Study on the nanoscale mechanical properties of graphene oxide–enhanced shear resisting cement

2.1 Raw materials and sample preparation

The ordinary Portland cement (OPC, Type P.O 42.5) and commercially available 4 mg·mL⁻¹ mass concentration GO suspensions were used to fabricate GO-enhanced cement. Polycarboxylate-based superplasticizer (PC) was used to improve the workability of fresh cement. The irregular polygonal-shaped GO used in this study has a thickness of around 1.5 nm and micron-sized side length, and it has wrinkled surface morphology (as shown in Figure S1). The Fourier transform infrared spectroscopy and Raman spectrums of GO are included in Figure S2. Both the hydroxyl and carboxyl groups can be found on GO sheets. Unlike carbon nanotubes that often naturally agglomerate in solid form in water [8], GO nanosheets are hydrophilic and highly dispersible in water because of these oxygen-containing groups [7,8,9]. The ultraviolet-vis spectrometer testing results show that when the GO suspension had been diluted 100 times, its absorbance was about 1.98 at the wavelength of 230 nm, and the absorbance remained almost unchanged after at least 1 hour. This observation indicated the full and stable dispersion of GO nanosheets in water [37], which is the prerequisite for dispersing GO in cement [19]. Besides, the addition of PC introduced the extra surfactant to enhance the dispersion of GO in the cementitious environment [8,37].

Table 1 presents the mixing design of GO-enhanced cement. Two series of GO-enhanced cements, with G/p of 0.02 wt% (Go-1) and 0.04 wt% (Go-2), were prepared. G/p represents the weight percentage of GO to cement. Plain cement (Re-0) with a G/p of 0 wt% was prepared for comparison. The water-to-cement ratio (w/c) of Re-0, Go-1, and Go-2 materials was 0.55, which is commonly used to fabricate cement sustaining shear [27,28].

The original GO suspensions were diluted to 2 mg·mL⁻¹ under water-bath ultrasonication for 2 min to separate the possibly agglomerated GO nanosheets [23]. After dilution and ultrasonication, the GO suspensions were mixed with PC aqueous solutions, and then, the mixtures were poured into the dry cement and stirred using a high-shear mixer (a Model 7000 constant speed mixer from Cement Test Equipment). The stirring process follows the procedure specified in ASTM Standard C1738 [38]. The high fluidity GO-enhanced cement was then cast into Φ50 × 50 mm cylindrical steel molds and vibrated to release any residual air bubbles. After curing for 24 h at 25°C and 85% relative humidity, the samples were demolded and cured in the same environment for another 27 days before the shear strength testing.

2.2 Instrumentation and characterization

Mini-slump tests were conducted to characterize the fluidity of the fresh cement [28,39]. The device for mini-slump test is an inverted-funnel shape with upper and lower diameters of 19 and 38 mm, respectively, and a height of 57 mm (more details are included in ref. [39]). The diameter of the free slump flow of fresh cement was measured at five different locations around the outline, and the average diameter was used to estimate the fluidity properties.

To characterize the bleeding rate of GO-enhanced cement, a certain amount of fresh cement was transferred into a sealed storage bottle and cured for 1 hr in the same environment used for curing the hardened samples, collecting the segregated water from the pastes every 10 min until no water was segregated [28]. The bleeding rate was defined as the weight percentage of the segregated water to the total water mixed into the cement pastes. For each kind of cement, three groups of pastes were used for testing the bleeding rate.

Following hardening, the shear strength properties of cement were characterized using a universal servo mechanical testing system (DDL 500 Type) and the variable angle shear test method. The variable angle shear test method has been widely used to estimate the shear strength of quasi-brittle materials like rock [40]- and cement-based composites [31]. As shown in Figure 1(a), on shear plane, σ, the normal stress, and τ, the shear stress, can be calculated using equation (1):

When the angle between the shear plane and horizontal direction (α) is changed, the vertical failure load (σ _n ) of the sample changes, and so do σ and τ. The variation laws of τ with σ can be linearly fitted based on Mohr-Coulomb criteria [41], based on which, two parameters can represent the shear strength of materials, cohesion (c) and internal friction angle (ϕ), can be calculated. Five different α were used in this study, 45, 50, 55, 60, and 65°, and three to four samples were tested under each α. σ _n was applied by applying a constant axial loading rate to shearing device [42]. The rate was 0.10 mm·min⁻¹ to simulate the static loading [31].

Figure 1

Schematic diagrams of (a) variable angle shear test, (b) peak force QNM measurement process, and (c) contact between the nanoindenter tip and material surface.

A Bruker iCon atomic force microscope (AFM) was used to conduct peak force QNM measurements for characterizing the nanoscale Young’s modulus (E) of different materials in hardened cement. As shown in Figure 1(b), the peak force QNM test includes the indentation and withdrawal processes. To be specific, a hand-crafted natural diamond tip was used as an indenter to compress the material until the force reaches the predetermined value, F; then, the indenter withdraws the indentation point and moves to the next location until all the points within the scanned regions are characterized. The indentation process makes the material surface deform, and the monitored indentation depth is defined as δ (Figure 1c). E of materials at the measured point can be identified using the modified Hertzian Model (equation (2)):

where R is the working radius of the indenter and ν is Poisson’s ratio.

R was derived with the relative method [31]. A highly ordered pyrolytic graphite (HOPG) with E of about 18 GPa was used as a reference sample. Based on equation (2) and the value of E for HOPG, R was determined as 30 nm (Figure 1c). For verification, clear fused quartz was characterized before testing the hardened cement. During testing, F was set to 4.0 μN and ν was set to 0.3, according to the previous study on the nanomechanical properties of cement [31]. About 164–655 points were characterized per square micron. To minimize the influence of the sample surface roughness [43], samples were first impregnated with epoxy, followed by surface polishing continuously down to 0.1 μm grit using diamond paste [31]. Through the impregnation, the original nanocracks and nanoholes can be filled with epoxy. The polishing treatment can reduce the surface roughness to meet the flatness requirements of the peak force QNM test. During polishing, a lubricant made of ethanol and propylene glycol was used.

3.1 The mini-slump diameter and bleeding rate

As shown in Figure 2, compared to plain cement, the mini-slump diameters of GO-enhanced cement are almost unchanged, whereas the bleeding rate is reduced. Whether or not the cement pastes contain GO, their mini-slump flow diameter varied in about 164 ± 2 mm range with a maximum variation of around 1.5%, indicating the little disturbed fluidity [39]. However, with increasing mass content of GO, the bleeding rate gradually reduces from about 1.7–1.1%, with a maximum reduction of about 34%. This is because single-layered GO has a large specific surface area [2]; therefore, GO in cement can absorb more PC and water molecules due to the strong van der Waals interaction, which helps to prevent the water from segregating from the fresh pastes. As a result, the bleeding rate of cement decreases when GO is added. The decreased bleeding rate indicates that the less settlement of cement grains is caused by gravity and the pastes were more stable during hardening [28].

Figure 2

(a) Mini-slump diameter and (b) bleeding rate of fresh cement. Error bars indicate standard deviations.

3.2 The shear strength

Figure 3 presents the shear strength of the hardened cement under different normal stresses (σ). From Figure 3, one can see that τ of the three types of cement increases when σ increases and the increasing trend is linear. The fitted functions have the same form as the Mohr–Coulomb function [41]. According to the fitted functions, the shear strength of the GO-enhanced cement is higher than that of the plain cement under the same normal stress, and the higher the mass ratios of GO, the higher the shear strength. For example, when σ is 10 MPa, the estimated τ of the Re-0 samples is about 12.9 MPa, whereas those of the Go-1 and Go-2 samples are 14.1 and 16.0 MPa, respectively.

Figure 3

The linear relationship between τ and σ of cement with GO mass ratios of (a) 0, (b) 0.02, and (c) 0.04 wt%. Error bars indicate standard deviations.

Figure 4 shows the cohesion (c) and internal friction angle (ϕ) of Re-0, Go-1, and Go-2 samples. c and ϕ were calculated based on the Mohr–Coulomb function and the linear relationship between τ and σ. From Figure 4, one can see that when G/p increases from 0 to 0.04 wt%, the average c of samples increases from about 7.10 to 9.89 MPa, and it increased by about 12.1 and 39.3%, respectively. However, the average ϕ remains almost unchanged when G/p varies, with a maximum variation of about 4%. The average ϕ of Re-0, Go-1, and Go-2 samples are about 30.3, 31.5, and 31.6°, respectively. These observations indicate that the higher shear strength of cement containing GO is primarily caused by the enhanced cohesion.

Figure 4

(a) Cohesion and (b) internal friction angle of the Re-0, Go-1, and Go-2 samples. Error bars indicate standard deviations.

3.3 Nanoscale mechanical properties

The nanoscale E of HOPG and clear fused quartz are presented in Figure 5(a) and (b), from which one can see that the E of HOPG and quartz are about 18–20 GPa and 60–80 GPa, respectively. The E of quartz measured here is consistent with literature values [44]. This indicates that materials with Young’s modulus range of 18–80 GPa can be reliably characterized with the peak force QNM method. Since the main compositions in cement matrices are often softer than quartz but harder than HOPG [31], the E characterizations of the cement materials at nanoscale are reliable.

Figure 5

Distribution maps of the nanoscale Young’s modulus of (a) HOPG; (b) clear fused quartz; and (c)–(e) different materials in cement with G/p of 0, 0.02, and 0.04 wt%. All these scanned regions share the same scale bar presented in (e).

Figure 5(c)–(e) shows the representative distribution maps of E of different materials in the characterized regions, and more distribution maps of E are included in Figure S3. The different E ranges are painted with different colors (as shown in the color bar). One can see that the multiple materials in hardened cement have a wide E range, from about 10 GPa to higher than 70 GPa, and there often exist unhydrated cement grains (U-c), hydration products, and some micro- or nanoholes/cracks [45]. The primary source of U-c is the original cement grains that are too large to be completely dissolved [45], and they were identified based on the already reported elastic modulus of cement clinkers [46]. The micro- or nanoholes/cracks exist because the hydration products often could not completely fill the space previously occupied by cement pore solutions. Because the nanoholes/cracks have been filled with epoxy during the impregnation process, they can be identified based on the E range of epoxy [31]. In Figure 5, the areas highlighted in red represent the nanoholes/cracks. Hydration products were categorized as the low-density hydration product (LD-h) and high-density hydration product (HD-h) based on the previous nanoindentation results [31]. It was also found that materials adjacent to U-c have higher E than HD-h but they are softer than U-c. These materials were defined as the transition products (T-p), and they are mainly made of hydrates and various ions dissolved from cement clinkers [45,47].

Figure 5(c)–(e) and Figure S3 show that U-c particles are randomly and individually embedded in cement matrices and that the original holes/cracks are surrounded by LD-h material. Compared to LD-h, HD-h is closer to U-c, indicating that during hardening, the deposition and crystallization of hydrates mainly occur on or nearby cement clinkers [45]. According to the distribution maps of E presented in Figure 5 and Figure S3, the probability density characteristics of E can be obtained, and they are shown in Figure 6.

Figure 6

The probability density of E of (a) Ref-0, (b) Go-1, and (c) Go-2 samples (with the goodness of fit of 0.987, 0.979, and 0.968, respectively).

The probability densities of E shown in Figure 6 are characterized as multiple peaks. The probability density can be fitted using a multimodal Gaussian function [31] (equation (3), with a goodness of fit of over 0.96). Based on fitting results, the average values of E of LD-h, HD-h, and T-p can be obtained, and they are represented by the abscissas of each peak [31]. Table 2 presents the average E of LD-h, HD-h, and T-p, from which one can see that the average E of T-p is about 48% higher than that of HD-h and that the addition of GO in cement brings little influence on the elastic modulus of HD-h (with E of around 27 GPa) but does enhance LD-h (with the increase in average E of about 1.6 and 11%). Additionally, the probability density curves also show that with the increasing mass content of GO, the peak of the sub curve representing U-c grains declines, whereas that of the sub curve representing HD-h increases and that the widths of these bell-shaped subcurves vary significantly, indicating that the cumulative probability (P _d) of E of different materials changes. P _d was calculated using the following equation:

where E _a to E _b , respectively, represent the lowest and greatest value of the E range of different materials in cement (as shown in Figure 6a).

Figure 7 shows P _d of different materials in cement with different G/p, from which one can see that when G/p increases to 0.04 wt% from 0 wt%, the P _d of LD-h decreases from about 55% to about 45%, whereas that of HD-h material increases from about 24.5% to about 36.9% and that of U-c grains declines from about 6.4% to about 3.9%. P _d of H&C in Go-1 and Go-2 cement also, respectively, declined by about 4.6 and 12.6% compared to Ref-0 cement. P _d can represent the volume ratio of each kind of material, and it can be deduced that the total hydration products occupy over 75% volume of the solid matrices. The less amount of H&C implies that the addition of GO in cement can make the microstructure denser. The increased amount of HD-h material and fewer U-c grains indicate that the hydration degree of cement was enhanced when GO was used [45].

Figure 7

The cumulative probability (P _d) of different materials in cement with different G/p. Error bars indicate standard deviations.

As demonstrated in Figure 8(a), average Young’s modulus of hydration products locating on a line in which the distance from the boundary of a certain U-c grain (U _j) is l can be calculated by equation (5):

where E(ρ, θ _m ) _j is Young’s modulus of grain located at the coordinate of (ρ, θ _m ), and this grain is with a distance of l from the boundary of U _j grain; n is the quantity of all grains of which the distance from the boundary of U _j grain is l; k is the total quantity of the U grains.

Figure 8

(a) Schematic diagram shows the identification of E of certain particles of which the distance from the boundary of a U-c grain is l. (b)–(d) Average E of particles with different distances from the boundary of U-c grains, and (e)–(g) schematic diagrams illustrate the evaluation of average E of materials in matrices with G/p of 0, 0.02, and 0.04 wt%.

Figure 8(b)–(d) shows that E̅ of materials near U grain in both Ref-0, Go-1, and Go-2 samples varies at a relatively high level (about 30 GPa). At l of about 4–6 μm, E̅declined by about 30% for the Ref-0 material because of the increased LD-h phase (Figure 5), whereas E̅ remained almost unchanged for the Go-1 and Go-2 materials. The difference in E̅ at l of about 4–6 μm indicates that the volume fractions of the HD-h phase in GO-reinforced cement are higher, consisting well with the distribution maps of E. For the Go-1 and Go-2 materials, E̅ starts to decline at l of about 13 μm and 15 μm.

Cement grains work as reactive core sites for the development of hydration products. Hydration products grown on or near the surface of core sites possess a higher density than those formed in the space that is originally occupied by the pore solution (Figure 5a and b). Therefore, average Young’s modulus of materials declines with the increasing distance. The increased l at which E̅ starts to decline indicates that the HD-h phase in GO-reinforced cement spread further from the core sites. The extending range of the HD-h phase in GO-reinforced cement was about 2–3 times that of the HD-h phase in plain cement. After reduction, E̅ of materials in GO-reinforced cement can be about 5–10% higher than that of hydrates in plain cement. It can also be noticed that after reduction to a plateau, E̅ may start to increase again at l of about 10 μm for the Ref-0 materials, which is due to the decreased distance between U-c grains.

3.4 Discussion

The hydration reaction of cement is a process of dissolving cement clinkers and forming and crystallizing hydration products [45]. During the normal hydration, the already formed hydrates mainly deposit on or near surfaces of cement grains. Researchers have claimed that the deposited hydrate particles are connected gradually to form a “shell structure” surrounding cement grains and that with the continuous precipitation and polymerization of hydrates, the shell structure thickens and becomes less porous [47]. As a result, both the further diffusion of hydrates through the shell structure to pore solutions and the passage of water molecules through the shell structure to contact cement clinkers can be retarded. When the hydrates and water molecules are no longer exchanged through the “shell,” the water molecules beneath the shell structure will be gradually consumed, at which point the hydration of this cement grain will stop. During this process, the newly formed hydrates will accumulate beneath the shell to form high-density matrices [47].

In fresh cement, most GO nanosheets distribute in pore solutions [48], and by virtue of the superior specific surface areas and high aspect ratios, GO nanosheets can work as extra substrates for the precipitation of hydrate particles [49]. Hydration products can thus be found in the pore space between cement grains rather than only on or near them [11]. The hydration products will be more uniformly distributed in the mixtures, and the medium surrounding the clinker grains will be less polymerized at the same hydration time [11]. As a result, the formation of the shell structure is retarded, and the retarding effect of the shell structure on ions/water exchange can be weakened; therefore, the hydration reaction of the GO-reinforced cement materials can be improved and more high-density hydration product is produced and less residual U-c are left (Figure 5).

As mentioned in Section 3.2, the higher shear strength of GO-enhanced cement is due to the higher cohesion. For hardened cement, the strength is predominately contributed by the calcium silicate hydrate (C–S–H) [50,51]. The charged C–S–H particles are bonded together through the ionic attractive interaction between them to form solid matrices that also bond other grains [50,51]. Young’s modulus of hydration products reflects the strength of the electrostatic interaction between C–S–H [52]. To explain this, here, the hydration products are simplified as collections of nanoscale crystals of idealized cubic geometry shape, and C–S–H grains are considered as the rigid nodes located at the vertices of the cubic crystal (as shown in Figure 9) [52].

Figure 9

Schematic diagram of the simplified cement hydration products’ crystal structure with cubic geometry shape and the variations of potential energy (U) of a bond with the bond length (r).

As shown in Figure 9, when the crystal is compressed in the y-direction, since this idealized crystal structure can be considered isotropic [53,54], the four bonds parallel to the y-direction will be compressed simultaneously and equally; therefore, the compression of the crystal can be simply considered a process in which two rigid spheres approach each other from their equilibrium positions (Figure 9) [55]. During this process, the bond can provide repulsive force to the nodes, and the whole stiffness can be calculated as follows [52]:

where U represents the potential energy and r ₀ and r represent the equilibrium bond length and the compressed bond length, respectively.

Equation (6) intuitively shows that E of this crystal is proportional to d U d r , which represents the bond strength, i.e., the interaction force f (Figure 9) [52,55]; therefore, the interaction force between the hypothetical nodes making up materials of higher E is stronger. To prove this viewpoint, when conducting the peak force QNM measurement, the pull-off force between cement materials and the diamond indenter at one characterized region was captured. The region is the one shown in Figure 5(e), and the result is drawn in Figure 10.

Figure 10

A map of the nanoscale pull-off force between GO-enhanced cement and diamond indenter (×10⁻⁷ N). This region is the one presented in Figure 5(e).

At nanoscale, the pull-off force is caused by the attractive interaction between C–S–H matrices and indenter [56,57,58]. Figure 5(e) and Figure 10 comprehensively show that in hardened cement, the regions occupied by hydration products with higher E have a stronger adhesion interaction with the indenter than those occupied by the softer materials. This observation indicates that at the nanoscale, materials with higher stiffness have stronger attractive interaction with the surrounding materials, which is consistent with the inference based on equation (5); that is, stronger force is required to overcome the attractive interaction force between C–S–H grains that compose high-density hydration products [52,55]. In hardened cement, cracks often tend to grow from the original defects and propagate into the solid materials gradually [31]. Due to the higher nanoscale Young’s modulus and the increased volume ratio of the high-density hydration products in GO-enhanced cement (Table 2 and Figure 7), it is more likely happen that the cracks propagate into materials with higher stiffness, and thus, more fracturing energy is required to break the material [31,52]. Due to the enhanced bonding effect between C–S–H particles, the cohesion of cement matrices is enhanced; thus, the shear strength is improved (as shown in Figure 4).

Additionally, it is also believed that the carboxylic acid groups attached to GO can chemically react with C–S–H or Ca(OH)₂ during hydration [23]. Such reactions have been mentioned in a previous study on the mechanical properties of functionalized carbon nanotube–enhanced cement [59], and these reactions can introduce strong covalent bonds on the interface between GO and hydrates [60]; therefore, the load transferring efficiency from cement to GO sheets can be increased, leading to the enhanced internal bonding effect, which also contributes to the enhanced cohesion.