Deciphering the Sulfate Attack of Cementitious Materials by High

ARTICLE pubs.acs.org/ac

Deciphering the Sulfate Attack of Cementitious Materials by High-Resolution Micro-X-ray Diffraction Moritz C. Schlegel,† Urs Mueller,† Ulrich Panne,†,‡ and Franziska Emmerling*,† † ‡

BAM Federal Institute for Materials Research and Testing, Richard-Willst€atter-Strasse 11, 12489 Berlin, Germany Department of Chemistry, Humboldt-Universitaet zu Berlin, Brook-Taylor-Strasse 2, 12489 Berlin, Germany ABSTRACT: The durability of cementitious materials depends, among others, on their resistance against chemical attack during the service life of a building. Here, we present an approach to analyze changes in the phase composition due to chemical attack in the form of sulfate ingress within the microstructure. Micro-X-ray (μX-ray) diffraction using synchrotron radiation in Debye
Scherrer (transmission) geometry allowed a spatial resolution of 10 μm. Phase transformations in the wake of damaging processes were observed in a detailed high-resolution imaging study. In comparison, samples containing supplementary cementitious materials were investigated and used to reconstruct the influence of different degeneration processes in detail. Additionally, reaction fronts within the bulk were localized by micro-X-ray fluorescence analysis. The experimental setup provided the possibility for analyzing the phase assemblage of a given sample without destroying the microstructure. The specimens for phase analysis are thick sections of the primary material and can be used for further microscopic analysis of the microstructure and microchemistry, e.g., scanning electron microscopy
energydispersive X-ray spectroscopy (SEM
EDX) or Raman spectroscopy.

C

ementitious building materials are a substantial part of our infrastructure and built environment.1 After construction, these materials are often exposed to chemical attacks due to anthropogenic sources, e.g., air pollution, or natural origin, e.g., weathering of surrounding sulfide- or sulfate-bearing bedrocks.2
5 Solutions containing sulfate ions (SO42
) penetrating concrete react with components of the cement paste to secondary phases. The crystallization of some of the secondary phases is accompanied with a volume expansion, which induces damaging processes such as crack growth and concrete spalling due to the crystallization pressure.3,4 These result in a reduction of mechanical strength and a decrease of the service life of an affected structure. The crack growth process increases the penetration rate of the sulfate solution, and the chemical attack is reinforced. The crystallization of secondary phases depends upon the concentration of the penetrating solution, the porosity of the cementitious material, and the chemical composition of the cement paste. Several studies were described using a scanning electron microscope (SEM) coupled with energy-dispersive X-ray spectroscopy (EDX).5,6 These measurements typically consist of a morphological determination of different phases (SEM) followed by an analysis of the chemical composition by EDX analysis. The structural information is not directly available, which could lead to imprecise assignments. The published chemical compositions differ strongly and indicate that the electron beam penetrates more than a single phase within the sample volume. The crystal structures of different modifications of secondary phases and their thermodynamics as well as the involved r 2011 American Chemical Society

chemical reactions and microstructural changes were also the subject of previous studies.7
13 Preparation techniques such as dissolving in acid or milling were common in former studies, e.g., IR spectroscopy or powder diffraction methods.14,15 In former studies, preparation techniques such as milling were commonly applied as a first sample preparation step especially for powder diffraction. Naturally, pressure and temperature during the milling process could result in artifacts. Moreover, the samples were cut into layers of several millimeters prior to the milling processes. This does not provide sufficient spatial resolution as all information about the phase assemblage within the microstructure is completely lost. Alternative studies reduced the spatial resolution by preparing the powder samples through grinding processes and measuring the resulting dust. Grinding devices may alter the sample properties due to chemical strain, and it is difficult to reach defined profile depths. Former investigations by X-ray diffraction (XRD) were limited by both sample preparation, especially the influences of the preparation devices, and by spatial resolution. Up to now the results of former investigations describing the chemical attack of cementitious materials were never verified by spatial phase analysis of the microstructure by XRD with high resolution. The transmission geometry described in this study provides a low-invasive characterization of the degradation Received: January 21, 2011 Accepted: April 4, 2011 Published: April 04, 2011 3744

dx.doi.org/10.1021/ac200181g | Anal. Chem. 2011, 83, 3744–3749

Analytical Chemistry

ARTICLE

Figure 1. Schematic representation of the sample preparation and experimental method.

processes within the preserved microstructure of the material with a resolution of 10 μm.

’ EXPERIMENTAL SECTION The sulfate attack was simulated under laboratory conditions by embedding four samples consisting of cement paste in an aqueous solution with a high content of Na2SO4 (30 g L
1) over time periods of 3 and 6 months. Typical samples prepared according to the standard EN 196-1 had a dimension of 40  40  160 mm3 and consist of ordinary portland cement paste (OPCP) and OPCP with 30 wt % of limestone (LL), slag (HS), or fly ashes (FS). After exposing the samples for different time spans to the sulfate attack, the samples were cut and polished into thick sections with a dimension of 20  40  0.2 mm3. During the preparation process, the samples were cooled with petroleum ether to counteract contamination and dissolution of watersoluble phases. This procedure ensures the coherence of the microstructure during the preparation. The penetration depths of the solutions and the secondary phases were localized by elemental mapping of sulfur using micro-X-ray fluorescence (μXRF) (Eagle III, EDAX, Wiesbaden, Germany, operating at 40 kV, 120 mA) with a spatial resolution of 40 μm. Finally, all profiles were studied by micro-X-ray diffraction (μXRD) (see Figure 1). The in situ XRD experiments were performed at the synchrotron microfocus beamline μSpot (BESSY II of the Helmholtz Centre Berlin for Materials and Energy).16 Profiles were measured across the reaction fronts, including the intact cement paste. To record the profiles, the samples were moved 10 mm parallel to the reaction front for 60 s in order to improve sampling statistics. The resulting diffraction patterns were represented in a top view to optimize the phase identification and provide an overview of both reflections intensities and positions. The ocean data viewer software was used to provide the respective images.17 The focusing system of the beamline provides a beam diameter of 10 μm at a photon flux of 1  109 s
1 at a ring current of 200 mA, providing a divergence of less than 1 mrad (horizontally and vertically). The experiments were carried out with a wavelength of 1.0657 Å using a double-crystal monochromator [Si(111)]. The diffracted intensities were collected 200 mm behind the sample position with a two-dimensional MarMosaic CCD X-ray detector with 3072  3072 pixels and a point spread function width of about 100 μm.16 The obtained scattering images were processed and converted into diagrams of scattered intensities versus scattering vector q (q = 4π/λsin θ) employing

an algorithm from the FIT2D software.18 For the graphical representations, q-values were transformed to the diffraction angle 2θ (Cu) to provide a direct comparison to results obtained by XRD with Cu radiation. Copper anodes are one of the most commonly used X-ray tube materials for laboratory XRD studies. A typical phase identification is shown in Figure 2. The phase identification was carried out with the search/match function of the EVA software.19 Additionally, profile scans using a μRaman spectrometer LabRam (Horiba Jobin-Yvon, Bensheim, Germany) were measured to validate the μXRD method. The system was operated at an excitation wavelength of 785 nm using a notch filter and a liquid nitrogen cooled CCD detector (256  1024 pixels), at a laser power of 1.8 mW (corresponding to an intensity of 3  104 W/cm2 when using a 50 microscope objective).

’ THEORETICAL BACKGROUND Dry cementitious materials can be described as a complex multiphase system consisting of different crystalline calcium silicates and aluminates (and calcium sulfates for controlling the hardening process). When mixed with water, these phases react from the start to structurally more complex hydration products, mostly calcium silicate hydrates (C
S
H) and portlandite (Ca(OH)2), but also calcium aluminate (ferrite) sulfate hydrate phases. The latter is present in the form of ettringite during the hardening process and is then transformed into calcium aluminate (ferrite) monosulfate hydrate (monosulfate). These sulfate-containing phases are part of two phase groups, the AFm ([Ca2(Al,Fe)(OH)6] 3 x 3 yH2O, x = SO42
, CO32
, Cl
, OH
= 1) and the AFt ([Ca3Al(OH)6 3 12H2O]2 3 x 3 yH2O, x = SO42
, CO32
= 3) phases, which all can be formed in the cement paste system. The AFm phases form layered structures consisting of edgelinked Ca or Ca and Al polyhedra. Different types of anions compensate the residual charge by occupying atomic positions between the layers and thus stabilizing the crystal structure. Depending on the ion (SO42
, CO32
, Cl
) occupying the interlayer positions, different types of AFm phases (monosulfate, monocarbonate, or Friedel’s salt) are formed. Kuzel’s salt and kuzelite crystallize when atomic positions occupied both by SO42
and Cl
or SO42
and OH
.20 The crystal structure of the AFt phases forms a network of Ca and O, which is arranged as tunnel structures. Between these tunnel structures, SO42
and CO32
compensate the residual charge and stabilize the crystal structure. The resulting secondary phases are ettringite or 3745

dx.doi.org/10.1021/ac200181g |Anal. Chem. 2011, 83, 3744–3749

Analytical Chemistry

ARTICLE

Figure 2. Example for phase identification based on the positions of the reflection. The lattice planes (hkl) are displayed in red for gypsum, blue for calcite, and black for ettringite.

tricarbonate. Under special conditions, both SO42
and CO32
occupy the intertunnel positions and Al3þ is substituted by Si4þ and results in the crystallization of the secondary phase thaumasite (Ca6Si2(SO4)2(CO3)2(OH)12 3 26H2O).21,22 Monosulfate is part of the cement paste and is formed during hydration by consuming ettringite. Ions (e.g., SO42
, Cl
, CO32
) which ingress from the outside can cause the formation of secondary AFm or AFt phases, most prominently secondary ettringite, which may cause considerable damage in the hardened cement paste. XRD methods are essential for the structural characterization of these secondary phases in detail. As indicated in Figure 3, it is possible to distinguish between the different ions occupying the interlayer and tunnel structures based on the diffraction pattern. The reflection positions change during the substitution of the residual charge-compensating ions. Furthermore, the reflection intensities and position of secondary phases representing solid solutions, e.g., Kuzel’s salt and thaumasite, differ clearly from their end members. This is an advantage compared to other spectroscopic methods, i.e., Raman spectroscopy, which cannot distinguish between a free ion and an ion stabilizing the crystal structure.

’ RESULTS AND DISCUSSION The samples were exposed to sulfate attack over time periods of 3 and 6 months. The corresponding results of sulfur mapping by μXRF are shown in Figure 4. The sulfur concentration is displayed from the surface to a profile depth of 10 mm. Each of the samples exhibits four areas of different sulfur concentrations: The sample surface exhibits elevated sulfur concentrations, followed by an area with the highest observed concentrations, a moderately concentrated area, and an area with the sulfur content of the bulk cement paste. All sulfur gradients were observed perpendicular to the sample surface. The sample surface itself interacts with the surrounding sulfate solution. This limits the crystallization processes of secondary phases and results in a first interface between the area containing low and the area containing high amounts of secondary sulfur phases. The second interface seems to be a reaction front between secondary sulfate phases and the intact bulk with a sulfur-saturated pore

space. When the different samples are compared, the profile depth of the second interface indicates the extent of crystallization of the secondary phases. The third interface is an indicator for the penetration depth of the sulfate solution. The reaction times of the crystallization processes of the secondary phases are influenced by the chemical composition and reactivity of each supplementary cementitious material. Additionally, the penetration depth of the solution is controlled by the water/cement ratio, the grain size distribution of the supplementary cementitious material, and the hydration rate. The phase identification by the high-resolution μXRD helps to elucidate the change of the phase composition due to the sulfate attack. When the phase composition within the microstructure is investigated, the detection of every (hkl) of the crystalline phases similar to conventional powder diffraction is not achievable. The particle size and particle size distribution limits the number of detectable (hkl). However, powder diffraction methods are limited by the powder sample properties as well, e.g., preferred orientation of the crystallites and penetration depth of the X-rays. Finally, it is possible to determine the phase composition completely. For each of the eight different samples, a change in phase composition could clearly be observed at defined profile depths (see Figure 5). Pure OPCP reacts gradually and is accompanied by the crystallization of the secondary phases, ettringite and gypsum, due to the high concentration of dissolved SO42
in the pore solution. From the sample surface to a depth of 420 μm, ettringite and portlandite were identified after an exposure time of 3 months. Below this profile depth only portlandite was found representing the intact bulk. After 6 months, ettringite and gypsum could be identified at the sample surface until a profile depth of 240 μm and no more portlandite was left. From 240 to 1180 μm, gypsum was the only existing crystalline secondary phase. From 1180 μm to the end of the profile (3 mm), the formation of portlandite indicates the intact microstructure. The sample with a high amount of limestone shows an increase of the penetration depths of the sulfate solution. During the first 3 months, calcite was identified and the crystallization of the secondary phases ettringite and monocarbonate could be observed from the sample surface to a profile depth of 390 μm. Calcite (CaCO3) was found after both time periods of the sulfate 3746

dx.doi.org/10.1021/ac200181g |Anal. Chem. 2011, 83, 3744–3749

Analytical Chemistry

ARTICLE

Figure 3. Unit cell of Friedel’s salt parallel c representing the crystal structure of the AFm phases (refs 23
27). Aluminum and calcium polyhedra are displayed in blue and red, respectively. The oxygen atoms are represented in red, and the chlorine atoms are in gray color. The Cl
can be replaced by SO42
, CO32
, or OH
.

Figure 4. Color-coded sulfur concentration within OPCP, OPCP þ 30 wt % limestone (LL), OPCP þ 30 wt % slag (HS), and OPCP þ 30 wt % fly ashes (FS) after exposure in sodium sulfate solution of 3 (left column) and 6 months (right column). The high-concentrated areas (sp) at the sample surface are representing the secondary phases, followed by lower concentrated areas (ss) representing the penetration depths of the sulfate solution and areas representing the intact bulk (ib).

attack and in every profile depth representing the limestone filler within the sample. It is the driving force for the crystallization of monocarbonate, because the high amount of limestone signifies a strong CO32
source for the crystallization process. It is most likely that, initially, monosulfate recrystallizes to monocarbonate by substituting SO42
though CO32
. Investigations of the sulfate attack carried out at early stages (< 3 months) would offer a more detailed understanding of this mechanism. In addition, gypsum was found between 390 and 810 μm, increasing the diversity of the phase composition. Below 810 μm, only portlandite and calcite were identified representing the intact bulk. After 6 months, calcite and only a small amount of ettringite is left near the sample surface down to a profile depth of 450 μm. Directly at the sample surface, ettringite was identified by reflections with high-indexed (hkl) values which correspond to small lattice plane distances. A possible explanation is that the initially crystallized secondary ettringite partially recrystallizes to gypsum and results in a strongly disordered ettringite structure. Underneath the surface, calcite and a high amount of gypsum was identified down to 1860 μm. Below this profile depth the intact bulk is recognized by the presence of portlandite and calcite. The cement with a high amount of slag displays a faster crystallization time of ettringite and gypsum as the pure OPCP

after 3 and 6 months. However, additional gypsum was found after 3 months of the sulfate attack from the sample surface down to a profile depth of 690 μm. Underneath this profile depth, portlandite and additional amount of ettringite indicates the accelerated crystallization of ettringite. After 6 months, only ettringite is crystallized and identified with a large number of (hkl) at the sample surface area down to a depth of 480 μm. Below this depth, portlandite was found in addition to ettringite, where the ettringite amount did not change significantly compared to the amount of ettringite after 3 months of the sulfate attack. By using fly ash as a supplementary cementitious material, the crystallization process of gypsum was decelerated. Probably, the grain size and the grain size distribution provides a more densely packed material, decreasing the open pore space and reducing the penetration rate of SO42
. After 3 months, gypsum is only identified from the sample surface down to a profile depth of 270 μm, but ettringite was found within the whole profile in addition to portlandite and a solid solution of AFm. After 6 months, only ettringite was left within the surface area down to a profile depth of 270 μm. Below 270 μm ettringite and gypsum were identified down to a profile depth of 720 μm. Below this depth, ettringite was observed as well as portlandite and small 3747

dx.doi.org/10.1021/ac200181g |Anal. Chem. 2011, 83, 3744–3749

Analytical Chemistry

ARTICLE

Figure 5. Phase identification of OPCP (A1, A2), OPCP þ 30 wt % limestone (B1, B2), OPCP þ 30 wt % slag (C1, C2), and OPCP þ 30 wt % fly ashes (D1, D2) after a chemical attack over a time period of 3 months (left column) and 6 months (right column). The phase names are sorted according to their content.

amounts of gypsum. The cement paste containing fly ash revealed the highest resistivity against the chemical attack; the crystallization of gypsum dominated in favor of ettringite. In summary, it can be stated that supplementary cementitious materials have strong influences on the reaction time of the secondary phases and the penetration of the sulfate solution. After 3 months, the crystallization of the ettringite is coupled with a strong decrease of the amount of portlandite. After 6 months, no portlandite is left within the areas where gypsum crystallized. Every sample with supplementary cementitious materials displayed a small layer of ettringite at the sample surface. Ettringite at the surface was identified by a large number of (hkl) with high-indexed values for h, k, and l. One possible explanation is that the ettringite is more stable at the sample surface where the sulfur concentration is lower compared to subsurface layers. The crystallization of gypsum could be limited due to the interaction at the interface between the open pore space and the surrounding sulfur solution.

The described experimental setup is not limited to spatial investigations of cementitious materials but can also be adapted to similar materials where crystallization processes within a complex matrix are studied.

’ CONCLUSION The elemental mapping by μXRF described here allowed the localization of secondary reaction products of concrete deterioration within the crystalline material as well as observing the penetration depths of the sulfate solution and areas representing the intact bulk. The advantages as compared to previous studies of the spatial analysis by μXRD are (i) the high resolution of 10 μm, which provides a detailed view of the phase assemblage, (ii) the reconstruction of the mechanisms leading to a sulfate attack in cement paste, and (iii) reuse of the sample specimen for further nondestructive microstructural analysis, e.g., by SEM or Raman spectroscopy. The reaction times of the crystallization 3748

dx.doi.org/10.1021/ac200181g |Anal. Chem. 2011, 83, 3744–3749

Analytical Chemistry processes of secondary phases could be determined and compared for different cements and their reaction with the penetrating solution. The influences of common supplementary cementitious materials were individually characterized. The degeneration processes are connected to the material properties, especially by the chemical composition and reactivity of each supplemental cementitious material. This low-invasive investigations offer new possibilities to in situ analysis of different chemical attacks in several types of cementitious materials, i.e., without destroying the microstructure of the cementitious material. In conclusion, the presented study shows the necessity of high-resolution structure analysis for the characterization of chemical attacks in the micrometer scale.28 On the basis of the presented data, the sulfate attack can be described as the formation of secondary gypsum which replaces the initially formed secondary ettringite. This indicates a recrystallization process which is only detectable by systematic investigations with highresolution diffraction methods. This method allows investigating the same samples with further experimental setups, and the results are directly comparable to results of imaging techniques, such as polarizing microscopy and SEM
EDX, or spectroscopic methods, such as Raman spectroscopy.

ARTICLE

(17) Schlitzer, R. Alfred-Wegener-Institut, Bremen, Germany, Ocean Data View, http://odv.awi.de, 2009. (18) Hammersley, A. P.; Brown, K.; Burmeister, W.; Claustre, L.; Gonzalez, A.; McSweeney, S.; Mitchell, E.; Moy, J. P.; Svensson, S. O.; Thompson, A. W. J. Synchrotron Radiat. 1997, 4, 67–77. (19) The International Centre for Diffraction Data (ICDD), 2008. (20) Mesbah, A.; Rapin, J. P.; Franc-ois, M.; Cau-dit-Coumes, C.; Frizon, F.; Leroux, F.; Renaudin, G. J. Am. Ceram. Soc. 2011, 94, 261–268. DOI: 10.1111/j.1551-2916.2010.04050.x. (21) Jallad, K. N.; Santhanam, M.; Cohen, M. D.; Ben-Amotz, D. Cem. Concr. Res. 2001, 31, 953–958. (22) Bensted, J. Cem. Concr. Compos. 2003, 25, 873–877. (23) Renaudin, G.; Kubel, F.; Rivera, J. P.; Francois, M. Cem. Concr. Res. 1999, 29, 1937–1942. (24) Allmann, R. Neues Jahrb. Mineral., Monatsh. 1968, 140–144. (25) Francois, M. R. G.; Evrard, O. Acta Crystallogr. 1998, C54, 1214–1217. (26) Renaudin, G.; Kubel, F.; Rivera, J. P.; Francois, M. Cem. Concr. Res. 1999, 29, 1937–1942. (27) Terzis, A.; Fillipakis, S.; Kuzel, H.-J.; Burzlaff, H. Z. Kristallogr. 1987, 181, 29–34. (28) Svarcova, S.; Bezdicka, P.; Hradil, D.; Hradilova, J.; Zizak, I. Anal. Bioanal. Chem. 2011, 399, 331–336.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]. Phone: þ49 (0) 30 81041133. Fax: þ49 (0) 30 8104-1137.

’ ACKNOWLEDGMENT M. C. Schlegel thanks the BAM Federal Institute for Materials Research and Testing, Berlin, Germany, for a Grant within its Ph. D. program. We thank S. Rolf and A. Gardei for technical support. ’ REFERENCES (1) Idorn, G. M. Cem. Concr. Res. 2005, 35, 3–10. (2) Bensted, J. Cem. Concr. Compos. 1999, 21, 117–121. (3) Crammond, N. J. Cem. Concr. Compos. 2003, 25, 809–818. (4) Loudon, N. Cem. Concr. Compos. 2003, 25, 1051–1058. (5) Jupe, A. C.; Stock, S. R.; Lee, P. L.; Naik, N. N.; Kurtis, K. E.; Wilkinson, A. P. J. Appl. Crystallogr. 2004, 37, 967–976. (6) Gollop, R. S.; Taylor, H. F. W. Cem. Concr. Res. 1992, 22, 1027–1038. (7) Crammond, N. Cem. Concr. Compos. 2002, 24, 393–402. (8) Taylor, H. F. W. Cement Chemistry, 2nd ed.; CPI Bath: London, 1997. (9) Christensen, A. N.; Jensen, T. R.; Hanson, J. C. J. Solid State Chem. 2004, 177, 1944–1951. (10) Matschei, T.; Lothenbach, B.; Glasser, F. P. Cem. Concr. Res. 2007, 37, 1379–1410. (11) Matschei, T.; Lothenbach, B.; Glasser, F. P. Cem. Concr. Res. 2007, 37, 118–130. (12) Castellote, M.; Andrade, C.; Turrillas, X.; Campo, J.; Cuello, G. J. Cem. Concr. Res. 2008, 38, 1365–1373. (13) Skibsted, J.; Hall, C. Cem. Concr. Res. 2008, 38, 205–225. (14) de la Torre, A. G.; Cabeza, A.; Calvente, A.; Bruque, S.; Aranda, M. A. G. Anal. Chem. 2001, 73, 151–156. (15) Christensen, A. N.; Jensen, T. R.; Scarlett, N. V. Y.; Madsen, I. C. J. Am. Chem. Soc. 2004, 87, 1488–1493. (16) Paris, O.; Li, C. H.; Siegel, S.; Weseloh, G.; Emmerling, F.; Riesemeier, H.; Erko, A.; Fratzl, P. J. Appl. Crystallogr. 2007, 40, S466–S470. 3749

dx.doi.org/10.1021/ac200181g |Anal. Chem. 2011, 83, 3744–3749