Figure 16a are the mean values and standard 498 functions to the peak profiles of 64 individual Raman measurements on different crystallites of each deviations of these 64 measurement values, revealing the described steady decrease in band widths 499 temperature step. The FWHMs that are listed in Figure 16a are the mean values and standard 500with increasingdeviations of temperatures these 64 measurement [62,63]. values, revealing the described steady decrease in band widths 501

502

503 Figure 16. The shapes of the most prominent band at 1017 cm−1 in the Raman spectra of anhydrite II Figure 16. The shapes of the most prominent band at 1017 cm 1 in the Raman spectra of anhydrite II 504 burnt from gypsum powder at different temperatures reveal a− correlation between their full widths 505 burntat from half maximum gypsum powder(FWHMs) at and diff theerent calcination temperatures temperatures reveal (a). a correlation Raman maps between of a remnant their fullbinder widths 506 at halfaggregate maximum in a (FWHMs) high-fired andmedieval the calcination gypsum mortar temperatures give access (a). to Raman the burning maps ofhistory: a remnant relative binder 507 aggregateintensities in a high-fired of two anhydrite medieval modes gypsum (B) reveal mortar differe givent access crystal to orientations, the burning the history: intensities relative of marker intensities 508 of twobands anhydrite provide modes the distributions (b) reveal diofff anhydriteerent crystal and gyp orientations,sum (c), and the finally intensities the band of markerwidth distribution bands provide 509 the distributionsmap of the anhydrite of anhydrite II mode and at 1017 gypsum cm−1 relates (c), and to a finally burning the temperature band width of approx. distribution 750 °C ± map 50K. of the 1 anhydrite II mode at 1017 cm− relates to a burning temperature of approx. 750 ◦C 50 K (d). 510 Note that the given numbers are exactly valid only for the actual Raman instrument± and the 511 measurement parameters (see Section 2.2), including, for example, a laser wavelength of 532 nm, a 512 100 µm wide spectrometer entrance slit, and a diffraction grating having 1800 lines per millimetre 513 [63]. A detailed description of the analytical method, comprising the measurement parameters and Heritage 2019, 2 1680

Note that the given numbers are exactly valid only for the actual Raman instrument and the measurement parameters (see Section 2.2), including, for example, a laser wavelength of 532 nm, a 100 µm wide spectrometer entrance slit, and a diffraction grating having 1800 lines per millimetre [63]. A detailed description of the analytical method, comprising the measurement parameters and data evaluation, as well as the correction of instrument-related broadening, can be found in Ref. [63]. Further application examples and details of the method are given in Ref. [62]. The shown example of a gypsum–anhydrite cluster (Figure 15) was mapped on an area of 75 µm 44 µm, with a step size of 1 µm 1 µm (Figure 16b–d). The resulting spectra only contained × × 1 1 the signatures of gypsum (most prominent band at 1008 cm− ) and anhydrite II (1017 cm− ). According to the previous Sections 3.1 and 3.3, the phase- and orientation-distribution maps were calculated by 1 evaluation of band intensities. In Figure 16b, the intensities of the main anhydrite band at 1017 cm− 1 (red pixels) and of a secondary mode of the same phase appearing at 1129 cm− (green) are plotted as a function of the mapping coordinates. This Raman map confirms the lath-shaped anhydrite crystals that are also seen in the electron and light micrographs in Figure 15, and the different crystal orientations, resulting in different (mixed) colours, enable a better identification of individual crystallites. The Raman 1 1 intensity map of the marker bands of gypsum (1008 cm− , green) and anhydrite II (1017 cm− , red) in Figure 16c is also in agreement with the microscopic observations. The Raman band width evaluations were restricted to relatively strong bands, because the weak bands, in some cases, do not yield meaningful results in the peak fitting procedure and their obtained widths are often significantly affected by the neighbouring and partly overlapping gypsum band (see Refs. [62,63] for details). Thus, not all anhydrite crystallites shown in Figure 16b were evaluated by 1 1 peak fitting. The resulting band widths in Figure 16d range from 3.41 cm− to 4.02 cm− , and their 1 average (n = 545 pixels or spectra, respectively, were evaluated) of 3.70 cm− is an FWHM between those that are assigned to 800 ◦C and 700 ◦C in Figure 16a (according to Ref. [63] the value determined for 750 C is 3.63 0.035), which is interpreted in the sense that the examined thermal anhydrite grains ◦ ± experienced calcination temperatures of approx. 750 C 50 K. ◦ ± 4. Conclusions The potential of Raman microscopy in the analysis of architectural heritage has been demonstrated by imaging unhydrated binder remnants in historical mortars. This approach is a very interesting complementation of the existing tools for petrographic analysis, because Raman microspectroscopic data combine high structural information content with spatial resolution on the micrometre scale, enabling the reconstruction of the process parameters that were applied during the production of mineral binders: The burning temperatures of high-fired medieval gypsum mortars can be determined by measuring the Raman band widths of unhydrated anhydrite grains that are found either in clusters of firing products or spread within the gypsum matrix. Indicators for the slow cooling of 19th-century cements in the batchwise operated shaft kilns, such as morphology, polymorphism, and (sub-)stoichiometric compositions of solid solutions, can be studied and imaged due to the high sensitivity of Raman spectroscopy to chemical and physical material properties.

Author Contributions: Conceptualisation, T.S. and P.D.; methodology, T.S. and P.D.; software, T.S.; formal analysis, T.S.; investigation, T.S. and P.D.; resources, P.D.; data curation, T.S.; writing—original draft preparation, T.S. and P.D.; writing—review and editing, T.S. and P.D.; visualisation, T.S.; project administration, P.D.; funding acquisition, P.D. Funding: We gratefully acknowledge a scholarship and funding of the equipment by DFG GSC 1013 SALSA. Acknowledgments: The authors gratefully acknowledge Tobias Hotz (TH-Conservations, Weinfelden, Switzerland) for providing samples from the balustrade of the pavilion in Feldbach and the “Zurileu” in Zurich. We likewise thank Doris Warger and Florian Nick (Doris Warger AG, Frauenfeld, Switzerland) for the sample from stucco elements in the Holy Cross Chapel of the convent of St. John in Müstair. Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. Heritage 2019, 2 1681

References

1. Kaszowska, Z.; Malek, K.; Staniszewska-Slezak, E.; Niedzielska, K. Raman scattering or fluorescence emission? Raman spectroscopy study on lime-based building and conservation materials. Spectrochim. Acta A 2016, 169, 7–15. [CrossRef][PubMed] 2. Schmid, T.; Dariz, P. Determination and imaging of binder remnants and aggregates in historic artificial stone by Raman microscopy. J. Raman Spectrosc. 2013, 44, 882–891. [CrossRef] 3. Schmid, T.; Dariz, P. Chemical imaging of historical mortars by Raman microscopy. Constr. Build. Mater. 2016, 114, 506–516. [CrossRef] 4. Jallad, K.; Santhanam, M.; Cohen, M.; Ben-Amotz, D. Chemical mapping of thaumasite in sulfate-attacked cement mortar using near-infrared Raman imaging microscopy. Cem. Concr. Res. 2001, 31, 953–958. [CrossRef] 5. Sahu, S.; Exline, D.; Nelson, M. Identification of thaumasite in concrete by Raman chemical imaging. Cem. Concrete Comp. 2002, 24, 347–350. [CrossRef] 6. García-Florentino, C.; Maguregui, M.; Morillas, H.; Balziskueta, U.; Azcarate, A.; Arana, G.; Madariaga, J. Portable and Raman imaging usefulness to detect decaying on mortars from Punta Begoña Galleries (Getxo, North of Spain). J. Raman Spectrosc. 2016, 47, 1458–1466. [CrossRef] 7. Irazola, M.; Olivares, M.; Castro, K.; Maguregui, M.; Martínez-Arkarazo, I.; Madariaga, J. In situ Raman spectroscopy analysis combined with Raman and SEM-EDS imaging to assess the conservation state of 16th century wall paintings. J. Raman Spectrosc. 2012, 43, 1676–1684. [CrossRef] 8. Maguregui, M.; Knuutinen, U.; Trebolazabala, J.; Morillas, H.; Castro, K.; Martínez-Arkarazo, I.; Madariaga, J. Use of in situ and confocal Raman spectroscopy to study the nature and distribution of carotenoids in brown patinas from a deteriorated wall painting in Marcus Lucretius House (Pompeii). Anal. Bioanal. Chem. 2012, 402, 1529–1539. [CrossRef] 9. Veneranda, M.; Irazola, M.; Pitarch, A.; Olivares, M.; Iturregui, A.; Castro, K.; Madariaga, J. In-situ and laboratory Raman analysis in the field of cultural heritage: The case of a mural painting. J. Raman Spectrosc. 2014, 45, 228–237. [CrossRef] 10. Raman, C.V.; Krishnan, K.S. A new type of secondary radiation. Nature 1928, 121, 501–502. [CrossRef] 11. Opilik, L.; Schmid, T.; Zenobi, R. Modern Raman imaging: Vibrational spectroscopy on the micrometer and nanometer scales. Annu. Rev. Anal. Chem. 2013, 6, 379–398. [CrossRef][PubMed] 12. Rayleigh, L. Investigations in Optics, with Special Reference to the Spectroscope. Philos. Mag. 1879, 8, 261–274. [CrossRef] 13. Schmid, T.; Schäfer, N.; Levcenko, S.; Rissom, T.; Abou-Ras, D. Orientation-distribution mapping of polycrystalline materials by Raman microspectroscopy. Sci. Rep. 2015, 5, 18410. [CrossRef][PubMed] 14. Schmid, T.; Schäfer, N.; Abou-Ras, D. Raman microspectroscopy provides access to compositional and microstructural details of polycrystalline materials. Spectrosc. Eur. 2016, 28, 16–20. 15. Campbell, D. Microscopical Examination and Interpretation of Portland Cement and Clinker, 2nd ed.; Portland Cement Association: Skokie, IL, USA, 1999. 16. Gille, F.; Dreizler, I.; Grade, K.; Krämer, H.; Woermann, E. Mikroskopie des Zementklinkers: Bilderatlas; Beton-Verlag: Düsseldorf, Germany, 1965. 17. Hofmänner, F. Portlandzement-Klinker. Kleine Gefügekunde; Holderbank: Heerbrugg, Switzerland, 1973. 18. Stutzman, P. Microscopy of clinker and hydraulic cements. Rev. Mineral. Geochem. 2012, 74, 101–146. [CrossRef] 19. Du Toit, P.; Caragacean, L. Benefits of microscopy for raw material preparation and clinker processing. In Proceedings of the 15th Euroseminar on Microscopy Applied to Building Materials, Delft, The Netherlands, 17–19 June 2015; Çopuroglu, O., Ed.; Delft University of Technology: Delft, The Netherlands, 2015; pp. 25–32. 20. French, W. Concrete petrography: A review. Q. J. Eng. Geol. 1991, 24, 17–48. [CrossRef] 21. Poole, A.; Sims, I. Concrete Petrography: A Handbook of Investigative Techniques, 2nd ed.; Taylor & Francis Group: London, UK, 2016. 22. Blezard, R. Technical aspects of Victorian cement. Chem. Ind. 1981, 17, 630–636. Heritage 2019, 2 1682

23. Pintér, F.; Gosselin, C. Material characteristics of prefabricated concrete elements from a late 19th century church in lower Austria. In Proceedings of the 15th Euroseminar on Microscopy Applied to Building Materials, Delft, The Netherlands, 17–19 June 2015; Çopuroglu, O., Ed.; Delft University of Technology: Delft, The Netherlands, 2015; pp. 131–138. 24. Pintér, F.; Gosselin, C. The origin, composition and early age hydration mechanisms of Austrian natural Portland cement. Cem. Concr. Res. 2018, 110, 1–12. [CrossRef] 25. Ando, Y.; Hirono, S.; Sawaki, D.; Katayama, T. Microscopy to evaluate the properties of cement and alterations in historic mortar/concrete of old Nobiru Port project, Northeast Japan. In Proceedings of the 36th Conference on Cement Microscopy, Milan, Italy, 13–17 April 2014; International Cement Microscopy Association: Montgomery, AL, USA, 2014; pp. 212–233. 26. Katayama, T.; Ando, Y.; Hirono, S.; Sawaki, D. Relicts of unhydrated cement clinker in a historic concrete from the 19th century—Microscopy with EDS analysis of old training dyke at Yokohama Port, Japan. In Proceedings of the 36th Conference on Cement Microscopy, Milan, Italy, 13–17 April 2014; International Cement Microscopy Association: Montgomery, AL, USA, 2014; pp. 432–458. 27. Katayama, T.; Sakai, K. Petrography of 100-year-old concrete from Otaru Port, Japan. In Proceedings of the 2nd International Conference on Concrete under Severe Conditions, Tromsø, Norway, 21–24 June 1998; Gjørv, O., Sakai, K., Banthia, N., Eds.; E & FN Spon: London, UK, 1998; pp. 250–261. 28. Francis, A. The Cement Industry 1796–1914: A History; David & Charles: Newton Abbot, UK, 1997. 29. Black, L. Raman spectroscopy of cementitious materials. Spectrosc. Prop. Inorg. Organomet. Compd. 2009, 40, 72–127. 30. Black, L.; Brooker, A. SEM-SCA: Combined SEM-Raman spectrometer for analysis of OPC clinker. Adv. Appl. Ceram. 2007, 6, 327–334. [CrossRef] 31. Potgieter-Vermaak, S.; Potgieter, J.H.; Van Grieken, R. The application of Raman spectroscopy to investigate and characterize cement, Part I: A review. Cem. Concr. Res. 2006, 36, 656–670. [CrossRef] 32. Conjeaud, M.; Boyer, H. Some possibilities of Raman microprobe in cement chemistry. Cem. Concr. Res. 1980, 10, 61–70. [CrossRef] 33. Handke, M. Vibrational spectra, force constants, and Si-O bond character in calcium silicate crystal structure. Appl. Spectrosc. 1986, 40, 871–877. [CrossRef] 34. Fukuda, K.; Takeda, A.; Yamaguchi, A.; Hashimoto, S. Characterization of liquid exsolved by remelting reaction of belite. J. Am. Ceram. Soc. 2001, 84, 1155–1160. [CrossRef]

35. Fujimori, H.; Komatsu, H.; Ioku, K.; Goto, S.; Watanabe, T. Vibrational spectra of Ca3SiO5: Ultraviolet Laser Raman Spectroscopy at high temperatures. J. Am. Ceram. Soc. 2005, 88, 1995–1998. [CrossRef] 36. Ibáñez, J.; Artús, L.; Cuscó, R.; López, Á.; Menéndez, E.; Andrade, M. Hydration and carbonation of

monoclinic C2S and C3S studied by Raman spectroscopy. J. Raman Spectrosc. 2007, 38, 61–67. [CrossRef] 37. Dariz, P.; Schmid, T. Ferruginous phases in 19th century lime and cement mortars: A Raman microspectroscopic study. Mater. Charact. 2017, 129, 9–17. [CrossRef] 38. Dariz, P.; Neubauer, J.; Götz-Neunhoeffer, F.; Schmid, T. Calcium aluminates in clinker remnants as marker phases for various types of 19th-century cement studied by Raman microspectroscopy. Eur. J. Mineral. 2016, 28, 907–914. [CrossRef] 39. Torréns-Martín, D.; Fernández-Carrasco, L.; Martínez-Ramírez, S.; Ibáñez, J.; Artús, L.; Matschei, T. Raman spectroscopy of anhydrous and hydrated calcium aluminates and sulfoaluminates. J. Am. Ceram. Soc. 2013, 96, 3589–3595. [CrossRef] 40. Torréns-Martín, D.; Fernández-Carrasco, L.; Martínez-Ramírez, S. Hydration of calcium aluminates and calcium sulfoaluminate studied by Raman spectroscopy. Cem. Concrete Res. 2013, 47, 43–50. [CrossRef] 41. Higl, J.; Köhler, M.; Lindén, M. Confocal Raman microscopy as a non-destructive tool to study microstructure of hydrating cementitious materials. Cem. Concrete Res. 2016, 88, 136–143. [CrossRef] 42. Lafuente, B.; Downs, R.T.; Yang, H.; Stone, N. The power of databases: The RRUFF project. In Highlights in Mineralogical Crystallography; Armbruster, T., Danisi, R.M., Eds.; W. De Gruyter: Berlin, Germany, 2015; pp. 1–30. 43. Hewlett, P. Lea’s Chemistry of Cement and Concrete, 4th ed.; Butterworth Heinemann: Oxford, UK, 2004. 44. Casanova, E.; Pelé-Meziani, C.; Guilminot, É.; Mevellec, J.-Y.; Riquier-Bouclet, C.; Vinçotte, A.; Lemoine, G. The use of vibrational spectroscopy techniques as a tool for the discrimination and identification of the natural and synthetic organic compounds used in conservation. Anal. Methods 2016, 8, 8514–8527. [CrossRef] Heritage 2019, 2 1683

45. Tennent, N.; Caen, J.; Courtney, P.; Lozano Diz, E. In-situ Raman spectroscopic characterisation of polymers used in past conservation treatments. e-Preserv. Sci. 2009, 6, 107–111. 46. Schmid, T.; Jungnickel, R.; Neuhaus, B.; Riedel, J.; Kneipp, J.; Lüter, C. Raman spectroscopy as a tool for the collection management of microscope slides. Zool. Anz. 2016, 265, 178–190. [CrossRef] 47. Insley, H. Structural characteristics of some constituents of Portland cement clinker. J. Res. Natl. Bur. Stand. 1936, 17, 353–361. [CrossRef]

48. Fukuda, K.; Maki, I. Orientation of β-Ca2SiO4 solid solution lamellae formed in the host α-phase. Cem. Concrete Res. 1989, 19, 913–918. [CrossRef] 49. Fukuda, K.; Maki, I.; Ito, S. Transformation-induced microtextures in belites. In Proceedings of the 10th International Congress on the Chemistry of Cement, Gothenburg, Sweden, 2–6 June 1997; Justnes, H., Ed.; Amarkai AB and Congrex Göteborg AB: Gothenburg, Sweden, 1997; pp. 1–52. 50. Remy, C.; Reynard, B.; Madon, M. Raman spectroscopic investigations of dicalcium silicate: Polymorphs and high-temperature phase transformations. J. Am. Ceram. Soc. 1997, 80, 413–423. [CrossRef] 51. Bhagavantam, S. Effect of crystal orientation on the Raman spectrum of calcite. Proc. Ind. Acad. Soc. A 1940, 11, 62–71. [CrossRef] 52. Hopkins, J.B.; Farrow, L.A. Raman microprobe determination of local crystal orientation. J. Appl. Phys. 1986, 59, 1103–1110. [CrossRef] 53. Hayward, I.P.; Baldwin, K.J.; Hunter, D.M.; Batchelder, D.N.; Pitt, G.D. Direct imaging and confocal mapping of diamond films using luminescence and Raman scattering. Diam. Rel. Mater. 1995, 4, 617–621. [CrossRef] 54. Loudon, R. The Raman effect in crystals. Adv. Phys. 1964, 13, 423–482. [CrossRef]

55. Fukuda, K.; Maki, I.; Ito, S. Phase stability study on the remelting reaction in Ca2SiO4 solid solutions. J. Am. Ceram. Soc. 1995, 78, 3387–3389. [CrossRef] 56. Fukuda, K.; Wakamatsu, N.; Ito, S.; Yoshida, H. Acceleration of early hydration in belite-rich cement by remelting reaction. J. Ceram. Soc. Jpn. 1999, 107, 901–906. [CrossRef] 57. Schlütter, F.; Kaiser, W.; Juling, H. High fired gypsum mortar for screeds, terrazzo and masonry repair on historic monuments. Production, properties and sample applications. In Proceedings of the 2nd Historic Mortars Conference HMC2010 and RILEM TC 203-RHM Final Workshop, Prague, Czech Republic, 22–24 September 2010; Válek, J., Groot, C., Hughes, J., Eds.; RILEM Publications SARL: Bagneux, France, 2010; pp. 1169–1180. 58. Schlütter, F. Mittelalterlicher Hochbrandgips. In 800 Jahre Kunststein—Vom Imitat Zum Kunstgut; für Denkmalpflege, B.L., Archäologisches Landesmuseum, A., Eds.; Wernersche Verlagsgesellschaft mbH: Worms, Germany, 2012; pp. 27–39. 59. Schlütter, F.; Jakubek, M.; Juling, H. Charakterisierung und Eigenschaften historischer Gipsmörtel aus unterschiedlichen Epochen und Anwendungsgebieten. In Gips als Baugrund, Mörtel und Dekorationsmaterial; Institut für Steinkonservierung e.V., IFS: Mainz, Germany, 2012; pp. 49–59. 60. Lenz, R.; Sobott, R. Beobachtungen zu Gefügen historischer Gipsmörtel. In Gipsmörtel im Historischen Mauerwerk und an Fassaden; Auras, M., Zier, H.-W., Eds.; WTA Publications: Munich, Germany, 2008; pp. 23–34. 61. Dariz, P.; Jakob, C.; Ectors, D.; Neubauer, J.; Schmid, T. Measuring the burning temperatures of anhydrite micrograins in a high-fired medieval gypsum mortar. ChemistrySelect 2017, 2, 9153–9156. [CrossRef] 62. Dariz, P.; Schmid, T. Phase composition and burning history of medieval high-fired gypsum mortars studied by Raman microspectroscopy. Mater. Charact. 2019, 151, 292–301. [CrossRef] 63. Schmid, T.; Jungnickel, R.; Dariz, P. Raman band widths of anhydrite II reveal the burning history of high-fired medieval gypsum mortars. J. Raman Spectrosc. 2019, in press. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).