Mineral-Based Coating of Plasma-Treated Carbon Fibre Rovings for Carbon Concrete Composites with Enhanced Mechanical Performance

Surfaces of carbon fibre roving were modified by means of a low temperature plasma treatment to improve their bonding with mineral fines; the latter serving as an inorganic fibre coating for the improved mechanical performance of carbon reinforcement in concrete matrices. Variation of the plasma conditions, such as gas composition and treatment time, was accomplished to establish polar groups on the carbon fibres prior to contact with the suspension of mineral particles in water. Subsequently, the rovings were implemented in a fine concrete matrix and their pull-out performance was assessed. Every plasma treatment resulted in increased pull-out forces in comparison to the reference samples without plasma treatment, indicating a better bonding between the mineral coating material and the carbon fibres. Significant differences were found, depending on gas composition and treatment time. Microscopic investigations showed that the samples with the highest pull-out force exhibited carbon fibre surfaces with the largest areas of hydration products grown on them. Additionally, the coating material ingresses into the multifilament roving in these specimens, leading to better force transfer between individual carbon filaments and between the entire roving and surrounding matrix, thus explaining the superior mechanical performance of the specimens containing appropriately plasma-treated carbon roving.


Introduction
The development of carbon fibre (CF) reinforced concrete has attracted much attention in recent years due to the emerging possibility of designing and producing sophisticated, thin-walled concrete structures, with extraordinary mechanical performance, very high durability, and enhanced potential for free-form designs in comparison to steel bar reinforced concrete [1,2]. The use of carbon reinforcement in the form of textiles or thin bars is not confined to new construction, but is also specifically interesting with respect to the strengthening and repair of existing structures [3]. However, the mechanical properties of such carbon concrete composites deteriorate dramatically under elevated temperatures, especially in the case of exposure to fire, with the result that key construction markets are still closed to them. A crucial issue is the common use of polymer coatings on the multifilament Figure 1 displays the representative load-crack opening displacement curves for the reference specimens, as well as for the specimens reinforced with Mikrodur-coated (MD-coated) CF, which had undergone 100 s of plasma treatment with various gas compositions.  Figure 1 displays the representative load-crack opening displacement curves for the reference specimens, as well as for the specimens reinforced with Mikrodur-coated (MD-coated) CF, which had undergone 100 s of plasma treatment with various gas compositions. At the beginning of the experiment, with deformations clearly below a mere 0.1 mm, each graph shows high transferred loads. This part of the curve represents the behaviour before and during crack formation in the concrete matrix at the notch. Hence, this part of the diagram was not taken into account when discussing any CF-concrete matrix interaction; see Figure 2 and Table 1. Beyond the low point of the sudden drop, the courses of the curves differed significantly, depending on the gas used for the plasma treatment. Table 1. Average values and standard deviations of maximum CF pull-out forces, the corresponding crack opening displacements, and the pull-out energy among replicate specimens (numbers in brackets indicate the standard deviation).  Figure 1. Representative pull-out force versus crack-opening-displacement curves for concrete specimens reinforced with a MD-coated and previously plasma-treated (for 100 s) CF roving; effect of various gas compositions can be seen and compared to reference samples.

Specimen Series
At the beginning of the experiment, with deformations clearly below a mere 0.1 mm, each graph shows high transferred loads. This part of the curve represents the behaviour before and during crack formation in the concrete matrix at the notch. Hence, this part of the diagram was not taken into account when discussing any CF-concrete matrix interaction; see Figure 2 and Table 1. Beyond the low point of the sudden drop, the courses of the curves differed significantly, depending on the gas used for the plasma treatment. Table 1. Average values and standard deviations of maximum CF pull-out forces, the corresponding crack opening displacements, and the pull-out energy among replicate specimens (numbers in brackets indicate the standard deviation).   Figure 2. Typical pull-out energies calculated from the curves presented in Figure 1 for two ranges of crack opening displacement (COD): 0.03 mm to 0.3 mm (black columns) and 0.03 mm to COD at the maximum pull-out force (red columns).

Specimen Series
As expected, the reference reinforced by the CF with neither plasma treatment nor an MD-layer yielded the lowest pull-out force, of around 0.25 kN, at the lowest crack opening displacement of 0.43 mm; see Table 1. The subsequent gradual decrease in the pull-out force with further crack opening indicates a decreasing resistance to the CF pull-out from the concrete matrix; see Figure 1. This curve progression is characteristic of the insufficient load-bearing ability of the bond between the inherently non-polar CF reinforcement and the highly polar, water-based concrete matrix. The MD-coating of untreated CF resulted in a slight increase in the average maximum pull-out force, to 0.27 kN, at an average crack opening displacement of 0.51 mm (Table 1). Beyond the maximum value of the pull-out force-crack opening, the curves exhibit a narrow plateau before a gradual descent of the curve sets in; see Figure 1. Principally, such behaviour can be interpreted as fairly similar to the MD-free reference. The slightly higher maximum pull-force may be related to an expected increase in friction forces and interlocking, due to the fine MD-layer coating of the CF. The performance is still governed by CF pull-out from the concrete matrix, indicating rather poor CF/concrete and CF/MD bonding interactions.
Two out of the three types of plasma-treatment caused significant changes in the pull-out force versus the crack opening behaviour. The samples made of CF modified with plasma generated from pure oxygen or with a mixture of oxygen and argon, featured a dramatic increase in the transferred forces over a wide range of crack openings. For both of these types of plasma treatment, a pronounced ascent in the pull-out force of up to approximately 0.1 mm to 0.2 mm in the crack opening was observed, followed by a further moderate increase in the force with increasing crack opening displacement ( Figure 1).
The CF treated with oxygen plasma exhibited an average maximum pull-out force of 0.50 kN, at a crack-opening displacement of 1.13 mm. After reaching the maximum pull-out force, a sharp drop in force was measured, followed by steady crack opening at a constant force. The samples made of CF modified with the plasma working with a mixture of oxygen and argon, showed an increase in the pull-out force up to 0.45 kN, at a 0.84 mm crack opening displacement; see Table 1. Subsequently, the force dropped steeply to the lowest value, indicating a full mechanical failure of the CF roving; cf. Figure 1. Apparently, both types of plasma treatment led to remarkably improved compatibility between the CF and the MD-coating and, finally, with the concrete matrix. Chemical and microstructural interpretation will be provided in Section 2.2.
As a further plasma mixture, a combination of oxygen and tetrafluoromethane (O2/CF4) was applied to CF. Cement-based composite specimens containing this CF only exhibited a marginal increase in the maximum force, to 0.31 kN, in comparison to the reference samples ( Table 1). The  Figure 1 for two ranges of crack opening displacement (COD): 0.03 mm to 0.3 mm (black columns) and 0.03 mm to COD at the maximum pull-out force (red columns).
As expected, the reference reinforced by the CF with neither plasma treatment nor an MD-layer yielded the lowest pull-out force, of around 0.25 kN, at the lowest crack opening displacement of 0.43 mm; see Table 1. The subsequent gradual decrease in the pull-out force with further crack opening indicates a decreasing resistance to the CF pull-out from the concrete matrix; see Figure 1. This curve progression is characteristic of the insufficient load-bearing ability of the bond between the inherently non-polar CF reinforcement and the highly polar, water-based concrete matrix. The MD-coating of untreated CF resulted in a slight increase in the average maximum pull-out force, to 0.27 kN, at an average crack opening displacement of 0.51 mm (Table 1). Beyond the maximum value of the pull-out force-crack opening, the curves exhibit a narrow plateau before a gradual descent of the curve sets in; see Figure 1. Principally, such behaviour can be interpreted as fairly similar to the MD-free reference. The slightly higher maximum pull-force may be related to an expected increase in friction forces and interlocking, due to the fine MD-layer coating of the CF. The performance is still governed by CF pull-out from the concrete matrix, indicating rather poor CF/concrete and CF/MD bonding interactions.
Two out of the three types of plasma-treatment caused significant changes in the pull-out force versus the crack opening behaviour. The samples made of CF modified with plasma generated from pure oxygen or with a mixture of oxygen and argon, featured a dramatic increase in the transferred forces over a wide range of crack openings. For both of these types of plasma treatment, a pronounced ascent in the pull-out force of up to approximately 0.1 mm to 0.2 mm in the crack opening was observed, followed by a further moderate increase in the force with increasing crack opening displacement ( Figure 1).
The CF treated with oxygen plasma exhibited an average maximum pull-out force of 0.50 kN, at a crack-opening displacement of 1.13 mm. After reaching the maximum pull-out force, a sharp drop in force was measured, followed by steady crack opening at a constant force. The samples made of CF modified with the plasma working with a mixture of oxygen and argon, showed an increase in the pull-out force up to 0.45 kN, at a 0.84 mm crack opening displacement; see Table 1. Subsequently, the force dropped steeply to the lowest value, indicating a full mechanical failure of the CF roving; cf. Figure 1. Apparently, both types of plasma treatment led to remarkably improved compatibility between the CF and the MD-coating and, finally, with the concrete matrix. Chemical and microstructural interpretation will be provided in Section 2.2.
As a further plasma mixture, a combination of oxygen and tetrafluoromethane (O 2 /CF 4 ) was applied to CF. Cement-based composite specimens containing this CF only exhibited a marginal increase in the maximum force, to 0.31 kN, in comparison to the reference samples ( Table 1). The general shape of the corresponding force-displacement curve indicates no obvious success in activating the CF surfaces; cf. Figure 1. Potentially, the formed fragments in the plasma, e.g., CF 4 , could react with the carbon-based epoxy sizing layer by breaking the bonding and dissolving the carbon, which would result in an increased oxygen content at the surface. However, the effect was obviously negligible with respect to enhancing the CF-to-matrix bonding. Moreover, it should be noted that oxidation is suppressed by the effect of the fluorine atoms' catching electrons and therefore prevents the formation of oxygen radicals and negative oxygen ions.
The interpretation of the mechanical performance of the carbon concrete composite specimens was continued with respect to the pull-out energy calculated as the areas under the force-crack opening displacement curves; see Figure 2 and Table 1. For reasons of better comparability, each curve was integrated within two particular ranges. Firstly, the initial part of the curve was analysed. Therefore, integration occurred for crack-opening displacements between 0.03 mm and 0.3 mm, for any sample; see black columns in Figure 2. The lower limit of 0.03 mm represents the border between a simple matrix fracture and the onset of CF pull-out. Within the integration limits, i.e., 0.03 to 0.30 mm, both the references and the sample treated with the O 2 /CF 4 plasma exhibited a pull-out energy between 0.05 kJ/mm 2 and 0.06 kJ/mm 2 . The other two plasma-treated samples showed significantly higher values, of 0.075 kJ/mm 2 (for O 2 plasma) and 0.09 kJ/mm 2 (for O 2 /Ar plasma). The increase in pull-out energy becomes much more pronounced when the upper limit is set to the crack opening displacement at the maximum load of each sample; see red columns in Figure 2. These integration limits were used for averaging and assessing the standard deviation as a statistical indicator of trustworthiness; see results in Table 1.
For the sample containing CF treated with pure oxygen plasma, a pull-out energy of 0.43 kJ/mm 2 was calculated, which corresponds to an increase of more than four times, in comparison to the reference samples. The sample containing CF modified with the O 2 /Ar plasma also exhibited a high increase in pull-out energy (0.30 kJ/mm 2 , i.e., an increase of three times, compared to the reference). In contrast, the samples made with CF modified by the O 2 /CF 4 plasma did not markedly differ from the reference samples. In addition to the qualitative curve shape, this result indicates a similar failure mechanism in these composites, which is dominated by friction between the CF and surrounding matrix, while the modification by oxygen plasma and oxygen/argon plasma resulted in an enhanced bonding between the components due to the introduction of polar groups (cf. Section 2.2).
For the two successful plasma treatments thus far, i.e., by using oxygen or the oxygen/argon mixtures, the treatment time was varied in the next step. Table 1 and Figure 3 disclose that in both systems, the maximum fibre pull-out force severely decreased in the case of the longer plasma treatment time of 200 s. By and large, the pull-out energies were nearly cut in half. However, the COD values at the maximum pull-out force did not change considerably, due to the prolonging of the plasma treatment. Despite the decreased maxima, the O 2 plasma remained superior in comparison to the O 2 /Ar plasma, with respect to the positive effects of both on the mechanical performance of carbon concrete composites. The observed decrease in the maximum pull-out force can be explained by the enhanced deterioration of the epoxy resin by plasma-induced oxidation, which was advanced by longer treatment; an effect which also known for ceramic fibres [31]. As the results show and as could be expected, this influence is higher in the case of pure oxygen; see also Figure 4. An intensified oxidation can lead to a physical surface modification due to locally varying abrasion [32]. The physical condition of the surface governs the tensile properties of the CF roving, even prior to its coating with inorganic particles.  . Element ratio of plasma-treated carbon fibres measured by X-ray photo electron spectroscopy; gas mixtures and treatment times are in accordance with Table 3.
To elucidate in this respect, the tensile strength of the carbon fibre roving subject to respective plasma treatments was determined; see Table 2. The plasma-treated roving exhibits a tendency for higher tensile strengths, a finding previously reported by Dilsiz et al. [33]. However, such an effect could be explained by the procedure of sample preparation before plasma treatment. Owing to this, the roving experienced a stretching process, which results in an alignment of the individual CF filaments within the roving's cross-section.
Consequently, a more uniform utilisation of all filaments was achieved in the case of mechanical loading and, hence, higher tensile strengths were achieved for CF. With an increasing duration of the plasma treatments, a decrease of 5% to 12% in the rovings' tensile strengths occurred, which agrees with the results reported in the literature [34]. The longer the plasma treatment duration, the rougher the CF surface becomes [32], ending in the destruction of some CF filaments. Consequently, there seems to be an optimum treatment time which enables the wetting of CF by maintaining moderate sizing and avoiding damage to the CF filaments.   . Element ratio of plasma-treated carbon fibres measured by X-ray photo electron spectroscopy; gas mixtures and treatment times are in accordance with Table 3.
To elucidate in this respect, the tensile strength of the carbon fibre roving subject to respective plasma treatments was determined; see Table 2. The plasma-treated roving exhibits a tendency for higher tensile strengths, a finding previously reported by Dilsiz et al. [33]. However, such an effect could be explained by the procedure of sample preparation before plasma treatment. Owing to this, the roving experienced a stretching process, which results in an alignment of the individual CF filaments within the roving's cross-section.
Consequently, a more uniform utilisation of all filaments was achieved in the case of mechanical loading and, hence, higher tensile strengths were achieved for CF. With an increasing duration of the plasma treatments, a decrease of 5% to 12% in the rovings' tensile strengths occurred, which agrees with the results reported in the literature [34]. The longer the plasma treatment duration, the rougher the CF surface becomes [32], ending in the destruction of some CF filaments. Consequently, there seems to be an optimum treatment time which enables the wetting of CF by maintaining moderate sizing and avoiding damage to the CF filaments. . Element ratio of plasma-treated carbon fibres measured by X-ray photo electron spectroscopy; gas mixtures and treatment times are in accordance with Table 3.
To elucidate in this respect, the tensile strength of the carbon fibre roving subject to respective plasma treatments was determined; see Table 2. The plasma-treated roving exhibits a tendency for higher tensile strengths, a finding previously reported by Dilsiz et al. [33]. However, such an effect could be explained by the procedure of sample preparation before plasma treatment. Owing to this, the roving experienced a stretching process, which results in an alignment of the individual CF filaments within the roving's cross-section.
Consequently, a more uniform utilisation of all filaments was achieved in the case of mechanical loading and, hence, higher tensile strengths were achieved for CF. With an increasing duration of the plasma treatments, a decrease of 5% to 12% in the rovings' tensile strengths occurred, which agrees with the results reported in the literature [34]. The longer the plasma treatment duration, the rougher the CF surface becomes [32], ending in the destruction of some CF filaments. Consequently, there seems to be an optimum treatment time which enables the wetting of CF by maintaining moderate sizing and avoiding damage to the CF filaments. From the results of mechanical testing, it can be concluded that the plasma process parameters play a crucial role in successful CF activation, enabling efficient load-bearing, a mineral-based coating of carbon roving, and ensuring an appropriate bond to the concrete matrix. The strikingly enhanced pull-out characteristics of the composites reinforced with oxygen, as well as the oxygen and argon plasmas, can be traced back to the presence of the oxygen-based functional groups on CF surfaces that had been introduced by these two plasmas. XPS results reveal a higher oxygen content at the surface of the carbon fibres after plasma treatment, both for oxygen, as well as for oxygen and argon as working gases, as compared to untreated CF; see Figure 4. The remarkably higher amounts of nitrogen on the fibre surfaces treated with the oxygen/argon plasma arise from the presence of nitrogen atoms and ions in this plasma. These reactive nitrogen species are induced by electrons, due to the electropositive effects of argon. On the other hand, oxygen has a high affinity for electrons, resulting in an electron-poor plasma that principally prevents the generation of nitrogen atoms or ions. Thus, the CF surface exhibits a lower nitrogen content when modified by an oxygen plasma. The presence of oxygen-based functional groups at the surface caused by treatment with oxygen-containing plasma results in a higher surface tension possessing a higher polar part [20,35]. This improves the binding interaction with the water-based slurry of MD and, eventually, with the reactive mineral particles.
To substantiate the macro-mechanical findings from the viewpoint of the cement-based matrix, the composites were characterised morphologically and chemically-mineralogically. Figure 5 provides an overview of the morphology of the carbon filament surfaces after the tensile tests. When considering the reference samples, huge areas of uncoated surfaces lay bare (Figure 5a,b). Close to CF, large clumps of non-homogeneously distributed coating material were found, which obviously had no intense bond to any carbon surface. The reference samples with neither plasma treatment nor MD coating even displayed big portlandite crystals in the near vicinity of CF, due to the direct contact with the surrounding concrete matrix (Figure 5a). However, no hydration products exhibited bonding to any fibre surface. Some few, fine-structured hydration products can be seen locally, but only to a very minor extent. In the reference sample with the MD-layer (Figure 5c,d), clumped structures were present close to CF, showing no significant affinity towards them. Their appearance was similar to the initial MD morphology, having a typical structure for slag-based fines; see Figure 6.
Moreover, at a higher magnification, the reference samples revealed only marginal areas of the carbon filaments' coatings; see Figure 5. It can be concluded that without plasma treatment, no affinity between CF filaments and MD matrix or fine-grained concrete matrix occurred. The hydrophobicity of unmodified CFs did not allow any growth of mineral structures on the surface. This finding explains the poor mechanical performance, as obtained from the pull-out tests. Some slight increase in pull-out forces may result from more pronounced friction and interlocking due to the MD coating. However, the missing electrostatic interaction between the constituents of the water-based mineral matrix and the inherently hydrophobic CF does not enable any considerable enhancement of load transfer between CF and the concrete matrix.   In contrast, the morphological appearance of any embedded plasma-treated CF was found to be fundamentally different to both untreated references. They possess prominent areas of mineral accretions of tightly bonded, finely structured hydration products (especially in cases when O2/Ar plasma and O2/CF4 plasma treatment), as well as residual MD particles (especially in case of O2 sample) on each single filament's surface (Figure 5e-j).
Among these plasma-treated specimens, the CF samples treated with O2/CF4 plasma had the most non-uniform CF coverage by such inorganic material (Figure 5i,j). The partially dense coating of CF filaments alternated with vast bare, uncoated CF areas. Such morphology explains the modest mechanical performance of this composite under tensile loading, with force-crack opening displacement curves similar to the curves obtained for the reference specimens without plasma treatment. There are too few build-ups bonded to CF surfaces, so that each filament can be pulled out smoothly from the surrounding concrete matrix. In contrast, the composite specimens reinforced with roving treated by O2 plasma and O2/Ar plasma featured numerous, large regions of mineral structures growing on the filament surface. These formations give rise to a superior mechanical performance during crack opening, as the consequence of both a high coating density and a very high number of bonding points of the minerals to the yarn. Thus, the surface of each single filament has an intense linkage to the matrix. The bonded mineral layer enables load transfer from the concrete matrix onto each single CF, resulting in increased load bearing capacity. Hence, these composites featured a fundamentally different pull-out behaviour in comparison to the reference samples. Interestingly, the ESEM images suggest that the prevailing minerals attached to the CF surfaces differ in the cases of the oxygen and the oxygen/argon plasmas. On the oxygen-treated filaments, MD grains prevail, whereas typical hydration products from cement and pozzolana are present, only to a lesser extent. Contrarily, more of these hydration products seem to be present on CF surfaces treated with O2/Ar plasma in comparison to the extent of unreacted or partially reacted MD grains. The hydration products are fine-structured interpenetrating needles, and their morphology strongly suggests that, chemically, they are intensely bonded to the CF surface. It appears that they originate from the interface, demonstrating heterogeneous nucleation, rather than having been precipitated out of solution. Furthermore, the remaining MD grains give the impression of an intense adhesion to the surface of each filament. Unlike in the case of the reference (no plasma treatment), no portlandite was found in the interphase layers of the composite specimens containing CF treated with oxygen and oxygen/argon plasmas, the reason for which being the extensive pozzolanic reaction with the locally available slag-based MD grains.
The distinct morphological differences are even more evident in the readily prepared crosssections of the samples; see Figures 7-9. Figure 7 presents selected optical micrographs. None of the CF roving featured a circular or elliptical shape. Instead, irregular shaped roving cross-sections were distributed non-uniformly within the surrounding concrete matrix. As a result, all of the samples possessed various conditions of CF-matrix interfaces. In both reference samples (no plasma treatment), In contrast, the morphological appearance of any embedded plasma-treated CF was found to be fundamentally different to both untreated references. They possess prominent areas of mineral accretions of tightly bonded, finely structured hydration products (especially in cases when O 2 /Ar plasma and O 2 /CF 4 plasma treatment), as well as residual MD particles (especially in case of O 2 sample) on each single filament's surface (Figure 5e-j).
Among these plasma-treated specimens, the CF samples treated with O 2 /CF 4 plasma had the most non-uniform CF coverage by such inorganic material (Figure 5i,j). The partially dense coating of CF filaments alternated with vast bare, uncoated CF areas. Such morphology explains the modest mechanical performance of this composite under tensile loading, with force-crack opening displacement curves similar to the curves obtained for the reference specimens without plasma treatment. There are too few build-ups bonded to CF surfaces, so that each filament can be pulled out smoothly from the surrounding concrete matrix. In contrast, the composite specimens reinforced with roving treated by O 2 plasma and O 2 /Ar plasma featured numerous, large regions of mineral structures growing on the filament surface. These formations give rise to a superior mechanical performance during crack opening, as the consequence of both a high coating density and a very high number of bonding points of the minerals to the yarn. Thus, the surface of each single filament has an intense linkage to the matrix. The bonded mineral layer enables load transfer from the concrete matrix onto each single CF, resulting in increased load bearing capacity. Hence, these composites featured a fundamentally different pull-out behaviour in comparison to the reference samples. Interestingly, the ESEM images suggest that the prevailing minerals attached to the CF surfaces differ in the cases of the oxygen and the oxygen/argon plasmas. On the oxygen-treated filaments, MD grains prevail, whereas typical hydration products from cement and pozzolana are present, only to a lesser extent. Contrarily, more of these hydration products seem to be present on CF surfaces treated with O 2 /Ar plasma in comparison to the extent of unreacted or partially reacted MD grains. The hydration products are fine-structured interpenetrating needles, and their morphology strongly suggests that, chemically, they are intensely bonded to the CF surface. It appears that they originate from the interface, demonstrating heterogeneous nucleation, rather than having been precipitated out of solution. Furthermore, the remaining MD grains give the impression of an intense adhesion to the surface of each filament. Unlike in the case of the reference (no plasma treatment), no portlandite was found in the interphase layers of the composite specimens containing CF treated with oxygen and oxygen/argon plasmas, the reason for which being the extensive pozzolanic reaction with the locally available slag-based MD grains.
The distinct morphological differences are even more evident in the readily prepared cross-sections of the samples; see Figures 7-9. Figure 7 presents selected optical micrographs. None of the CF roving featured a circular or elliptical shape. Instead, irregular shaped roving cross-sections were distributed non-uniformly within the surrounding concrete matrix. As a result, all of the samples possessed various conditions of CF-matrix interfaces. In both reference samples (no plasma treatment), the multifilament strands were embedded rather compactly in the concrete matrix or in the MD, as an adjacent layer. The cement-based matrix or MD had only marginally inserted itself between the CF filaments. In contrast, significant amounts of MD were present throughout the entire cross-section of the multifilament yarns treated with the O 2 plasma. Additionally, a dark-stained ring around the coating layer is to be observed, induced by a densification effect. the multifilament strands were embedded rather compactly in the concrete matrix or in the MD, as an adjacent layer. The cement-based matrix or MD had only marginally inserted itself between the CF filaments. In contrast, significant amounts of MD were present throughout the entire cross-section of the multifilament yarns treated with the O2 plasma. Additionally, a dark-stained ring around the coating layer is to be observed, induced by a densification effect.     the multifilament strands were embedded rather compactly in the concrete matrix or in the MD, as an adjacent layer. The cement-based matrix or MD had only marginally inserted itself between the CF filaments. In contrast, significant amounts of MD were present throughout the entire cross-section of the multifilament yarns treated with the O2 plasma. Additionally, a dark-stained ring around the coating layer is to be observed, induced by a densification effect.     ESEM images of the cross-sections demonstrate the described features even more clearly, and by additional EDX elemental mapping, mineralogical details were elucidated; see Figures 8 and 9. Figure 8 shows the back-scattered electron (BSE) images and the corresponding EDX mapping of the reference samples with an MD coating. At a lower magnification (upper images), the heterogeneous structure of CF (coloured in purple, representing carbon), the surrounding MD layer, and the embedment of that into the concrete matrix, are clearly visible. Moreover, shrinkage cracks can be observed within the MD layer, which appear in the upper EDX image in purple, due to the introduction of epoxy resin into the preparation process. The higher magnification shows a densely packed CF roving, in which only few, rather large cross-sectional areas with interpenetrated MD in the middle of the filament bundle are present. The majority of the single filaments' "neighbours" are further carbon filaments, but neither MD nor the cement-based matrix are in the close vicinity. Hence, the filaments are adjacent to each other in a large proportion, without adhesion-promoting linkage by the mineral matrix.
This gives rise to sliding pull-out without any significant load transfer over the cross-section of the roving. The poor bond properties result in modest mechanical characteristics of carbon-concrete- ESEM images of the cross-sections demonstrate the described features even more clearly, and by additional EDX elemental mapping, mineralogical details were elucidated; see Figures 8 and 9. Figure 8 shows the back-scattered electron (BSE) images and the corresponding EDX mapping of the reference samples with an MD coating. At a lower magnification (upper images), the heterogeneous structure of CF (coloured in purple, representing carbon), the surrounding MD layer, and the embedment of that into the concrete matrix, are clearly visible. Moreover, shrinkage cracks can be observed within the MD layer, which appear in the upper EDX image in purple, due to the introduction of epoxy resin into the preparation process. The higher magnification shows a densely packed CF roving, in which only few, rather large cross-sectional areas with interpenetrated MD in the middle of the filament bundle are present. The majority of the single filaments' "neighbours" are further carbon filaments, but neither MD nor the cement-based matrix are in the close vicinity. Hence, the filaments are adjacent to each other in a large proportion, without adhesion-promoting linkage by the mineral matrix.
This gives rise to sliding pull-out without any significant load transfer over the cross-section of the roving. The poor bond properties result in modest mechanical characteristics of carbon-concrete-composites containing CF without plasma treatment. In contrast, the samples treated with O 2 plasma exhibited much more pronounced infiltrated MD areas within the CF roving; see Figure 9.
The lower magnification already visualises a uniform distribution of carbon filaments (coloured in purple) within the surrounding mineral matrix. Higher magnifications strikingly disclose MD and matrix infiltration all over the cross-section of the multifilament yarn. The fibre/mineral interface was situated not only on the outer filaments of the roving, but was also spread inwardly, to the majority of the inner filaments. This enables a significantly better load transfer from the concrete matrix towards the coated carbon filaments, thus explaining the superior mechanical performance of such composites in the pull-out tests.
At this point, it is useful to recall the visual impressions during the experimental execution of the dip-coating. The MD slurry and the O 2 -plasma-treated CF must have been very affine, because the slurry was strongly drawn into non-submerged regions, which was not the case for the plasma-untreated CF samples during dipping. These findings and observations further underpin the notion that the oxygen-containing plasma induced the collection of oxygen-based polar groups on the CF surface, thus enabling distinctively better wetting [21,24]. Furthermore, the bonding force to mineral particles suspended in the liquid phase greatly increased.

Sample Preparation
A commercially available sample of CF roving with an epoxy-based sizing (Toho Tenax 6K, yarn fineness 400 tex, Toho Tenax Europe GmbH, Wuppertal, Germany) was selected for this fundamental study and used as obtained. A cubic plasma chamber (400 mm × 400 mm × 400 mm) was used for all processes (Figure 10a). Pieces of the roving were mounted on a frame (Figure 10b) and placed in the plasma chamber, so that overall treatment of every outer CF surface was ensured. Treatment of the fibres was performed under low pressure conditions (100 Pa), using pulsed microwave discharges (pulse: 10 ms on, 90 ms off). The discharge power was set at 1200 W. A vacuum was applied, before inserting the gas source and adjusting the pressure. Three different reactive gases or gas mixtures were used for the plasma-based modification of the carbon fibre surfaces: oxygen, oxygen and argon, and oxygen and tetrafluoromethane CF 4 (Table 3). Additionally, the exposure times were varied; they were set at 100 s and 200 s, respectively. composites containing CF without plasma treatment. In contrast, the samples treated with O2 plasma exhibited much more pronounced infiltrated MD areas within the CF roving; see Figure 9. The lower magnification already visualises a uniform distribution of carbon filaments (coloured in purple) within the surrounding mineral matrix. Higher magnifications strikingly disclose MD and matrix infiltration all over the cross-section of the multifilament yarn. The fibre/mineral interface was situated not only on the outer filaments of the roving, but was also spread inwardly, to the majority of the inner filaments. This enables a significantly better load transfer from the concrete matrix towards the coated carbon filaments, thus explaining the superior mechanical performance of such composites in the pull-out tests.
At this point, it is useful to recall the visual impressions during the experimental execution of the dip-coating. The MD slurry and the O2-plasma-treated CF must have been very affine, because the slurry was strongly drawn into non-submerged regions, which was not the case for the plasmauntreated CF samples during dipping. These findings and observations further underpin the notion that the oxygen-containing plasma induced the collection of oxygen-based polar groups on the CF surface, thus enabling distinctively better wetting [21,24]. Furthermore, the bonding force to mineral particles suspended in the liquid phase greatly increased.

Sample Preparation
A commercially available sample of CF roving with an epoxy-based sizing (Toho Tenax 6K, yarn fineness 400 tex, Toho Tenax Europe GmbH, Wuppertal, Germany) was selected for this fundamental study and used as obtained. A cubic plasma chamber (400 mm × 400 mm × 400 mm) was used for all processes (Figure 10a). Pieces of the roving were mounted on a frame (Figure 10b) and placed in the plasma chamber, so that overall treatment of every outer CF surface was ensured. Treatment of the fibres was performed under low pressure conditions (100 Pa), using pulsed microwave discharges (pulse: 10 ms on, 90 ms off). The discharge power was set at 1200 W. A vacuum was applied, before inserting the gas source and adjusting the pressure. Three different reactive gases or gas mixtures were used for the plasma-based modification of the carbon fibre surfaces: oxygen, oxygen and argon, and oxygen and tetrafluoromethane CF4 (Table 3). Additionally, the exposure times were varied; they were set at 100 s and 200 s, respectively.   The plasma-treated carbon fibres were investigated, with respect to their surface chemistry, by X-ray photo electron spectroscopy (XPS; Axis Ultra, Kratos, Manchester, UK). The maximum energy resolution was 0.45 eV (Ag 3d) and the local resolution was in the range of 7 µm. The information depth averaged 10 nm, with a detection limit of 0.1% to 1%. The hydrophilic properties were determined by contact angle measurements (Surftens, OEG GmbH, Frankfurt (Oder), Germany). Moreover, the tensile strengths of the CF rovings were measured according to DIN 3341.
Immediately following the plasma treatment, the mineral coatings were applied. The commercially available powder blend, Dyckerhoff MIKRODUR R-X, abbreviated as MD, based on granulated blast furnace slag [36] and produced by Dyckerhoff Beton GmbH & Co. KG, Wiesbaden, Germany, was used as the mineral coating material. It was manually mixed with tap water in a mass proportion of one-to-one, to obtain a homogenous suspension with a high workability. The freshly plasma-treated CFs were dip-coated by hand. They were immersed to the point of saturation by capillary suction forces for roughly 30 s, to ensure the best possible overall coating. No additional mechanical impregnation was performed. Subsequently, the coated yarns were dried by draping them over a clothesline in the standard atmosphere of the concrete laboratory (20 • C, 65% relative humidity) for seven days, before their incorporation into the concrete matrix.
The composition of the concrete matrix is summarised in Table 4. The cement was a normal Portland cement CEM I 32.5 R (Schwenk, Bernburg, Germany), according to EN 197-1. As pozzolanic additives, a hard-coal fly ash (Steament H-4, STEAG Power Minerals GmbH, Essen, Germany) and silica fume in the form of an aqueous slurry EMSAC 500 SE from BASF, Ludwigshafen, Germany, were used. A local quartz sand was obtained from Kieswerk Ottendorf-Okrilla, Ottendorf-Okrilla, Germany. To ease the workability, the β-naphthalene sulfonate (BNS) high-range water-reducing admixture (HRWRA) MasterRheobuild 30 from BASF, Germany, was added as received. The fine-grained concrete was prepared in a 5 L planetary mixer (HSM20, Hobart, Leipzig, Germany). Each batch had an overall volume of 2 L. Any dry components such as cement, fly ash, and sand were pre-blended for 1 min in the first gear. The liquid components, i.e., silica slurry, HRWRA, and water, were added and mixed for 1 min, also in first gear. Subsequently, stirring was stopped, and residual unmixed materials were removed from the bucket walls and added to the mixture. Homogenisation was continued for an additional 2 min in the second gear.
For each plasma treatment (cf. Table 3), five replicate carbon concrete composite specimens were prepared and characterised. One CF-reinforced concrete specimen series containing MD-coated carbon roving, but without prior plasma treatment, served as a reference (five replicates). Another reference composite, from which ten individual specimens were taken, was made of the virgin CF, with neither plasma treatment nor MD-coating. CF-reinforced composites were prepared in accordance with Figure 11a, in moulds having the specific dimensions displayed in Figure 11b. The formwork introduced a notch of 1 mm × 1 mm to initiate locally defined matrix cracking under tensile loading. After filling the first half of the mould with the matrix, the roving was placed in the centre of the specimen, along its longitudinal axis. Subsequently it was covered with a second layer of fresh, fine-grained concrete. Further formwork imprinting the specific geometry into the upper half of the specimen was tightly clamped on top. Neither vibrating nor densifying and, hence, no further energy input, was applied. The samples were kept for one day in the tight mould and were subsequently immersed in water for six days, according to DIN EN 12 390-2, followed by storage at a standard laboratory climate (20 • C, 65% relative humidity) until mechanical testing. The fine-grained concrete was prepared in a 5 L planetary mixer (HSM20, Hobart, Leipzig, Germany). Each batch had an overall volume of 2 L. Any dry components such as cement, fly ash, and sand were pre-blended for 1 min in the first gear. The liquid components, i.e., silica slurry, HRWRA, and water, were added and mixed for 1 min, also in first gear. Subsequently, stirring was stopped, and residual unmixed materials were removed from the bucket walls and added to the mixture. Homogenisation was continued for an additional 2 min in the second gear.
For each plasma treatment (cf. Table 3), five replicate carbon concrete composite specimens were prepared and characterised. One CF-reinforced concrete specimen series containing MD-coated carbon roving, but without prior plasma treatment, served as a reference (five replicates). Another reference composite, from which ten individual specimens were taken, was made of the virgin CF, with neither plasma treatment nor MD-coating. CF-reinforced composites were prepared in accordance with Figure 11a, in moulds having the specific dimensions displayed in Figure 11b. The formwork introduced a notch of 1 mm × 1 mm to initiate locally defined matrix cracking under tensile loading. After filling the first half of the mould with the matrix, the roving was placed in the centre of the specimen, along its longitudinal axis. Subsequently it was covered with a second layer of fresh, finegrained concrete. Further formwork imprinting the specific geometry into the upper half of the specimen was tightly clamped on top. Neither vibrating nor densifying and, hence, no further energy input, was applied. The samples were kept for one day in the tight mould and were subsequently immersed in water for six days, according to DIN EN 12 390-2, followed by storage at a standard laboratory climate (20 °C, 65% relative humidity) until mechanical testing.

Mechanical Tests, Morphological and Analytical Characterisation
Mechanical characterisation was performed as uniaxial quasi-static double-sided pull-out testing, according to the setup and procedure described by Butler et al. [37]. Notched specimens were tested at an age of 28 days. Five replicate specimens were produced and tested per material composition. A universal servo-hydraulic machine (Instron 8501, Instron, Darmstadt, Germany) equipped with a UK2518 load cell (max. 10 kN) was used under a crosshead displacement-control regime (Figure 12), with a displacement rate of 1 mm/min. The specimen (Figure 12) was clamped at both ends. Applied load versus crack-opening displacement, measured using an extensometer; cf. Figure 12a, was recorded. To ease visual perception, only typical, representative single-test curves are displayed for each parameter combination under investigation. Additionally, the average values of the maximum recorded load and of the crack opening displacement at the maximum load for the replicate specimens were calculated, along with the respective standard deviations. Furthermore, the pull-out energy was determined as the area under the individual load-displacement curves. Among replicate specimens, the pull-out energies were averaged and the standard deviations were calculated. Details of integration borders are provided and explained in Section 2.1.
Morphological characterisation was comprised of the assessment of MD-coated CF and CF-reinforced concrete specimens. In the latter case, two alignments per specimen were examined: (1) the specimens were split in the direction of the fibre; (2) cross-sections were prepared by infiltration with an epoxy resin to later characterise the intergrowth among CF, the MD coating, the concrete matrix, and the hydration products. The tools of instrumental analysis were an environmental scanning electron microscope (ESEM; Quanta 250 FEG, FEI, Eindhoven, The Netherlands), with which each specimen was inspected without further preparatory steps, an energy-dispersing X-ray a) b) Figure 11. (a) half-filled moulds with coated fibre in place; and (b) sample dimensions (in mm).

Mechanical Tests, Morphological and Analytical Characterisation
Mechanical characterisation was performed as uniaxial quasi-static double-sided pull-out testing, according to the setup and procedure described by Butler et al. [37]. Notched specimens were tested at an age of 28 days. Five replicate specimens were produced and tested per material composition. A universal servo-hydraulic machine (Instron 8501, Instron, Darmstadt, Germany) equipped with a UK2518 load cell (max. 10 kN) was used under a crosshead displacement-control regime (Figure 12), with a displacement rate of 1 mm/min. The specimen (Figure 12) was clamped at both ends. Applied load versus crack-opening displacement, measured using an extensometer; cf. Figure 12a, was recorded. To ease visual perception, only typical, representative single-test curves are displayed for each parameter combination under investigation. Additionally, the average values of the maximum recorded load and of the crack opening displacement at the maximum load for the replicate specimens were calculated, along with the respective standard deviations. Furthermore, the pull-out energy was determined as the area under the individual load-displacement curves. Among replicate specimens, the pull-out energies were averaged and the standard deviations were calculated. Details of integration borders are provided and explained in Section 2.1.
Morphological characterisation was comprised of the assessment of MD-coated CF and CF-reinforced concrete specimens. In the latter case, two alignments per specimen were examined: (1) the specimens were split in the direction of the fibre; (2) cross-sections were prepared by infiltration with an epoxy resin to later characterise the intergrowth among CF, the MD coating, the concrete matrix, and the hydration products. The tools of instrumental analysis were an environmental scanning electron microscope (ESEM; Quanta 250 FEG, FEI, Eindhoven, The Netherlands), with which each specimen was inspected without further preparatory steps, an energy-dispersing X-ray analyser (EDX; Quantax 400, Bruker, Billerica, MA, USA) for element mapping, and an optical microscope (AXIOTECH, Zeiss, Jena, Germany) equipped with a live camera (AxioCam ICc 3, Zeiss, Jena, Germany). (a) (b) Figure 12. (a) Setup for uniaxial quasi-static double-sided pull-out tests, according to [37]; (b) sample after testing.

Summary and Conclusions
In this investigation, the plasma treatment of carbon multifilament roving was successfully applied for enhanced interaction with a mineral-based coating and a subsequently improved load transfer to the concrete matrix.
By using three various process plasmas, different interfacial bonding characteristics towards the mineral coating were observed, by accomplishing mechanical pull-out tests on carbon-concretecomposite specimens. The CF treated with pure oxygen plasma showed the most improvement in the mechanical performance in this test, followed by the CF treated with oxygen/argon plasma. An increase in the pull-out energy measured up to the load maximum exhibited a rise of four times and three times the value of the references samples, respectively. In contrast, only a slight increase was detected for the CF treated with O2/CF4 plasma.
The striking differences in the mechanical performance observed for treated and untreated carbon fibres were explained by means of morphological characterisation using ESEM/EDX and optical microscopy. A superior mechanical performance was observed for carbon roving with a uniform coating, which impregnated the interior of the roving and was well bonded to the carbon filaments. Such a coating was unambiguously mediated by the functional groups on the CF surface introduced by plasma treatment, which enabled chemical bonding between the carbon filaments and the mineral coating. In contrast, mostly bare and uncoated CFs were found in the reference samples, which were not plasma-treated. They showed no interaction with the mineral coating or with the hydration products of the fine-grained concrete matrix.
The samples treated with oxygen plasma exhibited an intense impregnation of the mineral binder grains of the coating suspension between the carbon filaments, induced by the strong affinity of the functionalised CF surface and the polar, water-based slurry. This resulted in a pronounced increase in the CF-mineral coating interface and in enhanced bonding between the components, i.e., CF-to-coating and coated roving-to-concrete matrix, thus enabling a persistent mechanical load Figure 12. (a) Setup for uniaxial quasi-static double-sided pull-out tests, according to [37]; (b) sample after testing.

Summary and Conclusions
In this investigation, the plasma treatment of carbon multifilament roving was successfully applied for enhanced interaction with a mineral-based coating and a subsequently improved load transfer to the concrete matrix.
By using three various process plasmas, different interfacial bonding characteristics towards the mineral coating were observed, by accomplishing mechanical pull-out tests on carbon-concrete-composite specimens. The CF treated with pure oxygen plasma showed the most improvement in the mechanical performance in this test, followed by the CF treated with oxygen/argon plasma. An increase in the pull-out energy measured up to the load maximum exhibited a rise of four times and three times the value of the references samples, respectively. In contrast, only a slight increase was detected for the CF treated with O 2 /CF 4 plasma.
The striking differences in the mechanical performance observed for treated and untreated carbon fibres were explained by means of morphological characterisation using ESEM/EDX and optical microscopy. A superior mechanical performance was observed for carbon roving with a uniform coating, which impregnated the interior of the roving and was well bonded to the carbon filaments. Such a coating was unambiguously mediated by the functional groups on the CF surface introduced by plasma treatment, which enabled chemical bonding between the carbon filaments and the mineral coating. In contrast, mostly bare and uncoated CFs were found in the reference samples, which were not plasma-treated. They showed no interaction with the mineral coating or with the hydration products of the fine-grained concrete matrix.
The samples treated with oxygen plasma exhibited an intense impregnation of the mineral binder grains of the coating suspension between the carbon filaments, induced by the strong affinity of the functionalised CF surface and the polar, water-based slurry. This resulted in a pronounced increase in the CF-mineral coating interface and in enhanced bonding between the components, i.e., CF-to-coating and coated roving-to-concrete matrix, thus enabling a persistent mechanical load transfer between carbon yarn. This explains the superior mechanical performance of the composite specimens containing plasma-treated carbon roving in comparison to the reference specimens containing untreated carbon roving.
Along with the plasma source, the duration of treatment was found to be crucial. With a longer treatment time, the mechanical performance in the pull-out experiments severely decreased. This can be traced back to advanced oxidation and, therewith, partial damage to the carbon fibre.
It can be concluded that plasma-driven functionalisation of sized-only carbon roving is a very promising method for promoting the fibre-matrix bond in carbon concrete composites, without introducing a petrochemical impregnation of the carbon strands. The ongoing work of the authors deals with the systematic optimisation of plasma treatment, developing new compositions for mineral coatings, its application technology, and testing the performance of such carbon concrete composites under elevated temperatures.