Epoxy / Glass Fiber Nanostructured p- and n-Type Thermoelectric Enabled Model Composite Interphases

Featured Application: This work proves the thermoelectric principles and e ﬃ ciency of model composites towards the manufacturing of multifunctional ﬁber reinforced polymer composites (FRPCs) for large-scale structural applications enabling thermal energy harvesting technology. Abstract: This experimental study is associated with the modiﬁcation of glass ﬁbers with e ﬃ cient, organic, functional, thermoelectrically enabled coatings. The thermoelectric (TE) behavior of the coated glass ﬁber tows with either inherent p semiconductor type single wall carbon nanotubes (SWCNTs) or the n-type molecular doped SWCNTs were examined within epoxy resin matrix in detail. The corresponding morphological, thermogravimetric, spectroscopic, and thermoelectric measurements were assessed in order to characterize the produced functional interphases. For the p-type model composites, the Seebeck coe ﬃ cient was + 16.2 µ V / K which corresponds to a power factor of 0.02 µ W / m · K 2 and for the n-type − 28.4 µ V / K which corresponds to power factor of 0.12 µ W / m · K 2 . The p–n junction between the model composites allowed for the fabrication of a single pair thermoelectric element generator (TEG) demonstrator. Furthermore, the stress transfer at the interphase of the coated glass ﬁbers was studied by tow pull-out tests. The reference glass ﬁber tows presented the highest interfacial shear stress (IFSS) of 42.8 MPa in comparison to the p-and n-type SWCNT coated GF model composites that exhibited reduced IFSS values by 10.1% and 28.1%, respectively. evaluate interphases between modiﬁed GFs with p and n-type SWCNTs model epoxy composites. ﬀ orts the morphological inspection well as the evaluation of the structural changes that might be caused by the functional nanocoated GFs in the presence of temperature gradients. The thermal stability of the nanocoated GF tows was investigated through thermogravimetric analysis (TGA). Additionally, the chemical and structural stability of the TE nanoparticle-coated GF tows were studied by Raman measurements at elevated temperatures. The Raman spectra of p-type SWCNTs coated GFs and n-type SWCNTs coated GFs were acquired at ambient conditions, as well as at temperature di ﬀ erences of 50 ◦ C and 100 ◦ C by performing measurements at 75 ◦ C and 125 ◦ C. The results indicated that both the p-type and n-type TE enabled GF tows are chemically and structurally stable. The e ﬀ ect of the integration of the modiﬁed nanocoated GFs into a conventional epoxy resin was evaluated. The TE performance of the functional GF tows was only slightly a ﬀ ected by the resin encapsulation. Both p and n-type SWCNTs model composites exhibited a combination of high electrical conductivity and Seebeck coe ﬃ cient. N-type model composites exhibited higher electrical conductivity compared to the corresponding p-type, due to their higher intrinsic electrical properties, as measured on SWCNTs bucky paper ﬁlms. Moreover, the n-type model composites exhibited excellent behavior as thermoelectric materials with remarkably high values of Seebeck coe ﬃ cient very close to the properties of the bulk SWCNTs material. An epoxy-based TEG model composite was manufactured to demonstrate the potential application of the technology in a TE generator, using a pair of alternating p- and n-type junctions interconnected in series via a metallic junction. The interfacial shear strength for both p and n-type TE enabled model composites by tow pull-out tests. of SDBS ected


Introduction
During the last decades, the growing demand for the utilization of lightweight, high-performance structural composite materials has rapidly increased in various commercial sectors, such as the aerospace and automotive industries. Fiber reinforced polymers constitute an extremely interesting field of research due to their potential for the exploitation of additional functionalities. Tailoring or "architecting" the material in relation to both its structural and non-structural services is an inherent ability of advanced composites and is rapidly emerging technology for their next generation. Understanding their mechanical behavior requires the study of the interface, the common physicochemical areas between the distinct phases (reinforcing and matrix materials) [1,2]. This is a prerequisite for the design of a composite material, as its behavior is most of the times governed by its interfacial properties [3]. be reclaimed as electricity on account of the well-known 'Seebeck effect'. Thermoelectric materials exhibit the Seebeck effect, in which charge carriers move in response to temperature gradients. The Seebeck coefficient is directly related to the density of states, whereas the electrical conductivity can be limited by electronic and morphological defects. Charge carriers diffuse across the temperature gradient, creating an accumulation of charge and thus a potential difference, ∆V. The magnitude of this effect is described by the Seebeck coefficient (S) and is simply the voltage difference divided by the applied thermal gradient S = ∆V/∆T. The Seebeck coefficient is further used for the calculation of the power factor (PF = σ × S 2 ), a well-established quantity for comparing the TE performance of different thermoelectric materials. Finally, the overall thermoelectric efficiency for specific temperature range applications of a material is classified by the dimensionless thermoelectric figure of merit, ZT (ZT = (σ × S 2 ) × T/κ), where σ is the electrical conductivity, S is the Seebeck coefficient, κ is the thermal conductivity and T is the corresponding temperature.
An efficient thermoelectric material presents a high value of ZT, combining high electrical and poor thermal conductor characteristics. The three thermoelectric parameters σ, S and κ are correlated with the intrinsic carrier concentration of the material [26][27][28][29]. Thermoelectric energy harvesting is a conventional technique for converting wasted thermal energy into the electrical power by means of thermoelectric element generators (TEGs). A typical TEG device consists of p-n junctions properly interconnected electrically in series and thermally in parallel. When a temperature gradient is established between the hot and the cold side, free electrons and holes are able to move towards specific directions and generate a potential difference proportional to the temperature difference between the p-type and n-type semiconductors. Compared to the traditional TEG devices, high performance flexible TEGs that exploit the nanotechnology advancements are being extensively reported [17,30].
In this study, hierarchical glass fiber tows coated with p-type and n-type SWCNTs as well as reference tows were evaluated in terms of their thermal stability and physicochemical relevance with the epoxy resin. The thermoelectric response of the produced p and n-type epoxy model composites have been assessed for a single pair TEG fabrication. The tow pull-out test was used to determine the quality of interfacial bonding between the epoxy matrix and the nanocoated glass fibers. Fracture morphologies of the pulled-out tows obtained via optical microscopy revealed a cohesive type of failure in all cases, as the epoxy material remained attached on the surface of the tows after the pull-out process.

Materials
The 1062 Multi-End Roving (NEG) GFs were employed for this study with a tensile strength of 2300-2400 MPa and tensile modulus of 72 GPa, specifically designed for use with epoxy resin systems in applications that require high mechanical strength and corrosion resistance such as filament wound pipes. The filament diameter is 13.3 ± 0.6 µm as stated by the manufacturer.
RS Pro silver loaded epoxy conductive paste was utilized for the adhesive ohmic contacts. A commercial low viscosity liquid epoxy resin system was employed, consisting of EPIKOTE Resin 828LVEL (100.0 parts by weight) EPIKURE™, Curing Agent 866 (83.0 parts by weight), and EPIKURE™ Catalyst 101 (1.5 parts by weight). The curing cycle and post-curing included heating at 90 • C for 90 min and at 130 • C for 180 min, respectively.

Dispersion and Deposition of the TE p and n-Type SWCNTs Inks onto GF Tows
The SWCNTs were dispersed using the UP400S manufactured by Hielscher Ultrasonics tip-ultrasonic probe processor for 45 min at 10 a Watt amplitude to deliver the wave directly into the aqueous solvent. For the p-type TE ink, 2 mg/mL SWCNTs and 5 mg/mL of SDBS were utilized in order to produce a suspension with a proper critical micelle concentration (CMC) and achieve a stable dispersion. Accordingly, for the n-type TE ink, 2 mg/mL SWCNTs and 2 mg/mL of SDBS coupled with 5 mg/mL of poly(ethyleneimine) were used. The notation "p-type and n-type" refers to the type of the charge carriers induced in the two different formulations of the CNTs and does not imply any type of substitutional doping of the CNTs.
Then, the p and n-type TE inks were used for coating of the GF tows via a simple blade coating process. In more detail, 10 cm long single GF tows were cut from the reel, stabilized, and coated towards the creation of homogeneous continuous TE efficient p-type and n-type films onto the GFs. After the coating process, the tows were left to dry on a hot plate at 90 • C for 30 min.

Manufacturing of Model Composites for Thermoelectrical Characterization
Model composites were prepared and characterized for the study of the effect of the epoxy resin system addition on the thermoelectric response of the coated glass fibers. The coated single GF tows were placed flat onto a pre-cleaned silicon dog-bone shaped mould and secured in place on both sides with an adhesive tape. Then, two thin copper foil electrodes were adjusted tightly on each side of the functional single tows and glued with silver paste on the tow to ensure proper ohmic contacts. Upon drying of the contacts, a well-mixed and degassed epoxy resin/hardener/catalyst mixture was casted in the mould, which impregnated thoroughly the whole GF tow. The system was cured and post-cured following the aforementioned protocol. At least five model composites for each category were fabricated and the mean experimental values with the corresponding standard deviations are presented. The electrical properties and the thermoelectric response of the model composites were measured with the set up that is displayed in Figure 1.
Appl. Sci. 2020, 10, x 4 of 17 of the charge carriers induced in the two different formulations of the CNTs and does not imply any type of substitutional doping of the CNTs. Then, the p and n-type TE inks were used for coating of the GF tows via a simple blade coating process. In more detail, 10 cm long single GF tows were cut from the reel, stabilized, and coated towards the creation of homogeneous continuous TE efficient p-type and n-type films onto the GFs. After the coating process, the tows were left to dry on a hot plate at 90 °C for 30 min.

Manufacturing of Model Composites for Thermoelectrical Characterization
Model composites were prepared and characterized for the study of the effect of the epoxy resin system addition on the thermoelectric response of the coated glass fibers. The coated single GF tows were placed flat onto a pre-cleaned silicon dog-bone shaped mould and secured in place on both sides with an adhesive tape. Then, two thin copper foil electrodes were adjusted tightly on each side of the functional single tows and glued with silver paste on the tow to ensure proper ohmic contacts. Upon drying of the contacts, a well-mixed and degassed epoxy resin/hardener/catalyst mixture was casted in the mould, which impregnated thoroughly the whole GF tow. The system was cured and post-cured following the aforementioned protocol. At least five model composites for each category were fabricated and the mean experimental values with the corresponding standard deviations are presented. The electrical properties and the thermoelectric response of the model composites were measured with the set up that is displayed in Figure 1. Manufacturing process of the thermoelectric generator (TEG)-enabled model composites (a) silicon mould utilized for the preparation of dog-bone shaped specimens with copper electrodes placed at both sides ready for the epoxy resin casting and (b) demonstration of the thermoelectric (TE)-enabled p and n-type model composites exposed to temperature gradient.

Preparation of Model Composite Specimens for Micromechanical Tests
Tests for the interfacial shear strength calculations were performed on single GF tows using a pull-out configuration. These tests were performed in order to evaluate the effects of the nanostructured coatings on the interfacial mechanical properties between the hierarchical fibers and the epoxy resin system.
For the selected manufacturing process, both uncoated and nanocoated GF tows with a length of 100 mm were pre-impregnated with a thin layer of epoxy resin at 3/4 of their length, using a micropipette and were semi-cured until the resin was viscous enough so as not to leak from the Manufacturing process of the thermoelectric generator (TEG)-enabled model composites (a) silicon mould utilized for the preparation of dog-bone shaped specimens with copper electrodes placed at both sides ready for the epoxy resin casting and (b) demonstration of the thermoelectric (TE)-enabled p and n-type model composites exposed to temperature gradient.

Preparation of Model Composite Specimens for Micromechanical Tests
Tests for the interfacial shear strength calculations were performed on single GF tows using a pull-out configuration. These tests were performed in order to evaluate the effects of the nanostructured Appl. Sci. 2020, 10, 5352 5 of 17 coatings on the interfacial mechanical properties between the hierarchical fibers and the epoxy resin system.
For the selected manufacturing process, both uncoated and nanocoated GF tows with a length of 100 mm were pre-impregnated with a thin layer of epoxy resin at 3/4 of their length, using a micro-pipette and were semi-cured until the resin was viscous enough so as not to leak from the impregnated tow.
These epoxy resin semi-cured tows were lowered into the aluminum moulds by using purposely made fixture is presented in Figure 2b. The fixture was designed so as to hold the pre-impregnated tows aligned during the thermal treatment of the resin. Then, 20 g of the resin/hardener/catalyst system were poured into each of the aluminum moulds. Particular attention was paid in order to avoid buckling of the tows and simultaneously maintain the impregnated embedded length in the range of 1-2 mm. Finally, the specimens were cured and post cured according to the recommended curing cycles.
Appl. Sci. 2020, 10, x 5 of 17 system were poured into each of the aluminum moulds. Particular attention was paid in order to avoid buckling of the tows and simultaneously maintain the impregnated embedded length in the range of 1-2 mm. Finally, the specimens were cured and post cured according to the recommended curing cycles. The tow pull-out experiments were conducted at room temperature using the Universal Testing Machine by Jinan Testing Equipment IE Corporation (Jinan, China) with a 2 kN Load cell. A twopiece metal jig held the epoxy resin mould while the top part was clamped using the appropriate jaw faces. A constant cross-head speed of 5 mm/min was employed for tensile loading to induce the pull out shear force at the tow/resin interphase. Tests were considered successful when the GF tow was completely pulled out from the polymer matrix. The applied load (F) and the corresponding displacement were recorded and the interfacial shear stress (σIFSS) was calculated using the equation σΙFSS = Fmax/π×df×l, where Fmax is the maximum load prior to the fracture of the GF tow, df is the diameter-width of the tow, and l is the embedded length, including the meniscus, as determined by optical microscopy.
The apparent interfacial shear strength of the functional GF tows and the epoxy resin matrix was calculated and compared with reference uncoated GF tow model composites. The mean values of shear strength presented correspond to the average of 8 successful pull-out experiments for each type of specimens. Finally, the surface morphology of the pulled-out fiber tows from all categories was studied using an optical stereoscope at low magnification.

Characterization Techniques
Contact angle measurements were acquired via a commercial automated contact angle The tow pull-out experiments were conducted at room temperature using the Universal Testing Machine by Jinan Testing Equipment IE Corporation (Jinan, China) with a 2 kN Load cell. A two-piece metal jig held the epoxy resin mould while the top part was clamped using the appropriate jaw faces. A constant cross-head speed of 5 mm/min was employed for tensile loading to induce the pull out shear force at the tow/resin interphase. Tests were considered successful when the GF tow was completely pulled out from the polymer matrix. The applied load (F) and the corresponding displacement were recorded and the interfacial shear stress (σ IFSS ) was calculated using the equation where F max is the maximum load prior to the fracture of the GF tow, d f is the diameter-width of the tow, and l is the embedded length, including the meniscus, as determined by optical microscopy.
The apparent interfacial shear strength of the functional GF tows and the epoxy resin matrix was calculated and compared with reference uncoated GF tow model composites. The mean values of shear strength presented correspond to the average of 8 successful pull-out experiments for each type of specimens. Finally, the surface morphology of the pulled-out fiber tows from all categories was studied using an optical stereoscope at low magnification.

Characterization Techniques
Contact angle measurements were acquired via a commercial automated contact angle goniometer system manufactured by Ossila Ltd (Sheffield, UK).
The morphology of the hierarchical GFs, as well as the reference GFs were evaluated via SEM using the JEOL JSM 6510 LV SEM (JEOL Ltd., Akishima, Japan) with an operating voltage of 3.5 KV.
Thermogravimetric measurements were carried out in the STA 449C analyzer Netzsch-Gerätebau GmbH, Selb, Germany. The experimental procedure included a dynamic step from 25 to 1000 • C with a heating rate of 10 • C/min at ambient conditions to examine the reference and the nanocoated GFs mass loss.
The Labram HR-Horiba scientific micro-Raman system was used. The 514.5 nm line of an Ar + ion laser operating at a power of 1.5 mW at the focal plane was employed for the Raman excitation. An optical microscope equipped with a 50× long working distance objective served both for delivering the excitation light and collecting the back-scattered Raman light. Raman spectra in the range of 90-3500 cm −1 were collected.
In order to determine the basic electrical characteristics, p and n-type bucky paper films were prepared by overnight vacuum filtering of the respective SWCNTs dispersions and four probe sheet resistance measurements were conducted (Table 1). A typical 4-probe technique (Four-Point Probe commercial system manufactured by Ossila Ltd.) was used to calculate the electrical conductivity based on the following formula: σ = 1/R × L/S × (ln2/π), where R is the measured 4 probe sheet resistance, L is the length, and S is the cross section of each sample. It should be mentioned that the employed four-point probe system with four equally-spaced, co-linear probes to the material (interelectrode probe distance/spacing: 1 mm) is capable of delivering currents between 10 nA and 150 mA, and can measure voltages from as low as 100 µV up to 10 V, which results in a sheet resistance measurement range of 3 mOhm/ to 10 MOhm/ . The values reported are mean values of at least four measurements performed on different samples.
The DC measurements were conducted with a standard two-probe method using the Agilent 34401A multimeter with 6 1 2 digits resolution. Model composites were placed and measured at 25 mm electrode-electrode distance at room temperature and 25% relative humidity (RH).
The voltage output of the model composites was determined using a custom-made setup. The samples were mounted on two metal blocks, which enabled the generation of a temperature difference. For all measurements, the one aluminum block (T cold ) was kept at room temperature (~24 • C) via mild water circulation, while the other one (T hot ) was heated until 125 • C via calibrated temperature-controlled resistors, allowing for the generation of a temperature gradients. The generated thermovoltage (∆V) was measured across the electrodes by a multimeter-voltmeter. The temperature of the two blocks was measured with an IR-thermometer (OMEGA OS-VIR50 Dual Laser Video IR Video Thermometer) to verify the temperature difference (∆T), and the Seebeck coefficient (S) was derived from the ratio ∆V/∆T.

Results
In this study, organic materials were evaluated as emerging candidates for thermoelectric power generation. Chemical doping was employed for tuning the semiconductive nature of the graphitic material from p-type to n-type. In the literature, the terms p-type and n-type are frequently used to describe which apparent carrier mobility is in larger percentage in a material, rather than the position of the Fermi level [29]. Organic semiconductors can be ambipolar and in some cases the hole and electron mobilities are comparable, but one type of carrier prevails owing to the reactivity of the prefered dopant material to stably tune the initial material [31][32][33]. Contact angle (CA) measurements and scanning electron microscopy (SEM) were employed in order to assess the specific morphological characteristics of the produced nanocoated GFs, while their thermal stability was investigated via thermal gravimetric analysis (TGA). Raman spectroscopy (RS) was used to evaluate the quality of and distribution of the SWCNTs deposited on the surface of the GFs under temperature gradients. Finally, the p-n nanostructured interphases were estimated via thermoelectric measurements and micromechanical studies.

Morphological Measurements of the Coated GFs
Representative contact angle images for each case are represented in Figure 3. Contact angle is a major surface property in terms of assessing the solid liquid interphase. In this study, the static contact angles of the SWCNTs inks onto the GFs as a measure of the surface wetting properties of the coated glass fibers.  The coated glass fiber morphologies were also studied using SEM in order to determine the homogeneity of the coatings and illustrate the specific characteristics of the nanostructuring onto the microscale reinforcing phase. Figure 4a,b illustrates the hierarchical morphology of the coated GFs with homogeneously distributed SWCNTs covering the fiber surfaces for the p-and n-type SWCNT coated GFs, respectively, at two different magnifications. The coated glass fiber morphologies were also studied using SEM in order to determine the homogeneity of the coatings and illustrate the specific characteristics of the nanostructuring onto the microscale reinforcing phase. Figure 4a,b illustrates the hierarchical morphology of the coated GFs with homogeneously distributed SWCNTs covering the fiber surfaces for the p-and n-type SWCNT coated GFs, respectively, at two different magnifications.

Thermogravimetric Analysis of the TE-Enabled GFs
In order to assess the thermal stability of the produced functional hierarchical structures, dynamic thermogravimetric scans were performed from 25 • C to 1000 • C. Thermogravimetric analysis under air atmosphere is very sensitive to the increase of the surface area which promotes the oxidation of the examined material. High thermal stability of these hierarchical structures is critical for the thermoelectric applications that involve high operational temperatures. Figure 5 illustrates the TGA graph of the uncoated and the nanocoated GF tows.
Appl. Sci. 2020, 10, 5352 8 of 17 The coated glass fiber morphologies were also studied using SEM in order to determine the homogeneity of the coatings and illustrate the specific characteristics of the nanostructuring onto the microscale reinforcing phase. Figure 4a,b illustrates the hierarchical morphology of the coated GFs with homogeneously distributed SWCNTs covering the fiber surfaces for the p-and n-type SWCNT coated GFs, respectively, at two different magnifications.

Thermogravimetric Analysis of the TE-Enabled GFs
In order to assess the thermal stability of the produced functional hierarchical structures, dynamic thermogravimetric scans were performed from 25 °C to 1000 °C. Thermogravimetric analysis under air atmosphere is very sensitive to the increase of the surface area which promotes the oxidation of the examined material. High thermal stability of these hierarchical structures is critical for the thermoelectric applications that involve high operational temperatures. Figure 5 illustrates the TGA graph of the uncoated and the nanocoated GF tows.

Spectroscopic Measurements of the TE-Enabled GFs
The Raman spectra of the p-type SWCNTs coated GF and the n-type SWCNTs coated GF were acquired at ambient conditions (room temperature-RT), as well as at temperature gradients of 50 °C and 100 °C by performing measurements at 75 °C and 125 °C, respectively.
The respective results are presented in Figure 6. A linear baseline was subtracted from the spectra of the p-type and the n-type SWCNTs coated GFs.

Spectroscopic Measurements of the TE-Enabled GFs
The Raman spectra of the p-type SWCNTs coated GF and the n-type SWCNTs coated GF were acquired at ambient conditions (room temperature-RT), as well as at temperature gradients of 50 • C and 100 • C by performing measurements at 75 • C and 125 • C, respectively.
The respective results are presented in Figure 6. A linear baseline was subtracted from the spectra of the p-type and the n-type SWCNTs coated GFs.

Spectroscopic Measurements of the TE-Enabled GFs
The Raman spectra of the p-type SWCNTs coated GF and the n-type SWCNTs coated GF were acquired at ambient conditions (room temperature-RT), as well as at temperature gradients of 50 °C and 100 °C by performing measurements at 75 °C and 125 °C, respectively.
The respective results are presented in Figure 6. A linear baseline was subtracted from the spectra of the p-type and the n-type SWCNTs coated GFs.

Thermoelectric Response of the Model Composites
In Table 1 the TE properties of both p-type and n-type bucky paper films are presented. The conductivity and the TE response introduced from the coating can be attributed to the nanoparticle-based networks deposited onto the fiber surfaces creating the electron transport pathways. In fact, the electrical conductivity and the Seebeck coefficient values can be explained more precisely if we consider the existence of a continuous self-assembled layer with thickness (t) onto the fiber surface with electrical conductivity (σ) stemming from the electrostatic interactions [34].
More analytically, when the anionic surfactant (SDBS) is added to distilled water it tends to ionize and give a negative charge to the created micelles which include the dispersed de-aggregated SWCNTs. Conversely, the positively charged free silanol particles from the glass fibers surface interact with the ambient oxygen groups result in non-covalent bonding of the CNTs [35]. The resistivity (ρ) of an epoxy coated GF tow along the longitudinal direction is given by the formula: ρ = A × R/L (A: circular cross section area). Executing this equation, the coated GF tows with SWCNTs (p and n-type) consist of a highly conductive continuous network and could be assumed approximately as a tubular-like geometry resistor (V = π × r2 × L) with electrical conductivity of σ = (4L)/π × d2 × R [23]. The average diameter (d = r/2, r: radius) was~1654 µm for p GF tows within the epoxy resin and~1600 µm for the n-type GF tows within the epoxy resin, measured via a calibrated optical microscope. The electrical resistance (R) was measured along the length (L) via a two-probe method. It should be mentioned that the nanoparticle-based layers for both p-type and n-type SWCNTs coatings onto GFs possess lower electrical conductivity values in comparison to the aggregated and highly dense structure of the corresponding bucky paper films form, as presented in Table 1. Furthermore, the bulk thermoelectric properties of both p-and n-type bucky paper films were studied. The positive measured values of the Seebeck coefficient correspond to the p-type semi-conducting behavior and the negative to the n-type, respectively. The hierarchical GF tows model composites were evaluated in terms of their thermoelectric response as presented in Figure 7.

Study of the Nanostructured Interfacial Mechanical Properties
Apart from the thermoelectric functionality, the interfacial shear strength of the hierarchical GFs was also measured by tow pull-out tests and compared to the interfacial properties of the reference uncoated GF tows. While the influence of the CNTs on the interfacial bonding of the GF with epoxy resin is usually determined by measuring the interfacial shear strength through single fiber fragmentation [2] or pull-out tests [36,37], the pull out test configuration was deemed as more appropriate for this study and was used to quantify the quality of interfacial bonding between the epoxy matrix and the nanocoated glass fibers which were used in this study. Figure 8a shows the results of apparent interfacial shear strength (σIFSS) measurements, which is indicative of the stress transfer efficiency of the interphase. As can be observed, the uncoated glass fiber tows (REF GF) exhibited the highest interfacial shear strength, i.e., 42.8 ± 0.9 MPa. Comparatively, the p-type SWCNTs coated GF model composites exhibited lower interfacial shear strength values at 38.5 ± 1.5 MPa, that is a 10.1% decrease in comparison to the uncoated GFs. For the n-type SWCNTs coated GF model composites the interfacial shear strength was about 30.8 ± 3.4 MPa, which corresponds to a 28.1% reduction compared to the uncoated GF tows. (a) schematic illustration of a p-n junction between the model composites created by interconnecting in series the p-type and the n-type model composites by a copper foil at the hot side (125 • C), while at the cold side (~24 • C) the copper electrodes were attached with the multimeter for a 2-probe measurement, graphs of (b) electrical resistance-electrical conductivity, (c) Seebeck coefficient-Power factor and (d) ∆V vs. ∆T curves for the coated p and n-type TE model composites.

Study of the Nanostructured Interfacial Mechanical Properties
Apart from the thermoelectric functionality, the interfacial shear strength of the hierarchical GFs was also measured by tow pull-out tests and compared to the interfacial properties of the reference uncoated GF tows. While the influence of the CNTs on the interfacial bonding of the GF with epoxy resin is usually determined by measuring the interfacial shear strength through single fiber fragmentation [2] or pull-out tests [36,37], the pull out test configuration was deemed as more appropriate for this study and was used to quantify the quality of interfacial bonding between the epoxy matrix and the nanocoated glass fibers which were used in this study. Figure 8a shows the results of apparent interfacial shear strength (σ IFSS ) measurements, which is indicative of the stress transfer efficiency of the interphase. As can be observed, the uncoated glass fiber tows (REF GF) exhibited the highest interfacial shear strength, i.e., 42.8 ± 0.9 MPa. Comparatively, the p-type SWCNTs coated GF model composites exhibited lower interfacial shear strength values at 38.5 ± 1.5 MPa, that is a 10.1% decrease in comparison to the uncoated GFs. For the n-type SWCNTs coated GF model composites the interfacial shear strength was about 30.8 ± 3.4 MPa, which corresponds to a 28.1% reduction compared to the uncoated GF tows. appropriate for this study and was used to quantify the quality of interfacial bonding between the epoxy matrix and the nanocoated glass fibers which were used in this study. Figure 8a shows the results of apparent interfacial shear strength (σIFSS) measurements, which is indicative of the stress transfer efficiency of the interphase. As can be observed, the uncoated glass fiber tows (REF GF) exhibited the highest interfacial shear strength, i.e., 42.8 ± 0.9 MPa. Comparatively, the p-type SWCNTs coated GF model composites exhibited lower interfacial shear strength values at 38.5 ± 1.5 MPa, that is a 10.1% decrease in comparison to the uncoated GFs. For the n-type SWCNTs coated GF model composites the interfacial shear strength was about 30.8 ± 3.4 MPa, which corresponds to a 28.1% reduction compared to the uncoated GF tows.

Surface Structure and Wettability of the Coated GF Tows
The good wetting of the glass fiber tows by both the p and n-type SWCNTs ink droplets is indicative of the inherent hydrophilic behavior of the glass fibers, which promotes the forming of a uniform SWCNTs film on the GFs and is a measure of the level of adhesion. In Figure 3, images are shown for each case, captured just after the drop was suspended onto the surface of the glass fibers and also 60 s later to evaluate the progression of the wetting. For the GFs coated with p-type SWCNTs the average contact angle was found to be ca. 56.47 • and for the n-type 52.39 • .
As was observed in the SEM study (Figure 4), both the p-and n-type SWCNTs films formed on the GFs are satisfactorily uniform and dense exhibiting an interconnected network of SWCNTs which fully cover the surface. Some SWCNTs agglomerates are barely visible. The observed SWCNT-SWCNT junctions are impart the high values of electrical conductivity to the films, which alter the inherently dielectric behavior of the GF tows.

Thermal Stability of the Produced Coated GF Tows
Thermogravimetric analysis was used in order to characterize the thermal stability of the SWCNTs coated GFs in comparison to the reference GFs.
As depicted in Figure 5, a minor mass loss in the order of almost 0.05% was observed for the reference GFs as well as for the n-type SWCNTs coated GFs above 100 • C and up to 241.5 • C and 213.2 • C, respectively. The p-type SWCNTs coated GFs depicted a similar behavior at that temperature range losing, however a slightly higher amount of its mass, namely 0.28% up to 175.3 • C. This mass loss is attributed to the evaporation of residual moisture for all the samples and is observed to be increased in the p-type SWCNTs coated GFs due to the increased amount of carboxy moieties. At higher temperatures, between about 241.5 • C and 539.2 • C the reference GFs lost 0.7% of their mass due to the combustion of the silane type sizing [38][39][40]. Then, the mass of the E-glass fibers remained almost constant without considerable weight loss up to 1000 • C.
The thermal degradation of the p-type and the n-type molecular doped SWCNTs is a multistage process due to the existence of the various functional groups that are absorbed on their surface or are interacting with it. The functional chemical groups as moieties that SWCNTs contain are in general decomposed during the second temperature range that spans from roughly 200 • C to 500 • C. In more detail, the p-type SWCNTs exhibited the typical decomposition behavior of pristine SWCNTs that are inherently p-doped by a small amount of adsorbed oxygen containing groups. The oxygen-based moieties promote the electron withdrawing process from the CNT electron-rich back-bone, effectively introducing holes as main charge carriers (p-doping). The p-type SWCNTs coated GFs lost ca. 1% of their mass up to 373 • C that can be attributed to smaller in size oxygen containing functional groups like hydroxy and ether moieties. Another 1.35% mass loss was observed up to 531 • C that can be due to larger oxygen containing moieties like carboxyl and ester moieties [41]. The initial mass loss of SDBS molecules is owing to burning with the presence of oxygen (330-450 • C) and the burning of the SDBS continued to occur above 800 • C [42]. Overall, due to the fact that the SDBS concentration is lower (2 mg/mL) in the n-type SWCNTs coatings, this may also explain the loose of less weight compared to the p-type SWCNTs coatings (5 mg/mL). The observed loss amount is in agreement with the percentage of SDBS and CNTs that were used for the preparation of the inks. Additionally, any amorphous carbon that might be present into the SWCNTs coating would also be combusted at this temperature range. A further 0.7% mass loss occurred up to 756 • C that can be attributed to the combustion of the SWCNTs. Then, the mass of the underlying GFs remained almost stable up to 1000 • C with only a further 0.18% mass loss that can be attributed to some possible remaining percentage of catalyst particles from the production process of the CNTs [43].
A slightly different behavior was presented by the n-type SWCNTs coated GFs. The decomposition of the nitrogen containing moieties that are included in the PEI molecules, which are entangled in between the CNT network, was responsible for the 0.53% mass loss from 213 • C up to 368 • C. A further 1% mass loss was observed up to 626 • C that is attributed the total decomposition of the SWCNTs. Afterwards the mass of the underlying GFs remained almost stable up to 1000 • C, with only a small amount equal to a further 0.24% being decomposed up to 1000 • C maybe on account of some possible remaining percentage of catalyst particles due to the production process of the CNTs [43].

Raman Analysis of the TE-Enabled GF Tows
The p-type and the n-type SWCNTs coatings onto the GFs possess all the expected vibrational modes commonly encountered in graphitic materials. More specifically, the most prominent band of the spectrum of SWCNTs was the graphitic or G band with E 2g symmetry located at about 1590 cm −1 . The radial breathing mode, which is a unique phonon mode of SWCNTs was also observed at the 120 cm −1 to 140 cm −1 region. Another disorder-induced band appearing in the first order spectrum of the SWCNTs is the D band, at about 1340 cm −1 . The D band is caused by either amorphous carbon or sp 3 carbon atoms at defect sites on the CNT sidewalls [44]. Furthermore, the 2D peak dominated the second order spectrum of the SWCNTs coatings. This band was located at about 2685 cm −1 . This overtone results from a double resonance process and is a symmetry allowed band which requires two phonons to be completed [45,46].
As can be seen in Figure 6b,c, no changes were observed in the Raman spectra of both the p-type and the n-type SWCNTs coatings onto the GFs when they were subjected to temperature gradients of 50 • C and 100 • C, apart from minor positive frequency shifts of all vibrational modes which were attributed to the increased temperature. This result indicates that the hierarchical GFs which are coated with the organic SWCNTs inks of both p-type and n-type TE behavior are chemically and structurally stable.

Evaluation of the Model Composites Thermoelectric Efficiency
The electrical resistance as well as the electrical conductivity of the prepared model composites are presented in Figure 7b. The p-type SWCNTs coated GF model composites exhibited electrical resistance of 127.2 ± 2.9 Ohm, which corresponds to electrical conductivity of 1.097 ± 0.05 S/cm, while for the n-type SWCNTs coated GF model composites showed very similar values i.e., 95.6 ± 1.1 Ohm and 1.561 ± 0.023 S/cm, respectively. The high electrical conductivity of the coated GFs is a key parameter for thermoelectric performance, since it can facilitate the flow of the thermally induced charge carriers upon being exposed to a temperature difference. Additionally, as aforementioned, low thermal conductivity is also desirable for the sustainability of the thermoelectric effect. In the case of the SWCNT interphase, it may also be postulated that the infusion of the thermally insulating epoxy molecules reduces thermal conductivity, thus further enhancing thermoelectric performance. More analytically, this mechanism creates a barrier at the SWCNTs junctions and prevents the direct physical contact between them resulting in reduced thermal energy transport. [23,29].
The Seebeck coefficient (S = ∆V/∆T) was derived from the slope of ∆V vs. ∆T curves by linear fitting. Figure 7c shows the Seebeck coefficients of the p and the n-type model composites. For the p-type model composite, the Seebeck coefficient was found to be +16.2 ± 6 µV/K corresponding to a power factor of 0.02 µW/m·K2 and for the n-type −28.4 ± 4 µV/K corresponding to a power factor of 0.12 µW/m·K 2 .
As was observed, the p-type SWCNTs coated GF model composites exhibited slightly lower values of Seebeck coefficient compared to the corresponding bucky papers. This can be attributed to the high nitrogen amount of the hardener within the epoxy system, acting as a n-dopant [32]. On the other hand, for the n-type SWCNTs coated GF model composites the Seebeck coefficient values remain almost identical. Figure 7d depicts the generated thermovoltage upon exposure to various values of ∆T, for both p and n-type SWCNTs coated GF model composites. More specifically, the TE voltage of the in-series electrical and the in parallel thermal connection between the p-type and the n-type model composites creating a p-n junction are being displayed. As can be observed, within the studied temperature range, the generated voltage increased linearly with increasing temperature, R 2 > 0.98 for all cases. The measured voltage output was 4.46 ± 0.3 mV with a device resistance of 235.5 ± 2.5 Ohm at 100 • C ∆T. The maximum output of the p-n model composites TEG device was calculated 0.02 µW at a temperature difference of 100 • C, according to the following equation P max = ∆V 2 /4 × RTEG [24]. The derived maximum output power density of the device (p-n model composites pair) was 0.016 µW/cm 2 .
Overall, the relatively high values of the electrical resistance were directly related to the area of the evaluated coated GF tows. It is expected that by increasing the area of the thermoelement (in this case the single coated GF tow), the electrical conductivity could be significantly increased while the Seebeck coefficient would not be affected as it is an inherent material property and is independent of the geometry of the evaluated specimen. To further elaborate on this statement, when the cross-sectional area of the coated substrate is increased, by increasing the thickness of the CNT coating on the GF substrate, its electrical resistance is reduced. Generally, the electrical resistivity of the CNTs films decreases as the thickness of the coatings is being increased. The property of the electrical resistivity highly depends on carrier concentration and carrier mobility, while the electrical resistance depends on the resistor shape geometry and the specific material. Electrical resistivity is an intrinsic property of a conductive film which is directly proportional to the total resistance R, an extrinsic quantity that depends on the length and cross-sectional area of the corresponding film. Larger cross section areas present lower values of electrical resistance. In the case of SWCNTs, changes within the conductive network due to thickness optimization are able to positively impact the electrical conductivity. Both cases are true until the bulk resistance value is reached, thereby no further increments in electrical conductivity can be achieved. There also exists an optimal area that produces the highest possible thermoelectric values, as it is able to host the highest possible electron flow/current under a fixed ∆T. Therefore, since the Seebeck coefficient is a material property, the power factor of the coated GFs can be optimized by combining the most appropriate substrate dimensions and coating thickness. This is of the utmost importance for the exploitation of the nanomaterial coated reinforcements as thermoelectric elements in composite structures.

Evaluation of Interfacial Shear Strength
Tow pull out testing as opposed to single fiber pull-out testing was the preferred methodology for assessing the stress transfer between the nanocoated tow and the matrix, as it resulted in a more realistic measurement of the interfacial shear strength. Moreover, it was possible to effectively reduce the errors associated with the handling of single fiber filaments, which typically requires extremely careful manipulation due to their small size and brittle nature [47,48]. Several trials were performed with different specimen designs in order to consider all the different aspects that are involved in the particular test. Since the tow pull out process is not a standard test [49,50], preliminary tests were performed to optimize parameters such as the embedded length and the resin pre-impregnation process before insertion in the matrix.
Fracture morphologies of the pulled-out tows obtained via optical microscopy reveal a cohesive type of failure in all cases, since epoxy material remained on the tow surface after the pull-out process. In general, the GF tow was covered by a cone of resin and as such it was detached from the matrix. In the case of the n-type tows, a mixed type of failure mode was observed, whereby fiber/matrix-to-matrix failure was also present. The fiber-to-matrix failure, whereby the GF tow was completely detached from the resin, was not observed in any of the tested specimens. As should be noted, the persistent presence of cohesive failure within the matrix suggests that measured values were an underestimate of the true IFSS. This was also suggested during the process of experimental optimization where it was observed that small embedded length and low values of pull out force correlated more with fiber to matrix failure than cohesive failure.
In conclusion, it can be assumed that the presence of surfactant for the dispersions of both SWCNTs inks affected negatively the IFSS when a commercial epoxy resin was used as a matrix. Further studies could be performed oriented to optimize the coating process and simultaneously assess the effect of the surfactant existence on the interfacial shear strength [51,52]. As is well known, a strong interface provides efficient stress transfer but promotes brittle failure, whereas a weaker may promote energy dissipation and thus improve toughness. Balancing these conflicting requirements by engineering composite interphases to improve optimal strength and toughness values remains a great challenge.

Conclusions
The scope of this research study was to introduce and evaluate functional interphases between modified GFs with p and n-type SWCNTs model epoxy composites. Efforts were devoted to the morphological inspection as well as the evaluation of the structural changes that might be caused by the functional nanocoated GFs in the presence of temperature gradients. The thermal stability of the nanocoated GF tows was investigated through thermogravimetric analysis (TGA). Additionally, the chemical and structural stability of the TE nanoparticle-coated GF tows were studied by Raman measurements at elevated temperatures. The Raman spectra of p-type SWCNTs coated GFs and n-type SWCNTs coated GFs were acquired at ambient conditions, as well as at temperature differences of 50 • C and 100 • C by performing measurements at 75 • C and 125 • C. The results indicated that both the p-type and n-type TE enabled GF tows are chemically and structurally stable. The effect of the integration of the modified nanocoated GFs into a conventional epoxy resin was evaluated. The TE performance of the functional GF tows was only slightly affected by the resin encapsulation. Both p and n-type SWCNTs model composites exhibited a combination of high electrical conductivity and Seebeck coefficient. N-type model composites exhibited higher electrical conductivity compared to the corresponding p-type, due to their higher intrinsic electrical properties, as measured on SWCNTs bucky paper films. Moreover, the n-type model composites exhibited excellent behavior as thermoelectric materials with remarkably high values of Seebeck coefficient very close to the properties of the bulk SWCNTs material. An epoxy-based TEG model composite was manufactured to demonstrate the potential application of the technology in a TE generator, using a pair of alternating p-and n-type junctions interconnected in series via a metallic junction. The interfacial shear strength for both p and n-type TE enabled model composites was assessed by tow pull-out tests. The presence of SDBS as surfactant for the dispersion of the SWCNTs affected negatively the IFSS when epoxy resin was used as a matrix leading to a slight reduction of the IFSS values for both systems comparatively to the reference one. Finally, the fractographic study of the tested samples revealed that the matrix-to-matrix failure was more dominant than fiber/matrix-to-matrix failure, indicating strong interfacial adhesion.
Further optimization of the thermoelectric performance which involves the increase of both the electrical conductivity and the Seebeck coefficient could be as achieved via doping of SWCNTs, use of different types of SWCNTs and creating efficient p-n heterojunctions for the fabrication of TEG module devices. The generation of thermoelectricity using structural engineering materials that are routinely exposed to high temperatures will represent a breakthrough in multifunctional materials, rendering them an attractive technology for aerospace and automobile applications.

Conflicts of Interest:
The authors declare no conflict of interest.