CNT-grafted glass fibers as a smart tool for epoxy cure monitoring, UV-sensing and thermal energy harvesting in model composites

Abstract

A ‘hierarchical’ reinforcement of glass fibers (GFs) chemically grafted with multiwall carbon nanotubes (MWCNTs) has been utilized for epoxy cure monitoring, UV-sensing, and thermal energy harvesting in model composites. MWCNTs were covalently attached to the surface of glass fiber yarns (GF-yarns) in a dip-coating deposition process. Hereafter, the hybrid yarns are denoted as GF-CNT. Scanning electron microscopy (SEM) demonstrated a highly uniform CNT-layer covering the fiber surfaces. In turn, GF-CNT reached a maximum conductivity of 2060 S m⁻¹, being of the same order of magnitude as the CNT-only bucky paper film. A GF-CNT in a uni-directional arrangement within a dog-bone shaped mould was employed for epoxy cure monitoring, recording the resistance changes during the curing process. In addition, three yarns connected in parallel highlighted the potential for detecting the resin position upon filling a mould. GF-CNT embedded in epoxy has been proposed also as an integrated non-invasive composite UV-sensor, allowing polymer matrix health monitoring. Besides, the semi-conductive nature of MWCNTs offered the opportunity of thermoelectric energy harvesting by the GF-CNT and its model composite when exposed to a temperature gradient. This work reports some new insights into and potential of fiber/CNT multi-scale reinforcements giving rise to multi-functional structural composites.

---

1. Introduction

Since their discovery, carbon nanotubes (CNTs) have been widely used in advanced applications due to their unique electrical, mechanical and thermal properties.^1–5 Recently, the deposition of CNTs onto the surface of fibrous micro-scaled reinforcements has attracted the scientific interest of different research groups working in the area of high-performance polymer composites.⁶ The methods of the CNT deposition include: (i) chemical vapor deposition (CVD), (ii) electrophoretic deposition, (iii) conventional dip-coating, and (iv) sizing mixtures containing CNTs directly applied to the fibers during the spinning process. Namely, glass,⁷ carbon,^8–10 ceramic¹¹ and natural fibers^12–14 have been used, while the resulting ‘hierarchical’¹⁵ or ‘fuzzy’^16,17 reinforcements were incorporated in different polymeric matrices.^18,19 A review could be found where CNT-based hierarchical composites and the advantages for the formation of a multiscale ‘hierarchical’ reinforcement were elaborately discussed.²⁰ The CNT coated filaments have been shown to increase significantly the composite interfacial adhesion strength,²¹ as well as the interlaminar shear strength of laminate composites.²² Besides, multi-functional properties are endowed to the final composites arising from the nanostructured CNT-rich interfacial regions. In relation to that, different kinds of functionalities have been reported such as strain sensing,²³ temperature sensing,²⁴ etc.

Glass fiber reinforced polymer composites (GFRPs) offer a flexible design approach for structural materials with significantly enhanced specific properties such as strength and stiffness.²⁵ Epoxy resins are one of the most important thermosetting polymers used as the matrix in composite materials. This is attributed to their ability to be infused in fabrics at room temperature,²⁶ as well as their high chemical and temperature resistance.²⁷ It is well-known that the matrix-dominated properties such as interlaminar shear strength,²⁸ resin failure strain,²⁹ void content,³⁰ crosslink density,³¹ etc. can play a crucial role on the composite's damage development and progression. Particularly, the factors which can influence the properties of the epoxy matrix include: (i) the chemical composition and stoichiometry of the reagents, i.e. relative epoxy/amine concentration; (ii) the cure schedule used to process the resin; and (iii) the detection of resin location upon filling a mould containing fibers; for instance prepregs. Therefore, there is a great demand for sensors which can determine in situ the state of cure of epoxies at points remote from the composite surface.³² In relation to that, the cure monitoring can rely on various physical or chemical properties that can be used to follow the transformation of an initially liquid thermoset resin into its final rigid solid form. Optical fibre sensors,²⁵ conductive aramid fibers,²⁶ CNT networks giving a thermoresistive response,³³ dielectric analysis,³² ultrasonic wave propagation,³⁴ heat-flux³⁵ and viscosity measurements³⁶ are some of the techniques already reported for the cure monitoring of epoxy resins. Conventional spectroscopic techniques such as near-infrared (NIR),³⁷ mid-infrared, fluorescence³⁸ and Raman³⁹ have been also employed for the in situ cure monitoring providing even a quantitative information about the relative concentrations of the resin's chemical constituents. Since the fiber/CNT hierarchical structures have shown promising reinforcement effects, it would be ideal if they can provide some information about the curing state of the resin in which they are incorporated.

Polymers are known to be sensitive in the UV-light exposition because it causes the shortening of polymeric chains by time, phenomenon called as ‘photodegradation’. The photodegradation has been explained by different kind of mechanisms for thermosets and thermoplastics,⁴⁰ and apparently it has a detrimental effect on the polymer's mechanical performance. For polymer composites where the matrix is the dominant component providing the structural integrity, it is essential that it is free of defects. Therefore, the existence of an integrated, non-invasive sensor within the composite structure which can give an output signal when exposed to UV-illumination would be of great importance. Then, by testing separately the resistance and stability of the polymer itself towards the UV-light, the sensor's data can be correlated and determine the composites' health condition. In such a way, it could be realised whether the polymer matrix has undergone a significant damage.

Composite materials are exposed in several cases to environments where there exists a temperature difference, i.e. composite parts of airplanes, automotive, etc. Therefore, their potential to function as thermoelectric materials is a very intriguing field of research.⁴¹ Thermoelectric materials are emerging candidates for thermal energy harvesting (such as waste heat), due to their ability to generate voltage upon exposure to a temperature gradient. This so-called thermoelectric or Seebeck effect is described by the thermoelectric power (TEP), or thermopower, or Seebeck coefficient (S), which is the direct solid state conversion of thermal energy to electricity.^42–44 The thermoelectric power is defined as:

where ΔV is the electric potential difference (or thermovoltage, or thermoelectric voltage) created by the temperature gradient, ΔT, within the material. The Seebeck coefficient is used further to calculate the power factor (PF); PF = σ × S², σ is the electric conductivity, as a means to compare the efficiency of different thermoelectric materials. The dimensionless figure of merit (ZT); ZT = (σ × S²/κ)T, κ is the thermal conductivity and T is the absolute temperature, is also used to compare the thermoelectric efficiency.⁴⁵ However, in this study due to the difficulties to determine the thermal conductivity of the composite interphase, alternatingly the values of power factor are presented, often used to compare materials to one other.⁴⁶ It should be mentioned that the Seebeck coefficient is positive for p-type semi-conductors and negative for n-type ones.⁴⁷ Moreover, it is an intrinsic property of the materials related to their electronic properties, and independent of their geometry.⁴⁸ Thus, it can be realised that for an optimum thermoelectric efficiency; high electrical conductivity and Seebeck coefficient, combined with low thermal conductivity are required. Traditional thermoelectric materials are fabricated from low band gap semiconductors like Bi₂Te₃, PbTe, etc., but are limited in use due to toxicity and viability in the production on a large scale.⁴⁹ Recently, CNT-based polymer nanocomposites have been utilized for thermal energy harvesting, and they are of particular interest due to their low thermal conductivity, high electrical conductivity, ease of production, relatively low cost, flexibility and high specific properties. The low thermal conductivity is a result of the phonon scattering at the CNT–polymer–CNT interphases, and thus the temperature difference can be sustained within the material. Highly loaded SWCNT nanocomposites based on conductive polymer matrices⁵⁰ have reached maximum power factors (σ × S²) being in the range of ∼140 μW mK⁻².⁵¹ The optimisation of power factor by increasing both the conductivity as well as the Seebeck coefficient remains still an open field for investigation and for that, different approaches have been adopted such as doping of CNTs with different molecules,⁵² creation of structural geometries like a p/n heterojunction,⁵³ use of different kind of conjugated polymers,⁵⁴ etc. Until now, the thermoelectric power of CNT filled polymer nanocomposites,^44,55–57 short carbon fiber/polycarbonate composites,⁵⁸ and long carbon fiber reinforced laminates^41,59 have been investigated. Indeed, thermoelectric structural materials are very promising for large-scale thermal energy harvesting, and several factors should be addressed in order to increase their thermoelectric efficiency rendering them an attractive technology for the composite market. To the best of our knowledge, hierarchical fiber/CNT reinforcements exploited for harvesting waste heat energy has not yet been reported.

Herein, a ‘hierarchical’ reinforcement structure based on glass fibers (GFs) chemically grafted with multiwall carbon nanotubes (MWCNTs) is proposed for the first time as a smart tool for the epoxy cure monitoring, UV-sensing and thermal energy harvesting of model epoxy composites. Initially, GF-yarns were endowed with amine surface functionalities, and then dipped in a solution of acyl chloride modified MWCNTs (MWCNT-COCl). Amide chemical bonds were formed as revealed by the XPS analysis of a previous study.⁷ Hereafter, the hybrid yarns are denoted as GF-CNT. Scanning electron microscopy (SEM) depicted the highly entangled and densely packed MWCNT networks homogeneously distributed onto the fiber surfaces. The interphase microstructure of GF-CNT/epoxy composite was studied by transmission electron microscopy (TEM). The electrical properties of GF-CNT were investigated by ‘two-probe’ electrical resistance measurements, and after calculations, the electrical conductivity was found to be 2060 S m⁻¹. Single fiber pull-out test analysis revealed an increase of ∼95% for the interfacial adhesion strength of GF-CNT/epoxy model composites compared to bare GFs/epoxy composites (ESI data†). A single GF-CNT in a uni-directional arrangement within a dog-bone shaped mould was employed for the in situ cure monitoring of a commercial epoxy resin. Recording the GF-CNT resistance changes by time was found to give a fingerprint of the epoxy degree of cure and was further correlated to the real state of cure studied by differential scanning calorimetry (DSC). The principle of our study can be simply implemented to large-scale composites, since the GF-CNT yarns can be incorporated in composites during manufacturing; i.e. unidirectional lay-ups, prepregs, and warp-knits, allowing the in situ cure monitoring. Furthermore, three GF-CNT yarns connected in parallel demonstrated the potential of GF-CNT to detect the resin position upon filling a mould making it a viable technology for resin transfer moulding fabrication processes. In addition, GF-CNT embedded in epoxy exhibited an excellent response to UV-light demonstrating its function as a composite integrated and non-invasive UV-sensor. This could be an important feature for the polymer matrix health monitoring. Finally, thermal energy harvesting was manifested by the GF-CNT and the GF-CNT/epoxy model composite, while more investigation is required to increase the thermoelectric efficiency. The utilization of GF-CNT multiscale reinforcements for the epoxy cure monitoring, UV-sensing and thermal energy harvesting is for the first time addressed in this work and it could be envisaged that GF-CNT can be implemented on a larger scale giving rise to multi-functional structural composites.

2. Experimental

2.1 Materials

E-Glass fiber yarns without sizing and with a fineness of 120 tex consisting of 204 filaments (aver. diameter 18 μm) were manufactured by a continuous spinning process at the Leibniz Institute of Polymer Research Dresden. Commercially available MWCNTs (Nanocyl, NC 7000) with a carbon purity >90%, average length 1.5 μm and diameters around 10 nm were received from Nanocyl S.A. (Sambreville, Belgium). The silane coupling agent, gamma-aminopropyltriethoxysilane (γ-APS, 98%) was supplied by ABCR (Karlsruhe, Germany) and used for the fiber surface modification. Extra dry tetrahydrofuran (THF) was purchased from Sigma-Aldrich (Steinheim, Germany). A commercial low viscosity diglycidyl ether of bisphenol-A (DGEBA) based epoxy resin (η ∼ 0.7 Pa s) with triethylenetetramine (TETA) hardener (Epofix, Struers) was used, known to cure under a ring-opening addition polymerisation reaction. According to the supplier specification, the epoxy fully cures at room temperature within 24 h at the stoichiometric ratio was 100 : 12 (w/w) epoxy resin : hardener. The chemical formulas of the two constituents are depicted in Fig. 1.

Fig. 1 Chemical structure of (a) Epofix Struers resin and (b) triethylenetetramine hardener.

2.2 Chemical grafting of MWCNTs onto the surface of GF-yarns

In order to carry out the chemical grafting reaction, MWCNTs were chemically modified to introduce carbonyl chloride groups (–COCl), and GF-yarns were silanised using gamma-aminopropyltriethoxysilane (γ-APS) enabling amine surface moieties. Details about the modification steps can be found in a previous study.⁷ MWCNT-COCl (10 mg) were dispersed in 100 ml of extra dry THF. Several yarns of γ-APS modified GFs (GF-APS) were attached lengthwise to a glass frame and kept under vacuum at 80 °C overnight. Once the MWCNT-COCl suspension was ready, a conventional dip-coating apparatus was used and the glass frame was immersed into the MWCNT solution for 30 min. This time was sufficient to complete the chemical grafting through nucleophilic substitution reaction between the fibers' amine groups and the MWCNT acyl chloride groups. Afterwards, the frame was removed and the fibers were kept for drying with their axes perpendicular at 80 °C for 24 h. It should be mentioned that all the reaction steps were carried out under argon (Ar) atmosphere. By measuring the difference of the fiber yarn weight before and after coating with CNTs, it has been found that the weight fraction of CNTs is ∼0.5 wt%. Experiments of dip coating the glass fiber yarns in CNT dispersions with higher concentrations (1 mg and 10 mg ml⁻¹) have shown that the CNT loading can be increased to 0.8 wt% for 1 mg ml⁻¹ and 1.2 wt% for 10 mg ml⁻¹. Fig. 2 illustrates the chemical procedure followed to attach MWCNTs onto the GF-yarns via covalent bonds.

Fig. 2 Schematic illustration of the steps followed to deposit MWCNTs onto the GF-yarns via covalent bonds. A single glass fiber represents the fiber yarn which was used throughout all the experiments.

2.3 Characterisation techniques

2.3.1 Atomic force microscopy (AFM), scanning and transmission electron microscopy (SEM, TEM). The surface topography of GF-CNT was observed by Tapping Mode Atomic Force Microscopy (TM-AFM). AFM images (height data) were recorded with a scanning probe microscope (Nanoscope III, Dimension 3100™, Veeco Digital Instruments) at a resonant frequency of approximately 300 kHz. Commercially available silicon cantilevers were used (TM Nanoprobe, Veeco) with a cantilever spring constant of 35 N m⁻¹, a tip cone angle of 20° and tip radius of about 8 nm while the scanning frequency was 1 Hz. The values of surface mean roughness (R_a) and maximum height roughness (R_max) were calculated after 2^nd flattening operation over the captured area of 2 × 2 μm. For the AFM investigation, single GF-CNT was stabilized onto clean silicon wafers. The Nanoscope analysis software (Bruker, ver. 1.40) was used for the image analysis. The AFM R_max value has been correlated to the thickness of the CNT-layer, considering that the 8 nm cantilever can scan through the CNT network with a pore-like morphology reaching the surface of the glass fiber substrate. This can be supported by the SEM analysis depicting more precisely the GF-CNT morphological characteristics.

A NEON 40 (Carl Zeiss AG, Germany) scanning electron microscope operating at 1.0 kV was used to study the surface morphology of GF-CNT. Individual filaments debundled from the GF-CNT were placed onto the surface of cleaned silicon wafers and attached from both end-sides with a copper adhesive tape. Prior to the SEM analysis, the specimens were sputter with a thin layer of platinum (∼3 nm) to avoid charging effects.

The Libra 200 transmission electron microscope (TEM, Carl Zeiss AG, Germany) operating at 200 kV was used to investigate the interphase microstructure of GF-CNT/epoxy composite. Bright field TEM images were recorded, while the energy filtering and contrast apertures were inserted to enhance the image quality. The interphase-section of GF-CNT/epoxy composite was prepared by Focused Ion Beam (FIB) using the NEON 40 FIB/SEM chamber equipped with a gallium ion beam operating at 30 kV. Details about the preparation are given in the ESI data.†

2.3.2 Electrical resistance and conductivity measurements. The electrical resistance, R, of GF-CNT was represented by measuring the resistance on a single fiber level. Single filaments were thoroughly detached from the GF-CNT, and resistance measurements were carried out at 20.0 mm fiber length by a standard two-probe method using a semiconductor characterization system (Keithley 2400 source-measure unit, Keithley Instruments GmbH, Germany). The distance between the two electrodes defines the fiber's length and used further for the resistivity (ρ) and conductivity (σ) calculations. Details for the measurement setup as well as the experimental parameters can be found elsewhere.⁷ The I–V characteristics of a single GF-CNT filament embedded in epoxy were also determined in order to evaluate the interfacial electrical properties. The sample preparation and all the measurement parameters were described in our previous work.⁷ Each measurement set-up is shown schematically in ESI.†

2.3.3 In situ cure monitoring of epoxy using the GF-CNT smart tool. In order to conduct the real-time epoxy cure monitoring measurements, a GF-CNT was placed lengthwise on a dog-bone shaped Teflon mould and two silver wires were contacted at the ends of the yarn using silver paste. The epoxy mixture was prepared within 1 min, and then was added thoroughly into the mould until it was entirely filled. Afterwards, the current measurement immediately started providing a constant voltage bias of 0.1 V every two seconds for a period of 24 h (required time for the epoxy to fully cure). The resistance values were derived by the Ohm's law and used for the cure monitoring evaluation. For the detection of the epoxy position upon filling a mould, three GF-CNT were connected in parallel and the current measurement started at constant voltage of 0.1 V. Then, the epoxy mixture was added within three time intervals, covering fully each of the three yarns. The experimental set-up used for the epoxy cure monitoring and the epoxy position detection is depicted in Fig. 3a. In order to correlate the GF-CNT resistance change with the state of cure of the epoxy resin, similarly to the ‘optical degree of cure’ defined in another work⁶⁰ in which an optical fiber sensor was proposed as the cure monitoring tool; here we define for the first time the ‘electrical degree of cure’ as:where R_t is the resistance of the GF-CNT at any time, t, after the start of cure, R₀ is the resistance 10 s after the epoxy was introduced in the mould covering fully the GF-CNT (this time is considered as the beginning of the cure reaction), and R_f is the resistance at the end of the cure.

Fig. 3 Schematic illustration of the set-up used for (a) the epoxy position detection (in a similar way the epoxy cure monitoring has been performed using a single GF-CNT), (b) UV-sensing and (c) thermoelectric power measurements. For all measurements a single GF-CNT yarn was used, however, for simplicity in the schematics a single fiber is drawn. The AFM height image shows the topographic characteristics; mean and maximum roughness (R_a, R_max) of a single fiber grafted with MWCNTs (Z data scale: 100 nm).

2.3.4 Thermal analysis of the resin cure by differential scanning calorimetry (DSC). Thermal characterisation was performed using a Q2000 dynamic scanning calorimeter (TA Instruments Inc., USA) in order to analyze the epoxy cure behavior and establish the degree of cure. Prior to the DSC experiments, epoxy was mixed with the hardener under magnetic stirring for 1 min to ensure a homogeneous mixture. The curing process was studied using isothermal scans at 23 °C for 24 h, performed on ∼10 mg hermetically sealed sample in an aluminum pan. The experiments were carried out under a constant flow of nitrogen at a rate of 50 ml min⁻¹. For isothermal cure measurements, the degree of cure can be estimated as the reaction progresses by monitoring the DSC heat flow during the curing reaction. From the DSC scan, the degree of cure (α) was calculated from the enthalpy per unit mass (ΔH_t) at any time during the isothermal cure, divided by the total heat of the reaction (H_T) obtained by integrating the area under the heat flow curve. The degree of cure (α) can range from 0 (completely uncured) to 1 (fully cured) and it is defined as follows:

In the present work, the degree of cure was plotted as a function of time for direct comparison with the resistance change (or electrical degree of cure) results.

2.3.5 Rheology. Rheological experiments were conducted by an AR2000 rheometer (TA Instruments Inc., Delaware), in a rate control mode, using a parallel plate geometry at a frequency of 1 Hz. Isothermal time sweeps were performed at 23 °C for 24 h using a gap of 1 mm and a strain of 1%. The sample was placed on the lower plate, and the upper plate was lowered to make contact with the resin. The experiment started after the system was allowed to come to thermal equilibrium (1–2 min).

2.3.6 UV-sensing. The response of a GF-CNT embedded in epoxy matrix (a model composite) to the incident UV-light was detected by recording the generated photocurrent. This property is attributed to the excellent behavior of CNTs to generate charge carriers upon photon excitation. In particular, a continuous constant voltage bias (V_bias) of 0.1 V with a step of 0.1 s was applied to the GF-CNT through the two wires, as shown in Fig. 3b. Then, the current under two different UV lamps (Deutsche Mechatronics, Inc., λ = 254 nm; typical peak UV-intensity: 640 μW cm⁻²@25 cm and λ = 365 nm, typical peak UV-intensity: 1100 μW cm⁻²@25 cm) was measured using as previously the Keithley 2400 source-measure unit. ON/OFF (UV-irradiation/UV-turned off) cycles at time intervals of 650 s with the sample enclosed in a dark experimental chamber were carried out. To evaluate the GF-CNT sensor characteristics, the I_UV/I_D is defined as the ratio of the UV illuminating current (I_UV) versus the dark current (I_D) measured at a constant voltage of 0.1 V, while the distance between the sample and the lamp was kept constant at around 25 cm.

2.3.7 Thermoelectric energy harvesting. The thermoelectric power measurements were carried out using an experimental set-up schematically shown in Fig. 3c. In order to measure the S (Seebeck coefficient, S = ΔV/ΔT), a single GF-CNT or GF-CNT/epoxy composite (similar to that used for the UV-sensing) was mounted on two copper blocks. One block was kept at room temperature (∼298 K), while the other was heated up in a controlled way by 10 K steps up to 373 K. This created the temperature difference between the ends of the investigated sample. The generated electric potential difference, or thermovoltage (ΔV), was measured across the electrodes, while the temperature of the two blocks was continuously measured with K-type thermocouples to determine the temperature gradient, ΔT, within the material. The Seebeck coefficient was derived from the slope of ΔV vs. ΔT curves by linear fitting, and the power factors have been calculated as PF = σ × S².

3. Results and discussion

3.1 Surface morphology of GF-CNT and interphase microstructure of single fiber composite

Fig. 4a shows a scanning electron microscopy (SEM) image demonstrating the surface morphology of a single GF-CNT filament. A continuous MWCNT coating can be observed with the MWCNTs arranged in densely packed and interconnected networks. This MWCNT coating with the excellent CNT–CNT junctions is responsible for the electrical conductivity introduced to the electrically insulating GF-yarns. Fig. 4b depicts the TEM micrograph illustrating the GF-CNT/epoxy interfacial microstructure providing detailed information about the MWCNT network characteristics along the interphase region. It can be easily realised that the MWCNT networks' microstructure is affected by the epoxy resin's polymerization and cross-linking process during hardening. More specifically, the thickness of the CNT-rich interphase is about 120 nm, which is almost three times higher than the thickness of the CNT coating before embedding in epoxy (calculated roughly as the maximum roughness; AFM height image in Fig. 3). This difference could be attributed to the interdiffusion of the epoxy macromolecular chains through the MWCNT networks. This interdiffusion mechanism can be more precisely explained considering that the epoxy monomer reacts chemically with the MWCNT carboxylic groups via a nucleophilic ring-opening addition mechanism. This facilitates the grafting of epoxy chains onto the MWCNT surfaces. At the same time, the pores of the MWCNT surface coating are big enough to allow the epoxy molecules to penetrate through. Therefore, by the time, due to the increase of the epoxy molecular weight or similarly the increase of the polymer chain lengths, the distance between individual nanotubes of the MWCNT network also increases. Accordingly, it can be speculated that the epoxy results in a kind of ‘swelling’ of the MWCNT coating; with the thickness to increase approximately three times. This finding is of major importance for the accurate calculation of the electrical conductivity of this CNT-rich interphase region used further for the thermoelectric power factor determination.

Fig. 4 (a) SEM image of single filament obtained from the GF-CNT shown at two different magnifications, and (b) TEM interphase microstructure of a single fiber/epoxy composite.

3.2 Epoxy cure monitoring and position detection using the GF-CNT sensor

Fig. 5a depicts the electrical degree of cure, a_R, (black line) as a function of time within the 24 h curing cycle. At the same figure, the DSC degree of cure (blue line) is plotted for direct comparison. As it was expected, the conversion increases with increasing time, and 24 h were found to be sufficient for the fully-cure of the particular epoxy resin (according also to the manufacturers' data sheet). Fig. 5b demonstrates the epoxy rheological properties showing that about 5 h are required for the epoxy to reach the gel point from its initial liquid state. The time between the onset of the rapid increase of the dynamic viscosity (∼3000 s = 50 min), and that at which it exceeded 100 kPa s was taken as the gelation period, similarly to other studies.^53,54 At the end of gelation period, the gel point is reached considered as the point in which storage and loss modulus cross.⁵⁵ At that time, the onset of vitrification occurs and the curing reaction continues until the end of vitrification, which is considered practically as the end of the epoxy curing reaction. The DSC and rheological investigations were performed in order to have a full image about the curing reaction and its completion, as well as the epoxy molecular mobility during hardening.

Fig. 5 (a) Electrical and DSC degree of cure at an isothermal cure cycle within 24 h, (b) epoxy rheological properties within 24 h curing cycle, and (c) illustration of the fast resistance change of the GF-CNT sensors upon filling a mould with epoxy resin.

Regarding the electrical degree of cure, it can be observed that it increases with time until it reaches a certain maximum. Then, it slightly decreases until a final plateau is reached. This can be explained by two factors. Predominant one is that the epoxy interdiffuses/penetrates through the fiber surface MWCNT networks resulting in a kind of “swelling” of the CNT layer increasing thus the distance between adjacent MWCNTs and hampering the direct physical contact of CNT–CNT junctions. Therefore, this can affect the electron transport and mobility, which follows a tunneling or a hoping mechanism. The other, which is secondary factor for the electrical resistance change, is the possible alternation of the MWCNT electronic properties due to the absorption of the epoxy molecules, which chemically react with the MWCNT functional groups. Taking into account the epoxy's reaction progression, as well as its molecular mobility revealed by rheological analysis, we can distinguish three areas of the electrical degree of cure curve. In the first one, the epoxy is in a liquid state and penetrates rapidly and easily through the MWCNT network (“swelling”); thus the resistance changes with a high rate. The rate of resistance change starts to decrease at a time around 3000 s, which is consistent with the rapid increase of the viscosity from the rheological investigation (onset of gelation). Further, the resistance changes with a lower rate and the epoxy comes into the rubbery state until it reaches a maximum at around 18 000 s (∼5 h), which represents the gel point according to the rheological analysis. At this point, the molecular mobility of the epoxy starts to become very slow, thereby the interdiffusion through the MWCNT networks becomes practically zero without increasing any more the CNT–CNT distances. Comparing this point with the DSC degree of cure, it can be deduced that at the maximum of the electrical degree of cure, epoxy has been 62% cured, and this maximum corresponds to the gelation point. Moreover, it should be mentioned that the slight decrease of the electrical degree of cure from this maximum until the end of cure cycle is attributed to the chemical interaction of the epoxy or amine hardener with the MWCNT carboxyl groups. This can act as a molecular doping decreasing slightly the MWCNT resistance.

In terms of practical application utilising GF-CNT for the epoxy cure monitoring, it is not very desired the fact that resistance changes at a high percent at the initial minutes due to “swelling” of the CNT layer by the epoxy. However, in the rubbery state (∼80% of initial resistance change) and onwards until the 24 h end of cure of the particular epoxy, the resistance changes are very characteristic for the curing phenomena during hardening explained also in detail and correlated to the rheological and DSC results.

Fig. 5c demonstrates that three GF-CNT connected in parallel can very effectively give a fast response of their resistance change when epoxy comes in contact. The resistance of GF-CNT shows immediately an increase, and this is explained similarly due the reasons which were mentioned above. The response of the GF-CNT sensor at each time that epoxy comes in contact can be utilized to detect the location of epoxy resin upon filling a mould in large-scale composite manufacturing processes.

3.3 UV-sensing properties

Fig. 6 depicts the ratio of I_UV/I_D (I_UV: UV-illumination current/I_D: dark current) during ON and OFF states exposing the sample to UV-illumination (λ ≈ 254 nm with 640 μW cm⁻² and λ ≈ 365 with 1100 μW cm⁻²). A significant increase in the forward current occurred upon UV-irradiation at both UV-light energies used (640 μW cm⁻² and 1100 μW cm⁻²), which showed a complete elastic and recoverable behavior. The fast response of the GF-CNT can be attributed to the highly entangled and concentrated MWCNT networks located at the interphase region, allowing high carrier mobility and enabling thus the fast transport of the photo-generated carriers. From the test experiments, it was observed that the measured light-induced current increased linearly as the UV-power increased, which is consistent with the results found in the literature.^57,61 The responsivity of the GF-CNT/epoxy device with an active area of approximately 20 × 0.15 mm² (20 mm is the GF-CNT length and 0.15 mm is the yarn diameter knowing that it has a fineness of 120 tex) was found to be 3.61 × 10⁻³ A W⁻¹ at a forward bias of 0.1 V.

Fig. 6 The ratio of photocurrent divided by the dark current at a constant bias of 0.1 V versus time. The time intervals between On/Off state of the UV-irradiation are 650 s.

Since the MWCNTs used for coating the GF-yarns contained –COCl functionalities (considered to have been transformed to –COOH groups by possible hydrolysis after the grafting reaction), the MWCNTs are p-doped because carboxyl groups facilitate the e⁻ withdrawing from the MWCNT-backbone, creating holes as the main charge carriers. Therefore, the observed response of the GF-CNT sensor was due to generation of UV-excited hole charge carriers and slightly due to the rise in temperature.⁵⁶ In addition, the current changes very slightly under visible light illumination and considering its high response to UV-illumination, the hierarchical GF-CNT reinforcement can act as an excellent integrated non-invasive composite UV-sensor. The UV-sensing ability could be utilized to monitor the health condition of the polymer matrix, by counting the time that a composite material has been exposed to UV-light, and correlating the effect of this exposition time on the degradation of the polymeric chains.

The responsivity of the GF-CNT/epoxy device in our study is much lower than values that have been reported in literature for UV-sensors (0.3 A W⁻¹ for p-type polymer/n-type ZnO structure, and 2.4 A W⁻¹ for graphene oxide:CNT/ZnO⁶¹), however, the GF-CNT can act as a non-invasive composite UV-sensor. Further investigation can be performed to grow for instance ZnO nanoparticles onto the CNT network and increase responsivity, however, micromechanical tests are required to prove that such nanostructured coating onto the surface of reinforcing fiber will not destroy and hamper the interfacial strength of the hierarchical composite.

3.4 Electrical resistance and conductivity of GF-CNT and GF-CNT/composite

Table 1 summarizes the values of the electrical resistance (R) and conductivity (σ) of GF-CNT single filaments. The average resistance was 3.85 ± 0.76 Mohm and its variation was relatively low as a result of the homogeneous MWCNT-coating characteristics, observed previously by the SEM analysis. The specific conductivity of a GF-CNT single filament with an average diameter of d ≈ 18 μm was calculated by the formula: σ_GF-CNT = (4L)/πd²R, and it was found 20.40 ± 4.02 S m⁻¹. The conductivity introduced to the intrinsic electrically insulating glass fibers arises from the continuous ultrathin MWCNT networks deposited onto the fiber surfaces, with the extensive CNT–CNT junctions creating the electron transport pathways. Therefore, it can be more precisely explained if we consider the existence of a continuous MWCNT thin layer with thickness, t, on the fiber surface, with electrical conductivity, σ_CNT. The conductivity of the CNT layer (σ_CNT) is given as proposed also in other studies by the following equation:^33,58

Table 1 DC electrical resistance (R) and conductivity (σ) of single fibers chemically grafted with MWCNTs and their corresponding single fiber composite

Material under investigation	Electrical resistance, R (Mohm)	Conductivity, σ (S m^−1)	Conductivity due to the CNT deposited networks, σCNT (S m^−1)
a The conductivity of the CNT layer was very close to the value of the CNT-only bucky paper film, which was found to be 2.3 × 10^3 S m^−1 (ESI data).b This conductivity value corresponds to the conductivity of the CNT-rich nanostructured interphase (σinterphase).
GF-CNT	3.85 ± 0.76	20.40 ± 4.02	2.06 × 10^3^a
GF-CNT/epoxy single fiber composite	7.52	10.46	3.92 × 10^2^b

---

Taking this formula as a rough estimate, the fiber average diameter d ≈ 18 μm, the thickness of the CNT layer 44.6 nm (according to the AFM maximum roughness), and σ_GF-CNT the conductivity of MWCNT-coated GFs, the conductivity of the CNT layer is calculated to be 2.06 × 10³ S m⁻¹. The thin CNT layer possesses high electrical conductivity, which is very close to the aggregated and highly dense CNT-only bucky paper film (see details in ESI†). The maximum conductivity of the GF-CNT single filaments is to our knowledge the highest value compared to existing ones reported in literature.³³

Table 1 shows also the resistance of the single fiber composite, which was found to be 7.52 Mohm. The specific conductivity of single GF-CNT filament in epoxy at the longitudinal direction is given by the formula: σ_GF-CNT/epoxy = (4L)/πd²R, and after calculations was found to be 10.46 S m⁻¹; approximately two times lower than the conductivity of the respective fibers. This can be explained by the presence of the epoxy insulating chains through the CNT networks generating a barrier at the CNT junctions preventing direct physical contact between them.⁵⁷ In the same way like previously, if we consider the thickness, t* of the MWCNT networks after embedding in epoxy as revealed by the TEM interphase-section image (t* ≈ 120 nm), using eqn (4), we can correlate the conductivity of the nanostructured interphases (σ_interphase) with the respective specific conductivity of single GF-CNT filament in epoxy following the expression: σ_interphase ≈ (d/4t*) σ_GF-CNT/epoxy. After calculations, the interphase conductivity of single GF-CNT filament/epoxy composite was 3.92 × 10² S m⁻¹. The high conductivity of the interphase can be compared to polymer/MWCNT composites with high MWCNT loadings.³³ This is of utmost importance for the utilization of the GF-CNT hierarchical reinforcement as thermoelectric elements in composite materials.

3.5 Thermoelectric power investigations

Fig. 7 demonstrates the generated thermovoltage upon exposition to a temperature gradient for the different samples tested. It could be observed that within the temperature range, the thermovoltage increased linearly with the increased temperature as revealed by the perfectly matching linear fitting. The slope of Seebeck voltage versus temperature difference gives the Seebeck coefficient, and it was found to be 12.26 μV K⁻¹ for the GF-CNT, 14.05 μV K⁻¹ for the GF-CNT/epoxy composite and 16.1 μV K⁻¹ for the MWCNT bucky paper film. It can be observed that the bucky paper exhibits the highest Seebeck coefficient. The slightly higher Seebeck coefficient of the CNT bucky paper film compared to the GF-CNT and GF-CNT/epoxy composite is explained due to the fact that bucky paper is a compact bulk film of interconnected CNTs with a thickness of ∼100 μm (ESI†). Therefore, it can more efficiently transport the thermoelectrically induced charge carriers generating a slightly higher Seebeck coefficient. Accordingly, the power factors revealing the thermoelectric performance have shown the best value also for the MWCNT film (0.649 μW mK⁻²), while the GF-CNT and GF-CNT/epoxy showed 0.31 μW mK⁻² and 0.08 μW mK⁻², respectively (Fig. 7b). The GF-CNT depicts higher Seebeck coefficient value after being incorporated in the epoxy matrix due to the phonon scattering, which facilitates the preservation of the temperature gradient within the material. Moreover, due to the high electrical conductivity of the highly loaded CNT interphase region, a high power factor can be achieved reaching the value of the CNT film. In this work, the extremely high conductivity of the GF-CNT delivered to the resulting composite interphases, combined with the high Seebeck coefficient, result in very promising power factor values higher than highly filled MWCNT composites prepared via solution blending techniques.⁴⁷ The use of GF-CNT with p- and n-type semi-conductive characteristics connected in series, as well as decoration of highly efficient thermoelectric nanoparticles onto the CNT networks covering the fiber surfaces, is still subject to ongoing research of our group. It is expected that the results from the future experimental investigations will lead to new insights and improvements in thermoelectric performance of the composite interphases.

Fig. 7 (a) Seebeck voltage versus deltaT with the corresponding linear regression fits giving the values of Seebeck coefficients (equal to the slope of the corresponding curve), and (b) electrical conductivity, Seebeck coefficients and power factors calculated for the different samples.

4. Conclusions

This paper reports the development of glass fibres fully coated with MWCNTs (GF-CNT) by covalent bonds following a solution based dip-coating deposition process. The MWCNT coating was found to be sensitive in the epoxy molecular mobility during the hardening process, and therefore GF-CNT can serve as a non-invasive composite sensor for the real time epoxy cure monitoring. The findings from this study could be of interest for industrial applications, making it feasible to monitor the composite quality as for e.g. unidirectional or textile fabric composites. The GF-CNT embedded in epoxy exhibited also a high sensitivity to UV-light offering the possibility to be used as an integrated composite UV-sensor. Finally, another major finding in this study is the utilization of the GF-CNT hierarchical reinforcement for the thermal energy harvesting. The wt% of CNT loadings on the glass fiber yarns has been found to slightly affect the performance of the GF-CNT for the epoxy cure monitoring and UV-sensing function. However, the increased CNT-loading very positively affected the thermoelectric performance of the hybrid GF-CNT reinforcements and their model composites, which is a study under investigation for future work. Overall, it can be anticipated that the all the results presented in this study could be implemented to real composites in an industrial scale.

Acknowledgements

This work was partly funded by the ‘POCO’ European Project (CP-IP213939-1, http://www.poco-project.org) under the 7^th Framework Program (FP7) and call NMP-2007-LARGE-1. The authors would like to thank Mrs L. Häusler and Mr M. Göbel for help with DSC measurements and preparation of the FIB interphase-sections, respectively. The electronic supporting information (ESI†) is available online.

Notes and references

1. S. Iijima, Nature, 1991, 354, 56–58.
2. Z. Spitalsky, D. Tasis, K. Papagelis and C. Galiotis, Prog. Polym. Sci., 2010, 35, 357–401.
3. D. G. Papageorgiou, L. Tzounis, G. Z. Papageorgiou, D. N. Bikiaris and K. Chrissafis, Polymer, 2014, 55, 3758–3769.
4. F. Wode, L. Tzounis, M. Kirsten, M. Constantinou, P. Georgopanos, S. Rangou, N. E. Zafeiropoulos, A. Avgeropoulos and M. Stamm, Polymer, 2012, 53, 4438–4447.
5. M. Liebscher, L. Tzounis, P. Pötschke and G. Heinrich, Polymer, 2013, 54, 6801–6808.
6. T.-W. Chou, L. Gao, E. T. Thostenson, Z. Zhang and J.-H. Byun, Compos. Sci. Technol., 2010, 70, 1–19.
7. L. Tzounis, M. Kirsten, F. Simon, E. Mäder and M. Stamm, Carbon, 2014, 73, 310–324.
8. R. J. Sager, P. J. Klein, D. C. Lagoudas, Q. Zhang, J. Liu, L. Dai and J. W. Baur, Compos. Sci. Technol., 2009, 69, 898–904.
9. L. Ma, L. Meng, G. Wu, Y. Wang, M. Zhao, C. Zhang and Y. Huang, Compos. Sci. Technol., 2015, 114, 64–71.
10. O. Boura, E. K. Diamanti, S. A. Grammatikos, D. Gournis and A. S. Paipetis, Surf. Interface Anal., 2013, 45, 1372–1381.
11. N. Yamamoto, A. John Hart, E. J. Garcia, S. S. Wicks, H. M. Duong, A. H. Slocum and B. L. Wardle, Carbon, 2009, 47, 551–560.
12. L. Tzounis, S. Debnath, S. Rooj, D. Fischer, E. Mäder, A. Das, M. Stamm and G. Heinrich, Mater. Des., 2014, 58, 1–11.
13. N. Yamamoto, A. J. Hart, E. J. Garcia, S. S. Wicks, H. M. Duong, A. H. Slocum and B. L. Wardle, Carbon, 2009, 47, 551–560.
14. W. Zhang, Y. Y. Tan, C. Wu and S. R. P. Silva, Particuology, 2012, 10, 517–521.
15. H. Qian, A. Bismarck, E. S. Greenhalgh, G. Kalinka and M. S. P. Shaffer, Chem. Mater., 2008, 20, 1862–1869.
16. E. J. Garcia, B. L. Wardle, A. John Hart and N. Yamamoto, Compos. Sci. Technol., 2008, 68, 2034–2041.
17. T. R. Pozegic, I. Hamerton, J. V. Anguita, W. Tang, P. Ballocchi, P. Jenkins and S. R. P. Silva, Carbon, 2014, 74, 319–328.
18. E. Bekyarova, E. T. Thostenson, A. Yu, H. Kim, J. Gao, J. Tang, H. T. Hahn, T. W. Chou, M. E. Itkis and R. C. Haddon, Langmuir, 2007, 23, 3970–3974.
19. J. Karger-Kocsis, H. Mahmood and A. Pegoretti, Prog. Mater. Sci., 2015, 73, 1–43.
20. H. Qian, E. S. Greenhalgh, M. S. P. Shaffer and A. Bismarck, J. Mater. Chem., 2010, 20, 4751–4762.
21. H. Qian, A. Bismarck, E. S. Greenhalgh and M. S. P. Shaffer, Compos. Sci. Technol., 2010, 70, 393–399.
22. Q. An, A. N. Rider and E. T. Thostenson, ACS Appl. Mater. Interfaces, 2013, 5, 2022–2032.
23. L. Gao, E. T. Thostenson, Z. Zhang and T.-W. Chou, Adv. Funct. Mater., 2009, 19, 123–130.
24. J. Zhang, J. Liu, R. Zhuang, E. Mäder, G. Heinrich and S. Gao, Adv. Mater., 2011, 23, 3392–3397.
25. P. A. Crosby, G. R. Powell, G. F. Fernando, C. M. France, R. C. Spooncer and D. N. Waters, Smart Mater. Struct., 1996, 5, 415.
26. G. Pandey, H. Deffor, E. T. Thostenson and D. Heider, Composites, Part A, 2013, 47, 102–108.
27. S. K. Ooi, W. D. Cook, G. P. Simon and C. H. Such, Polymer, 2000, 41, 3639–3649.
28. K. J. Bowles and S. Frimpong, J. Compos. Mater., 1992, 26, 1487–1509.
29. R. J. Morgan, in Epoxy Resins and Composites I, Springer, Berlin Heidelberg, 1985, pp. 1–43.
30. S. F. M. de Almeida and Z. d. S. N. Neto, Compos. Struct., 1994, 28, 139–148.
31. C. H. Lau, K. A. Hodd and W. W. Wright, Br. Polym. J., 1986, 18, 316–322.
32. R. Hardis, J. L. P. Jessop, F. E. Peters and M. R. Kessler, Compos. Struct., 2013, 49, 100–108.
33. K. L. Lasater and E. T. Thostenson, Polymer, 2012, 53, 5367–5374.
34. A. Davis, M. M. Ohn, K. Liu and R. M. Measures, Composite cure monitoring with embedded optical fiber sensors, Proc. SPIE 1489, Structures Sensing and Control, 33, 1991, pp. 264–274,  DOI:10.1117/12.46585.
35. M. J. Perry, L. J. Lee and C. W. Lee, J. Compos. Mater., 1992, 26, 274–292.
36. J.-S. Kim and D. G. Lee, J. Mater. Process. Technol., 1993, 37, 405–416.
37. N. A. St John and G. A. George, Polymer, 1992, 33, 2679–2688.
38. E. Pyun and C. S. P. Sung, Macromolecules, 1991, 24, 855–861.
39. L. Merad, M. Cochez, S. Margueron, F. Jauchem, M. Ferriol, B. Benyoucef and P. Bourson, Polym. Test., 2009, 28, 42–45.
40. R. S. C. Woo, Y. Chen, H. Zhu, J. Li, J.-K. Kim and C. K. Y. Leung, Compos. Sci. Technol., 2007, 67, 3448–3456.
41. S. Han and D. D. L. Chung, Carbon, 2013, 52, 30–39.
42. C. J. Vineis, A. Shakouri, A. Majumdar and M. G. Kanatzidis, Adv. Mater., 2010, 22, 3970–3980.
43. M. Liebscher, T. Gärtner, L. Tzounis, M. Mičušík, P. Pötschke, M. Stamm, G. Heinrich and B. Voit, Compos. Sci. Technol., 2014, 101, 133–138.
44. L. Tzounis, T. Gärtner, M. Liebscher, P. Pötschke, M. Stamm, B. Voit and G. Heinrich, Polymer, 2014, 55, 5381–5388.
45. A. I. Boukai, Y. Bunimovich, J. Tahir-Kheli, J.-K. Yu, W. A. Goddard Iii and J. R. Heath, Nature, 2008, 451, 168–171.
46. T. M. Tritt and M. A. Subramanian, MRS Bull., 2006, 31, 188–198.
47. M. Baxendale, K. G. Lim and G. A. J. Amaratunga, Phys. Rev. B: Condens. Matter Mater. Phys., 2000, 61, 12705–12708.
48. M. T. Terry, in Thermoelectrics Handbook, CRC Press, 2005, pp. 23–21-23-20.
49. Y. Choi, Y. Kim, S.-G. Park, Y.-G. Kim, B. J. Sung, S.-Y. Jang and W. Kim, Org. Electron., 2011, 12, 2120–2125.
50. C. Yu, Y. S. Kim, D. Kim and J. C. Grunlan, Nano Lett., 2008, 8, 4428–4432.
51. G. P. Moriarty, S. De, P. J. King, U. Khan, M. Via, J. A. King, J. N. Coleman and J. C. Grunlan, J. Polym. Sci., Part B: Polym. Phys., 2013, 51, 119–123.
52. C. Yu, A. Murali, K. Choi and Y. Ryu, Energy Environ. Sci., 2012, 5, 9481–9486.
53. C.-K. Mai, B. Russ, S. L. Fronk, N. Hu, M. B. Chan-Park, J. J. Urban, R. A. Segalman, M. L. Chabinyc and G. C. Bazan, Energy Environ. Sci., 2015, 8, 2341–2346.
54. O. Bubnova, Z. U. Khan, A. Malti, S. Braun, M. Fahlman, M. Berggren and X. Crispin, Nat. Mater., 2011, 10, 429–433.
55. M. Piao, J. Na, J. Choi, J. Kim, G. P. Kennedy, G. Kim, S. Roth and U. Dettlaff-Weglikowska, Carbon, 2013, 62, 430–437.
56. M. Liebscher, T. Gärtner, L. Tzounis, M. Mičušík, P. Pötschke, M. Stamm, G. Heinrich and B. Voit, Compos. Sci. Technol., 2014, 101, 133–138.
57. L. Tzounis, M. Liebscher, E. Mäder, P. Pötschke, M. Stamm and S. Logothetidis, in International Conferences and Exhibition on Nanotechnologies and Organic Electronics (Nanotechnology 2014): Proceedings of NN14 and ISFOE14, AIP Publishing, 2015, pp. 138–148.
58. S. Saqán, A. M. Zihlif, S. R. Al-Ani and G. Ragosta, J. Mater. Sci.: Mater. Electron., 2008, 19, 1079–1085.
59. V. H. Guerrero and D. D. L. Chung, Compos. Interfaces, 2002, 9, 395–401.
60. A. Dimopoulos, S. J. Buggy, A. A. Skordos, S. W. James, R. P. Tatam and I. K. Partridge, J. Appl. Polym. Sci., 2009, 113, 730–735.
61. T. V. Cuong, H. N. Tien, V. H. Luan, V. H. Pham, J. S. Chung, D. H. Yoo, S. H. Hahn, K.-K. Koo, P. A. Kohl, S. H. Hur and E. J. Kim, Phys. Status Solidi A, 2011, 208, 943–946.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra09800b

---