Elsevier

Cement and Concrete Research

Volume 103, January 2018, Pages 49-65
Cement and Concrete Research

Automated coupling of NanoIndentation and Quantitative Energy-Dispersive Spectroscopy (NI-QEDS): A comprehensive method to disclose the micro-chemo-mechanical properties of cement pastes

https://doi.org/10.1016/j.cemconres.2017.08.016Get rights and content

Abstract

Engineering cement-based composites requires a comprehensive understanding of the microstructure features governing macroscopic properties. This work aims to foster the latest chemo-mechanical technique to disclose the micro-mechanical properties of intimately intermixed phases. Considering a typical Portland cement paste as a study case, we compared the so-far developed nanoindentation analysis methods with a new automatic method coupling Nano-Indentation and Quantitative Energy-Dispersive Spectroscopy (NI-QEDS). Besides evincing advantages and disadvantages of previous methods, the NI-QEDS analyses enabled distinguishing chemical phases having strongly overlapping mechanical properties. Results suggested the presence at the sub-micrometre scale of extra calcium, aluminum and sulfur into (or in the vicinity of) outer C–S–H. The quantitative chemistry allowed identifying the rare occurrence of nanoindentation volumes located on “pure” phases (e.g., inner C–S–H). Finally, NI-QEDS provided new knowledge on cement pastes' micro-chemo-visco-mechanical properties and showed to be a powerful tool for investigating even more heterogeneous systems.

Introduction

Engineering modern concrete made of ordinary Portland cement (OPC) requires a better knowledge of its complex microstructure properties, which – due to its high heterogeneity – still holds mysteries. Moreover, the emergence of new eco-friendly alternatives to OPC-only systems urges the development of methods for investigating their highly heterogeneous microstructure features which govern the macroscopic properties.

Typically, an OPC paste is composed of a porous matrix of hydrated phases, which are rather intermixed at the micron scale and embed anhydrous clinker grains with sizes up to tens of microns. The hydrates include calcium-silicate-hydrates (C–S–H), Portlandite (CH), aluminum-rich hydrates (e.g., aluminoferrite-mono [AFm] such as monosulfoaluminate or monocarboaluminate, and aluminoferrite-tri [AFt] such as ettringite) and possibly hydrotalcite-like phases (incorporating Mg and Al atoms) [1], [2]. As the most important binding hydrate, C–S–H has been widely investigated in the last decades in terms of morphology, sorption-desorption, chemistry, or mechanical properties [1], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12]. In a landmark work, Taplin [6] distinguished the Inner Products (IP), which lie within the original boundaries of the clinker grains, and the Outer Products (OP), which grow outside these boundaries. The IP/OP classification of C–S–H was widely accepted and investigated intensively by Richardson using transmission electron microscopy [7], [8], [13]. He found that the IP C–S–H formed by hydration of clinker grains larger than ~ 5 μm had typically a compact, fine-scale and homogeneous morphology, with pores smaller than ~ 10 nm, whereas the OP C–S–H had a fibrillar morphology. The latter was found to be composed of longer coarse fibrils growing in large pores and of a fine fibrillar morphology in constrained spaces. Based on nitrogen surface area and accessible porosity measurements, the Tennis and Jennings model [10] suggested a different classification of Low-Density (LD) and High-Density (HD) types of C–S–H. Jennings and co-workers [14], [15] further extended the classification into a colloid model considering C–S–H as a nanogranular material (or colloid) with globules of ~ 5 nm as basic building blocks. The same building blocks may organize with two preferential packing densities corresponding to LD and HD C–S–H. The structure of LD C–S–H is more porous and, thus, more deformable during internal loading (e.g., drying) or external loading [14], [15]. Eventually, the different classifications may potentially be reconciled: low density C–S–H forms as fibrillar (or foil-like) structures in the water between cement grains (outer product) during the early and middle stages of hydration; whereas high density C–S–H forms as a dense, fine-scale homogeneous structure “inside” the original boundaries of cement grains (inner product) during the late stage of hydration.

In 2004, the investigation of micro-mechanical properties of C–S–H became possible with the combined use of statistical nanoindentation and micromechanics theory, as developed by Ulm and co-workers [12], [16], [17], [18], [19]. The nanoindentation test consists of establishing contact between an indenter and a material surface, before prescribing a loading profile and measuring the penetration depth. The indentation Modulus (M) and the indentation Hardness (H) may then be derived from the measured load-penetration curve. Heterogeneous microstructures composed of distinct phases can be investigated using the grid nanoindentation method [19].

The method was based on statistical deconvolution of nanoindentation results, which consists of decomposing a mixture distribution for the two measured mechanical variables (M and H) with supposed normal distributions. The method was employed on different cementitious systems and the analysis approach evolved over the past decade: first, the frequency distributions were fitted manually [16], [19]; then, least-square fitting was employed to fit univariate probability density functions (i.e., one variable, M or H) [11], [12], [18], [20], [21], [22], [23], [24], [25], [26] and bivariate probability density functions (i.e., two variables, M and H) [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38]; eventually, better results were obtained by least square fitting of joint cumulative density functions [39], [40], [41], [42], [43], [44], [45], [46], [47], [48]. Finally, the method evolved into Gaussian mixture modeling using the Maximum Likelihood (ML) via the Expectation Maximization (EM) algorithm and the Bayesian Information Criterion (BIC) [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65].

Based on the statistical analysis of the distribution of indentation mechanical properties, the deconvoluted phases have been associated with phases of the cement paste with respect to the indentation modulus, for instance: (i) M  [0  13] GPa was attributed to regions dominated by defects or porosity; (ii) M  [13  26] GPa to LD C–S–H; (iii) M  [26  39] GPa to HD C–S–H; (iv) and E > 39 GPa to Portlandite and anhydrous clinker [12]. In recent publications [66], [67], data points with M  65 GPa and H  3 GPa were considered as hydrates and data points above those limits were associated with anhydrous clinker. In early experiments [12], white Portland cement with high siliceous clinker content (malite + mbelite  90%) was employed to limit the presence of minor hydrates, disclosing the mechanical property of two different kinds of C–S–H.

Over the past decade, such statistical nanoindentation methods have been extensively employed and similar micromechanical values were reported for LD and HD C–S–H by several authors for different types of cement-based systems in different contexts, as summarized in Table 1. The average quoted values for LD C–S–H reach MLD C–S–H  22.0 GPa and HLD C–S–H  0.7 GPa, whereas HD C–S–H reach MHD C–S–H  32.0 GPa and HHD C–S–H  1.1 GPa. A third hydrous phase having an ultra-high density (UHD) was proposed by Vandamme [48] along with estimated values of MUHD  42.8 ± 2 GPa and HUHD  1.4 ± 0.2 GPa. Although the properties of Portlandite evaluated using several methods are also in the same range with MCH  36–46 GPa and HCH  1.3–1.35 GPa [12], [16], [68], [69], [70], [71], [72], the UHD phase was separately defined because it was also found in low w/c heat treated samples (where CH was less likely to occur in high amounts) [48]. Finally, investigations of the pure phases of anhydrous clinker provided the highest mechanical properties with Mclinker  125–145 GPa and Hclinker  8.8–10.8 GPa [68], [73].

In the works of Vandamme and Ulm [40], [42], [65], the contact creep modulus (C) was defined as an additional mechanical parameter, which measures the asymptotic logarithmic creep rate of the penetration depth as a function of time during the holding phase of the load. This contact creep modulus has proved to be particularly interesting as well correlated with the basic creep of a cement paste over a much longer time [74], [75].

However, when applied to OPC with very low cement-to-water ratio (w/c) and silica fume, the statistical deconvolution based on mechanical properties showed the difficulty in distinguishing the mechanical properties of 8 phases with overlapping properties [29]. Furthermore, based on investigations using focussed-ion beam nanotomography, Trtik et al. [76] questioned the interpretation of statistical nanoindentation results arguing that there was a limited chance to perform nanoindentation on pure outer product C–S–H with an indenter–solid interaction volume of about 1 μm3 or larger. They also suggested that the peak commonly attributed to HD C–S–H could represent mixtures of C–S–H with other hydrous and anhydrous phases, which started a discussion on the application of statistical nanoindentation to cementitious materials [52], [67], [77].

A breakthrough occurred in this research endeavour with the coupling of chemo-mechanical measurements. First, Vanzo [49] statistically combined Wavelength Dispersive Spectroscopy (WDS) analyses with nanoindentation to better understand the effect of carbonation on the hydrated phases of a cement paste. However, this chemo-mechanical coupling was not carried on the same microvolumes as those indented, and therefore, the statistical analysis of the chemical and mechanical properties had to be carried out separately. Hu et al. [53], [54], [57] investigated different types of cement pastes using both nanoindentation and SEM-EDS, and they suggested that, at the scale of the interaction volume of nanoindentation, the ‘C–S–H gel’ was not a single phase but rather a composite of porous C–S–H and other phases. They also proposed a method based on micro-poromechanics to investigate these bi-phase mixtures of C–S–H with other phases. In a review article [78], those authors argue that the mechanical properties of the IP and OP C–S–H depend on the cement paste mix proportions.

Notably, Krakowiak et al. [55] developed an automatic algorithm for rapid statistical coupling of nanoindentation with qualitative Energy-Dispersive Spectroscopy (NI-EDS), by extraction of chemical intensities at the location of each nanoindentation from chemical mappings. Furthermore, they applied multivariate Gaussian mixture deconvolution algorithm on the coupled chemo-mechanical dataset to recognize the preferential chemical phases of an oil well cement paste and determine their mechanical properties. The method was also employed as part of a comprehensive validation of advanced mesoscale modeling of C-S-H [79] to isolate nanoindentation performed on the C-S-H phase. Another NI-EDS investigation of High Volume Natural Pozzolan (HVNP) cement composite revealed the importance of anhydrous pozzolan grains on the overall macroscopic properties [56]. As for low w/c ratio OPC (i.e., 0.2), Chen et al. [80] combined statistical nanoindentation and quantitative EDS by visually and manually tracing the location of each point of the nanoindentation grid. Such analysis allowed to discover that the phase proposed to be “Ultra High Density” phase was indeed an intermix of HD C–S–H and nanocrystaline CH, as also accepted by other authors [42].

Following the development trend, the coupling of NanoIndentation and Quantitative Energy-Dispersive Spectroscopy (NI-QEDS) with an automated procedure appears the next step to undertake. Hopefully, this will refine knowledge on the influence of the different anhydrous and hydrous phases on the chemo-mechanical properties of the C–S–H phase. Moreover, further understanding of the main phases occurring in low w/c systems, or in systems including OPC and/or supplementary cementitious materials (SCMs) is today paramount (e.g., CH, AFm, AFt, C–A–S–H, anhydrous SCMs, etc.).

The contribution of this work is to foster this research direction by developing the automated NI-QEDS method with an optimized analysis approach. After having reviewed the previous analysis methods, the application of NI-QEDS to a typical OPC paste allowed further distinction of the intimately-intermixed main phases of its microstructure, such as: (i) anhydrous silicates; (ii) anhydrous alumino-ferrites, (iii) C–S–H; (iv) CH; and (iv) Al-rich hydrates (e.g., AFm and AFt). Furthermore, NI-QEDS enabled the identification of the properties of the “pure” phases by selecting the points having chemistry close to the theoretical values. For the same OPC, this work compared the NI-QEDS results with the previous methods for statistical analysis of nanoindentation tests. The comparison evinced advantages and limitations of up-so far developed methods which are widely employed worldwide.

Section snippets

Materials and specimen preparation

The cement paste sample investigated in this study was prepared with Canadian General Use Ordinary Portland Cement (i.e., OPC type GU). The chemical and physical properties of this cement are presented in Table 2. The cement paste was prepared using a high-shear mixer at a water-to-cement ratio of w/c = 0.40. Samples were cast in bar molds of 25 mm × 25 mm × 285 mm and kept moist until demolding at 24 h. The samples were then cured for one year in lime water at 23 ± 2 °C. The combination of low w/c ratio

Results

In the following, we have analyzed the same nanoindentation results (or dataset) using several approaches to provide a comprehensive comparison of the existing methods which are widely employed in current open literature works (see Section 1). Note that 44 of 630 indentation points were disregarded due to the irregular shape of their load-penetration curve which was attributed to the presence of pores, cracks or defects.

Impact of the method on the identification of hydrates' properties

The results obtained using the three levels of statistical analyses and a deterministic approach showed the influence of the deconvolution method and the classification variables on the identified properties of the hydrates. In particular, Figs. 12ab and 12de compare the results obtained with the cluster analysis of level 1 and level 3, with respect to the mechanical axes of M vs. H, and C vs. H, respectively. The analysis using 3 mechanical properties (level 1, Figs. 12a and d) provided a

Conclusion

By considering a “relatively simple” system such as an OPC cement paste, this work compared existing methods for statistical deconvolution of nanoindentation measurements, while also developing a state-of-the-art NI-QEDS technique for further disclosing the micro-chemo-mechanical properties. Based on the presented results, the following conclusions can be drawn:

  • 1.

    Deconvolution based solely on mechanical properties resulted in clusters including data points having similar mechanical properties but

Acknowledgements

The authors wish to thank the Natural Sciences and Engineering Research Council of Canada (NSERC) [grant numbers RGPIN-2016-05985 and 6636-438380-2013] and the Research Center on Concrete Infrastructure (CRIB) for the financial support provided to perform this research work.

References (110)

  • G. Constantinides et al.

    Grid indentation analysis of composite microstructure and mechanics: principles and validation

    Mater. Sci. Eng. A

    (2006)
  • J. Němeček et al.

    Nanoindentation characteristics of alkali-activated aluminosilicate materials

    Cem. Concr. Compos.

    (2011)
  • J. Němeček et al.

    Tensile strength of hydrated cement paste phases assessed by micro-bending tests and nanoindentation

    Cem. Concr. Compos.

    (2016)
  • N. Shanahan et al.

    Multi-technique investigation of metakaolin and slag blended portland cement pastes

    Appl. Clay Sci.

    (2016)
  • J. Němeček et al.

    Micromechanical analysis of heterogeneous structural materials

    Cem. Concr. Compos.

    (2013)
  • M.J. DeJong et al.

    The nanogranular behavior of C-S-H at elevated temperatures (up to 700C)

    Cem. Concr. Res.

    (2007)
  • W. Zhu et al.

    Nanoindentation mapping of mechanical properties of cement paste and natural rocks

    Mater. Charact.

    (2007)
  • L. Sorelli et al.

    The nano-mechanical signature of ultra high performance concrete by statistical nanoindentation techniques

    Cem. Concr. Res.

    (2008)
  • L.P. Singh et al.

    Quantification and characterization of C-S-H in silica nanoparticles incorporated cementitious system

    Cem. Concr. Compos.

    (2017)
  • M. Zhang et al.

    A multiscale investigation of reaction kinetics, phase formation, and mechanical properties of metakaolin geopolymers

    Cem. Concr. Compos.

    (2017)
  • S. Barbhuiya et al.

    Microstructure, hydration and nanomechanical properties of concrete containing metakaolin

    Constr. Build. Mater.

    (2015)
  • W.R.L. Da Silva et al.

    Methodology for nanoindentation-assisted prediction of macroscale elastic properties of high performance cementitious composites

    Cem. Concr. Compos.

    (2014)
  • J. Xu et al.

    Modification effects of nanosilica on the interfacial transition zone in concrete: a multiscale approach

    Cem. Concr. Compos.

    (2017)
  • L. Xu et al.

    Nano-mechanical behavior of the interfacial transition zone between steel-polypropylene fiber and cement paste

    Constr. Build. Mater.

    (2017)
  • W.R.L. Da Silva et al.

    Application of multiscale elastic homogenization based on nanoindentation for high performance concrete

    Adv. Eng. Softw.

    (2013)
  • M. Miller et al.

    Surface roughness criteria for cement paste nanoindentation

    Cem. Concr. Res.

    (2008)
  • M. Vandamme et al.

    Nanoindentation investigation of creep properties of calcium silicate hydrates

    Cem. Concr. Res.

    (2013)
  • M. Sebastiani et al.

    High-resolution high-speed nanoindentation mapping of cement pastes: Unravelling the effect of microstructure on the mechanical properties of hydrated phases

    Mater. Des.

    (2016)
  • W. Ashraf et al.

    Multiscale characterization of carbonated wollastonite paste and application of homogenization schemes to predict its effective elastic modulus

    Cem. Concr. Compos.

    (2016)
  • H. Lee et al.

    Mechanical and micromechanical properties of alkali activated fly-ash cement based on nano-indentation

    Constr. Build. Mater.

    (2016)
  • M. Vandamme et al.

    Nanogranular packing of C-S-H at substochiometric conditions

    Cem. Concr. Res.

    (2010)
  • D. Davydov et al.

    Critical aspects of nano-indentation technique in application to hardened cement paste

    Cem. Concr. Res.

    (2011)
  • C. Hu et al.

    Micromechanical investigation of Portland cement paste

    Constr. Build. Mater.

    (2014)
  • C. Hu et al.

    Property investigation of calcium-silicate-hydrate (C-S-H) gel in cementitious composites

    Mater. Charact.

    (2014)
  • W. Wilson et al.

    The micromechanical signature of high-volume natural pozzolan concrete by combined statistical nanoindentation and SEM-EDS analyses

    Cem. Concr. Res.

    (2017)
  • C. Hu et al.

    Statistical nanoindentation technique in application to hardened cement pastes: influences of material microstructure and analysis method

    Constr. Build. Mater.

    (2016)
  • C. Hu et al.

    Micro-mechanical properties of calcium sulfoaluminate cement and the correlation with microstructures

    Cem. Concr. Compos.

    (2017)
  • C. Hu et al.

    Micromechanical investigation of magnesium oxychloride cement paste

    Constr. Build. Mater.

    (2016)
  • C. Hu et al.

    Property investigation of individual phases in cementitious composites containing silica fume and fly ash

    Cem. Concr. Compos.

    (2015)
  • F.-J. Ulm et al.

    Does microstructure matter for statistical nanoindentation techniques?

    Cem. Concr. Compos.

    (2010)
  • P.J.M. Monteiro et al.

    The elastic moduli of calcium hydroxide

    Cem. Concr. Res.

    (1995)
  • C.J. Haecker et al.

    Modeling the linear elastic properties of Portland cement paste

    Cem. Concr. Res.

    (2005)
  • K. Velez et al.

    Determination by nanoindentation of elastic modulus and hardness of pure constituents of Portland cement clinker

    Cem. Concr. Res.

    (2001)
  • J. Frech-Baronet et al.

    New evidences on the effect of the internal relative humidity on the creep and relaxation behaviour of a cement paste by micro-indentation techniques

    Cem. Concr. Res.

    (2017)
  • P. Trtik et al.

    A critical examination of statistical nanoindentation on model materials and hardened cement pastes based on virtual experiments

    Cem. Concr. Compos.

    (2009)
  • P. Lura et al.

    Validity of recent approaches for statistical nanoindentation of cement pastes

    Cem. Concr. Compos.

    (2011)
  • C. Hu et al.

    A review on the mechanical properties of cement-based materials measured by nanoindentation

    Constr. Build. Mater.

    (2015)
  • I.N. Sneddon

    The relation between load and penetration in the axi-symmetric Boussinesq problem for a punch of arbitrary profile

    Int. J. Eng. Sci.

    (1965)
  • K.J. Krakowiak et al.

    Nano-chemo-mechanical signature of conventional oil-well cement systems: effects of elevated temperature and curing time

    Cem. Concr. Res.

    (2015)
  • N. Venkovic et al.

    Nanoindentation study of calcium silicate hydrates in concrete produced with effective microorganisms-based bioplasticizer

    Cem. Concr. Compos.

    (2014)
  • Cited by (79)

    • High-speed nanoindentation mapping: A review of recent advances and applications

      2023, Current Opinion in Solid State and Materials Science
    View all citing articles on Scopus
    View full text