Dynamic mechanical response of carbon nanotube yarns and their in situ electrical measurements

Abstract Carbon nanotube yarns (CNTYs) are outstanding hierarchical fibers with electrical properties that depend on temperature and mechanical stimuli, which makes them attractive as smart materials for sensing applications. In order to better understand the electrical response of CNTYs under a combination of dynamic loading and variations of temperature, their monotonic and cyclic tensile mechanical response, in situ Raman spectroscopy, mechanical hysteresis, and dynamic mechanical analysis (DMA) are investigated herein. The piezoresistivity and thermoresistivity of CNTYs were characterized to correlate the contribution of temperature and strain to the effective electrical response of the CNTY under DMA testing. It was found that the tensile load bearing of the CNTYs is governed mainly by structural changes of its bundles/fibrils, rather than by stretching of the carbon bonds. The large energy dissipation capabilities of the CNTYs arise mainly from friction among their fibrils and bundles, and also due to irreversible morphological changes in their hierarchical structure. The (mechanically constrained) thermoresistive characterization showed that the electrical resistance of the CNTYs decreases with increasing temperature, yielding an average temperature coefficient of resistance of −8.63 × 10−4 K−1. The electrical response of the CNTYs during DMA temperature scans is governed by their thermoresistive response.


Introduction
Carbon nanotubes (CNTs) are an allotropic form of carbon composed of sp 2 hybridized carbon atoms rolled up forming a hollow cylindrical structure.Since the discovery of CNTs (Iijima, 1991), they have been widely studied due to their outstanding mechanical, electrical, thermal, and coupled (electromechanical, thermo-electrical) properties.CNTs can be used to produce all-carbon macroscopic assemblies such as CNT arrays, films, bucky papers, and carbon nanotube yarns (CNTYs), which have a broad range of applications as multifunctional and smart materials (Zhang et al., 2005;Shahidi and Moazzenchi, 2018;Kim et al., 2017;Ali et al., 2022).As smart materials, CNTYs are of particular interest for artificial muscles, motion detection, and structural health monitoring applications due to their high specific (per unit weight) mechanical properties, and high electrical conductivity, which changes with strain (Abot et al., 2010;Hehr et al., 2014;Ryu et al., 2015;Mikhalchan and Vilatela, 2019).CNTYs are twisted fibers made of millions of CNTs bonded by van der Waals forces and CNT entanglement (Misak et al., 2014;Miao et al., 2020;Miao, 2013).The CNTs that comprise the CNTYs are grouped into CNT bundles, and those may further group into fibrils that make up the yarn.This hierarchical structure renders a very strong property-structure relationship (Miao et al., 2020;Miao, 2013;Jung et al., 2018).Studies of the mechanical behavior of CNT assemblies and CNT bundles have provided initial knowledge of the mechanisms that influence the mechanical response of CNTYs (Filleter et al., 2011;Gupta et al., 2021;Liew et al., 2005;Kumar and Cronin, 2007).This knowledge might help to improve properties and tailor the mechanical/electro-mechanical response of the CNTYs.For example, large energy dissipation capacity may be achieved through mechanisms such as zipping/unzipping, self-organization, and bundle sliding (Liu et al., 2015;Zhao et al., 2018;Zhang et al., 2020).However, CNTYs exhibit complex mechanical behavior that depends on multiple factors, such as their twist level (Mirzaeifar et al., 2015;Miao, 2016;Anike et al., 2019), porosity (Miao et al., 2010), and packing density of CNT bundles (Li et al., 2020).Parameters such as the test length (gage length) (Liz ak, 2002) and strain rate (Zhang et al., 2012) also impact the CNTYs mechanical response.All these factors may also affect the electrical response of the CNTYs under quasi-static and dynamic loading.For this reason, the study of electromechanical properties of CNTYs is still challenging, but at the same time presents opportunities for research and development in the smart materials area.Lekawa-Raus et al. (Lekawa-Raus et al., 2014) measured the electrical resistance changes of CNTYs during cyclic tensile tests.They found that the electrical resistance of the CNTYs increases with strain, which was attributed to the decrease in the contact area of CNT bundles.In addition, they observed that CNTYs present a hysteresis loop in both electrical and mechanical responses during the cyclic tensile tests.This hysteresis loop was attributed to energy loss due to the breaking and remaking of bonds between CNT bundles.Hehr et al. (Hehr et al., 2014) studied the energy dissipation of dry-spun CNTYs during cyclic tensile tests and as a function of temperature.They observed that the energy dissipation capacity of the CNTYs is lower than rubber but higher than epoxy resins.On the other hand, they found a slight variation in the energy dissipation capacity with temperature, but the governing mechanisms remain unknown.Thus, questions such as the relative contribution of C-C bond stretching and structural deformation of the fibrils during tensile loading, and their impact on the electrical response of the CNTY, remain unsolved.The role of temperature variations concurrent with this electro-mechanical response has also not been tackled, and they represent the main issues that this work aims to tackle.Therefore, the objective of this work is to investigate the dynamic mechanical response of dry-spun CNTYs, coupled with in situ electrical measurements, to advance the understanding of the relationship between the CNTY hierarchical structure and their electro-mechanical response.For this, monotonic and cyclic tensile testings of CNTYs were conducted, with and without in situ electrical measurements.In situ Raman spectroscopy was also performed during tensile testing of CNTYs to better understand the load/strain transfer and energy dissipation capabilities of the yarns at different hierarchical levels (individual CNTs vs. bundles).In addition, the electrical response of the CNTYs was measured during dynamic mechanical (DMA) tensile testing.The study of the piezoresistive and thermoresistive response of the CNTYs helps to separate each contribution and better explain the DMA-electrical results.

Materials
The CNTYs were dry-spun from the sides of 500 mm-high vertically aligned arrays composed of CNTs grown by water-assisted chemical vapor deposition.According to the supplier (Nanoworld Laboratories, Cincinnati, USA), the CNTs that make up the CNTYs have 2 to 3 walls with an outer diameter ranging from 9 to 12 nm (Anike et al., 2019;Jayasinghe et al., 2011).The CNTs have a volumetric mass density of 0.78 g/cm 3 , which was obtained considering their number of walls (3), their inter-shell distance (0.34 nm), and their outer diameter (11 nm) as described in (Laurent et al., 2010).Measurements indicate that CNTYs have a twist angle (a) of 29.9 ± 5.29 .In addition, dedicated statistical measurements of the CNTY diameter (d) indicated that a statistical distribution of diameters exists (28-39 mm), with mean value d ¼ 33.3 ± 1.72 mm.This diameter was obtained by fitting a 3-parameter log-logistic distribution from 480 diameter measurements from scanning electron microscopy (SEM) images (see Figure 1(a)).The measurements were obtained from 4 samples of the CNTYs (6 images per sample and 20 measurements per image).SEM images were obtained using a JEOL JSOL-6360-LV microscope (Tokyo, Japan) with magnifications of 500Â and acceleration voltages of 20-25 kV.The CNTYs comprise CNT bundles/fibrils with a diameter ranging from 99-436 nm, and a root mean square roughness of 62.2 ± 12.5 nm, obtained from atomic force microscopy (AFM) images (see Figure 1(b)), using the tapping mode (dynamic contact mode) in the air of an SPM-8 equipment from Bruker (Kontich, Belgium).The CNTYs have a volumetric mass density (q) of 0.25 g/cm 3 with calculated dry (unstretched) porosity (/ P ) of 0.69, and a linear density (q L ) of 0.21 tex (tex ¼ g/km).Table 1 shows a summary of the morphological and structural properties of the CNTYs. .The CNTYs were prestretched with a very small load ($0.73 mN), in order to minimize variability in the initial state of the CNTY.A 75 mg mass (fragment of a needle) was attached to one end of the yarn and subsequently bonded onto a cardboard or polyimide (Kapton) frame with a rectangular internal window, as shown in Figure 2. The ends of the CNTYs were fixed to the frame using a cyanoacrylate adhesive and the  62.2 ± 12.5 frame was cut in the middle before running the tests (except for the constrained thermoresistive characterization).To measure the electrical resistance of the CNTYs by the point probe method, four 38-gauge copper wires were cemented to the CNTY with carbon-based conductive paint (Bare four-Conductive, London, England) or silver paint (Structure Probe, West Chester, USA), as shown in Figures 2b,c.All tests that involved temperature rising used a Kapton frame (instead of cardboard) and silver paint (instead of carbon-based conductive paint).Five specimens (replicates) were considered for all tests.

Raman spectroscopy and in situ tensile testing
Raman spectroscopy of individual CNTYs was carried out with a confocal Renishaw inVia Raman spectrometer (Wotton-under-Edge, England), using a 50Â objective lens, a 532 nm (green) laser of 50 mW (using 1% of its power), 60 s exposure time, a 1800 lines/mm grating and laser spot size of 1 mm.The CNTY specimens were prepared as described in Section 2.2 with the dimensions shown in Figure 3.For the in situ Raman spectroscopy analysis, the CNTYs were strained in multiple steps using a special test stage (rig) designed to that aim; see Figure 3.Each strain step deformed the CNTY an amount DL ¼ 50 mm (corresponding to 0.33% strain) in $30 s until CNTY failure was reached.The strain of the CNTY (e) was calculated by dividing DL by the gage length of the specimen (L g ¼ 15 mm, free length in the window of the frames; see Figure 3).Before the first step (before loading the specimen) and right after each strain step, a 5 mm Raman line scan mapping was performed at the central region of the specimen and along the CNTY axial direction (x), as schematized in Figure 3.The Raman line scan covered 5 mm with steps of 1 mm, for a total of 6 Raman spectra.Each line scan mapping took about 20 min (time between steps) using the automated scanning function of the Renishaw Raman equipment.The Raman peaks were fit to a Lorentzian function to obtain the center position and peak parameters.The Raman parameters of each line scan were averaged, using the standard deviation as a metric of dispersion.Since Raman line mapping collects information from a wider region than a simple point analysis, it was verified that the mapping technique greatly assists in reducing data variability and obtaining more reliable information about the yarn.

Electromechanical characterization under tensile loading
CNTY specimens (Figure 2(b)) were tested in uniaxial tension under monotonically increasing load up to failure using a Shimadzu AGS-X (Kyoto, Japan) universal testing machine, with a crosshead displacement speed of 0.5 mm/min.The electrical resistance of the CNTYs was measured in situ (during tensile testing) by the four-point probe method, using a Keysight 34980 A multifunction switch (Santa Clara, USA) with 34921 A terminal block module controlled by the BenchLink Data Logger Keysight software.The strain (e) was calculated by dividing the crosshead displacement over the gage length, L g ¼ 25 mm (see Figure 2(b)).Instead of using the volumetric stress (r) obtained by considering the yarn as a continuous solid, the specific stress (r=q, widely used for textiles) was considered.r=q was by obtained dividing the applied load (F) by the linear density q L (r=q ¼ F=q L ).The units of r=q are N/tex (1 N/tex ¼ 1 GPa/(g/cm 3 )).The change of electrical resistance (DR) was divided by the electrical resistance of the CNTYs at the reference state (R 0 ), that is, at room temperature (25 C) and e ¼ 0. This produced a fractional change of electrical resistance (DR=R 0 ).The gage factor (GF, piezoresistive sensitivity) of the CNTY was obtained from the slope of the linear fit of the DR=R 0 vs. e curve in the 0.2% e 1% range.The specific tensile modulus (E=q) was obtained from the slope of the linear fit of the r=q vs. e curve, in the same strain interval as the GF: In addition, the specific electrical conductivity of the CNTY (n=q, in SÁcm 2 /g) was calculated by dividing the distance between electrodes (same as L g ¼ 25 mm; see Figure 2(b)) by the product of R 0 and q L : All properties of the CNTYs were reported using the arithmetic mean as the nominal value and the standard deviation as a metric of dispersion.

Cyclic tensile testing and hysteresis
The CNTY specimens (Figure 2(a)) were tensile tested under strain-controlled cyclic loading by using a Shimadzu AGS-X universal testing machine.Two types of uniaxial cyclic tensile tests were conducted, viz.cycling under constant strain, and cycling under incremental strain.For the first one (constant strain), all test cycles consisted of stretching the specimen to DL ¼ 25 mm (e ¼ 1%), and then returning the specimen to the original position (DL ¼ 0).All cycles were carried out with a crosshead displacement speed of 0.5 mm/min, both in the loading and unloading cycles.A total of 300 cycles were performed on each specimen.The second type of cycling testing involved increasing the crosshead displacement by 25 mm (e ¼ 0.1%) with each new cycle until DL ¼ 400 mm (e ¼ 1.6%, 16 cycles) was reached.Further details of this incremental loading test can be found in Section S1 of the supplementary information.
To quantify the mechanical hysteresis of each cycle, the residual specific stress (r Res =q) and the hysteresis loop (H) were obtained as depicted in Figure 4.The hysteresis loop (a trajectory function) is the energy dissipated due to internal friction within the material at the molecular or structural level.On the other hand, the residual specific stress (a point function) indicates the amount of specific stress that remains in the specimen upon unloading, which results from irreversible strain.The hysteresis loop was quantified by the area between the loading and unloading curves, while the residual specific stress was obtained from the difference between r=q at the end of each cycle and r=q at the end of the first cycle.It should be noted that, for plotting purposes, all curves were slightly shifted to a very small constant amount on the vertical axis to avoid surplus negative values.This small force arises from the fine adjustment (preload) of the fiber specimen upon tightening the clamps.Finally, the hysteresis loop was used to calculate the normalized hysteresis parameter (H N ) as, where e max and r max =q are the maximum values of strain and specific stress of each cycle, respectively (see Figure 4).The normalized hysteresis parameter is used to provide a more appropriate metric to compare the path-dependent hysteresis.

Constrained thermoresistive characterization
The (constrained) thermoresistive characterization of the CNTYs was carried out to better understand the electrical response of the CNTY during DMA testing and, thus, was conducted emulating the test rig conditions of the DMA.
The test was conducted by measuring the electrical resistance of the specimen (Figure 2(c)) while heating inside a Perkin Elmer DMA 7 dynamic mechanical analyzer (Waltham, USA).This test was conducted inside the DMA equipment without loading the specimen, that is, the DMA chamber was initially used just for heating.Since the CNTY is fixed within the DMA tensile clamps (and not free to expand), the test is referred to as "constrained".In order to measure the electrical response of the CNTYs within the DMA equipment, four 38-gauge copper wires were cemented to the CNTY with silver paint.The electrical resistance of the CNTYs was measured by the four-point probe method, using a Keysight 34465 A digital multimeter (Santa Clara, USA).The DMA setup (see Figure 2(c)) requires the application of either a static or a dynamic force (or both) to simulate the experiment and run the heating chamber.The dynamic force was set to zero for this test.To avoid significant loading of the CNTY specimen, the constrained thermoresistive test was conducted keeping the Kapton frame of the specimen (without cutting the side legs in Figure 2(c)) and applying a very small constant static force of 20 mN.Using the strength of materials concepts, the partition of the total force supported by the CNTY is estimated as $1 mN, while the remaining force ($19 mN) is supported by the (uncut) Kapton frame.The force supported by the CNTY (1 mN) corresponds to $0.68% of the CNTY ultimate tensile force (144 mN), so it is deemed negligible for this analysis.During the test, the CNTY was first heated at a rate of 5 C/min from 30 to 300 C.Then, a temperature (T) dwell of 5 min at 300 C was maintained.
Finally, the specimens were cooled down at À5 C/min from 300 to 30 C. After the thermoresistive tests inside the DMA chamber, conventional DMA testing was performed by cutting the side legs of the Kapton frame, as discussed in the following section.

Electrical monitoring of the CNTY response during dynamic mechanical testing
DMA was carried out for two groups of CNTY specimens (Figure 2(c)), using a Perkin Elmer DMA 7 dynamic mechanical analyzer.The CNTYs of the first group were tested right after they had been characterized for thermoresistivity, as described in Section 2.6; in this sense, the thermoresistive test acted as an initial preheating cycle for subsequent DMA testing.The second group of CNTYs was tested using pristine CNTYs, that is, without being previously heated.In both cases, the Kapton frame was cut in the middle before running the DMA test.The CNTYs were tested under uniaxial tensile cyclic loading at 1 Hz with a static force of 18 mN and using 70% of the static force as the sinusoidal dynamic force.The test was carried out by performing temperature scans from 30 to 300 C, with a heating rate of 5 C/min.The electrical response of the CNTYs was simultaneously (in situ) measured by the four-point probe method using a Keysight 34465 A digital multimeter.

In situ Raman spectroscopy during tensile testing
The Raman spectrum of CNTYs preserves the basic features of the CNTs comprising the yarn, although a few bands may be slightly shifted due to CNT pre-stressing within the yarn.A typical Raman spectrum of the CNTYs investigated is shown in Section S2 of the supplementary information.In our case, the D band (located at $1350 cm À1 ) was the only Raman band that showed a statistically significant correlation (a shift in its peak position) with the applied strain (e).Thus, only shifting in the D band with strain will be discussed herein.One of the key features of this detailed Raman analysis is to distinguish the uncertainty in the band position due to the natural variability of the CNTY, from changes in the D band position due to strain.Averaging eleven Raman spectra from the line mapping turned out to be a great tool to capture more spatially broad information and assess the uncertainty in this task; such results comprise the scattering bars in Figure 4.The D band peak position of the CNTY as a function of the axial strain applied to the yarn (e) is shown in Figure 5.At small strains (e 0.5%) the shift in the D band position with strain cannot be distinguished from the experimental scattering of the experiment.This implies that below e ¼ 0.5% the CNTY exhibits mainly (or only) structural changes of its constitutive fibrils and bundles, and C-C bond stretching is negligible.As shown in Figure 5, above this level of strain, the CNTs that make up the CNTY suffer C-C bond stretching upon tensile loading, which is seen by a blue shift in the D band with strain.For e > 0.5%, the peak position of the D band is shifted towards lower wavenumbers as the strain increases.This change in the peak position of the D band is approximately linear, with a Raman shift strain factor (sensitivity) obtained from the slope of the linear fit of À0.30 cm À1 /%.This value of Raman shift with strain factor is very small.It is at least one order of magnitude smaller than the one observed for other carbon-based materials such as carbon fibers ($18.8 cm À1 /% for the 2660 cm À1 band (Gamstedt et al., 2002)), aramid (Kevlar) fibers ($4.85 cm À1 /% for the 1610 cm À1 band (Day and Cauich-Rodrigez, 1998)), and graphene deposited on a flexible substrate ($64 cm À1 /% for the 2D band (Mohiuddin et al., 2009)).This means that, because of the twist and hierarchical architecture of the yarn, C-C bond stretching is limited as a mechanism of deformation and strain transfer, and most of the strain energy is dissipated by the relative motion between the fibrils and bundles.In this sense, Raman spectroscopy suggests that tensile loading of the CNTY is strongly governed by structural changes of the fibrils/bundles comprising the yarn, and this information is of great assistance in understanding the mechanics of the yarn.
It has been argued that, when a CNTY is subjected to tensile loading, the CNT bundles exhibit slippage and stretching which generates a reduction of the diameter and untwisting (Jung et al., 2015;Abu-Obaid et al., 2015;Gao et al., 2018).As the strain increases, slippage among the constitutive CNT bundles and CNTs increases (Jung et al., 2015;Abu-Obaid et al., 2015).Since more energy is required to deform or break the C-C bond than to break the secondary bonds (van der Waals, dipolar, hydrogen bonding) between the CNTs, the secondary bonds break earlier, rearranging the structure of the CNTYs.Most CNTs are grouped in bundles and fibrils, which behave as a single structure in the CNTY, in analogy to the behavior of a single filament in a conventional yarn.This hinders C-C bond stretching of individual CNTs, since the strain is first transferred from the CNTY to the CNT fibrils, then to bundles, and then to CNTs.All this strain transfer process is conducted through the secondary bonds (Miao et al., 2020;Miao, 2013;Jung et al., 2018).For this reason, the G band shift that typically occurs when a CNT is subjected to tension (Kumar and Cronin, 2007) is not observed herein.Despite this, there is a small shift of the D band with strain, which indicates that C-C bonds do stretch, but not much.The D band can be affected by the modifications on the CNT sidewalls due to the introduction of defects and the attachment of different chemical species on the CNTs and also can be affected by the loss of symmetry in the crystalline structure of a material (Saito et al., 2011;Dresselhaus et al., 2005).In this case, the shift of the D band may be caused by changes in the crystalline structure of the CNTs, due to the loss of symmetry introduced by strain.The stress relaxation of the CNTY due to the relatively long measurement times between each deformation ($20 min) could also affect the peak position of the D band.

Electromechanical response under tensile loading
Before tensile testing, the specific electrical conductivity of the CNTYs (n=q) at zero strain and 25 C was measured from the tensile coupons, using the distance between electrodes (L g ¼ 25 mm, see Figure 2(b)), yielding 1211 ± 36.9 SÁcm 2 /g.The electrical conductivity of the CNTYs is governed by the conductivity of the individual CNTs as well as the degree of electrical contact (by physical contact or by tunnel effect due to their proximity) between the CNTs and among all hierarchical entities comprising the yarn (Miao et al., 2020;Obitayo and Liu, 2012).
The piezoresistive response of the CNTYs subjected to monotonic tensile loading is shown in Figure 6.The r=q vs. e response (solid circles) starts with a small linear zone and approximately around e ¼ 0.5% presents some degree of nonlinearity.This strain coincides with the value of strain where important shifts in the Raman D band are observed in Figure 5, indicating a change in the loading mechanism and such a strain.This means that more efficient load transfer between the bundles occurs after e ¼ 0.5%.Due to the inherent structural variability of the CNTY, there was some variability in the ultimate (failure) strain (e ult ) among the five replicates tested, reaching values between 2.4% and 6.3%.The CNTY specimens also showed an ultimate (failure) tensile load (F ult ) between 178 mN and 259 mN, corresponding to a specific tensile strength (r ult =q) between 0.84 N/tex and 1.21 N/tex.Table 2 summarizes the electromechanical properties of the CNTYs, including the thermoresistive properties that will be discussed in the following sections.In the literature, the mechanical response of the CNTYs for small strains (ca.e < 1%), has been attributed to the straightening and untwisting of the CNT bundles (Jung et al., 2015;Abu-Obaid et al., 2015).This is confirmed by the Raman spectrum of Figure 5 herein, where below e ¼ 0.5% the D band exhibit negligible peak shifting.The specific tensile modulus (E=q) of the CNTYs is 26.6 ± 12.8 N/tex, which was calculated herein in the 0.2% e 1% range.The mechanical response for e > 0.5% has been mainly attributed to slippage of the CNTs and bundles, as a result of breaking the van der Waals and possibly dipolar bonds (Jung et al., 2015;Abu-Obaid et al., 2015).The in situ Raman experiments of Figure 5 indicate that for this 'large strain' range, there are also some changes in the symmetry and stretching of the C-C bond, but this is small.
Regarding the piezoresistive response observed in Figure 6 (empty circles), it is observed that for e 1%, the fractional change of electrical resistance increases almost linearly with increased strain.The gage factor (GF) obtained in the range of 0.2% e 1% was 0.36 ± 0.13.For e > 1%, the rate of increase of the fractional change of electrical resistance (DR=R 0 ) increases rapidly and suddenly peaks close to failure.Therefore, it is expected that structural changes of the bundles and fibrils within the CNTY due to the applied strain play a paramount role in its piezoresistive response.Since CNTYs are hierarchical materials with a strongly propertystructure dependence, there is not a universal piezoresistive response for CNTYs.Anike et al. (Anike et al., 2019;Anike et al., 2017;Anike et al., 2018) hypothesized that there are two major physical phenomena governing the piezoresistive response of CNTYs.The first phenomenon occurs when the CNTY is stretched during loading, generating a decrease in the contact length of the CNT bundles, and therefore, an increase in its electrical resistance.The second phenomenon occurs when the CNTY relaxes during the loading segments, presenting inter-CNT and inter-CNT bundles slippage (inelastic shear motion), and therefore, a decrease in its electrical resistance.This last phenomenon causes negative piezoresistivity, which has also been observed for very low strain rates (Anike et al., 2018).In our case, the strain rate is 0.5 mm/min and the electrical resistance of the CNTYs always increases (positive piezoresistivity), which implies that the piezoresistive behavior of these CNTYs is governed by the first phenomenon.As discussed in Section 3.1, when a CNTY is subjected to tensile loading, their CNT bundles stretch and slide in direction of the applied load, which generates untwisting (decrease in twist angle), and therefore, reduction in diameter (Jung et al., 2015;Abu-Obaid et al., 2015;Gao et al., 2018).This decrease in the diameter of the CNTY implies a decrease in porosity since it increases the lateral contact between the CNT bundles, which decreases the electrical resistance of the CNTY.However, the sliding of the CNT bundles also decreases the length of the lateral contact between the CNT bundles, which increases the electrical resistance (Anike et al., 2019;Anike et al., 2017;Anike et al., 2018).Additionally, there is an increase in the distance between electrodes, which implies an increase in the electrical resistance.Herein, the phenomena that increase the electrical resistance with increased strain are those that govern the electromechanical response of the CNTYs under tension.These competing factors could also explain why the electrical response of the CNTYs is nonlinear, despite the mechanical behavior under tension is quite linear.

Cyclic tensile properties
The mechanical response of a representative CNTY subjected to cyclic tensile loading up to e ¼ 1% is shown in Figure 7(a).For increased clarity, the response of selected cycles (1, 100, 200, and 300) is shown in Figure 7(b).It is observed that as the number of cycles increases, the r=q vs. e curves move towards higher values of specific stress.Similarly, as the number of cycles increases, the separation between the loading and unloading curves decreases, that is, the area between loading and unloading curves (the hysteresis loop, H) decreases.Figure 7(c) shows that the residual specific stress (r Res =q) of the CNTYs presents some variations from cycle to cycle, but overall presents an increasing trend that stabilizes after cycle $230.When a CNTY is stretched beyond a certain strain, its structure changes irreversibly, and if this is repeated cyclically continuous structural changes are expected.The analysis of the in situ Raman spectra during tensile testing suggests that the deformation of the C-C bonds of the CNTs is very small or negligible for small strain levels (1% in this case).Thus, the residual specific stress is attributed to the structural rearrangement of the fibrils and CNT bundles within the CNTY.In addition, it is expected that the CNTY present residual stresses in their fibrils and bundles due to the twisting that is applied when they are spun.This explains why the residual value of r Res =q upon unloading to zero strain is different for each cycle.However, above cycle 230 the residual specific stress stabilizes around 0.06 N/tex.It is known that the CNTYs exhibit internal slippage and stretching of their fibrils and CNT bundles, as well as untwisting and reduction in diameter when they are subjected to tensile loading (Miao et al., 2020;Jung et al., 2015;Gao et al., 2018).As the CNTY is stretched, their fibrils slide and line up with the loading direction.However, unlike a monotonic tensile test where the stretching and alignment of the fibrils are continuous and constant, in the cyclic tensile test the rearrangement and alignment of the fibrils is discontinuous and not constant.This could lead to the oscillations in the residual specific stress observed in Fig. 7c.With each loading cycle, the fibrils and CNT bundles slip and rearrange, breaking some of the secondary bonds (van der Waals, dipolar, hydrogen bonding) between them, and forming new ones (Li and Kr€ oger, 2012;Xu et al., 2010;Li and Kr€ oger, 2012;Yang et al., 2011).On the other hand, with each unloading cycle, the fibrils that have kept their original secondary bonds during the loading cycle might return to their initial position, while the fibrils that formed new secondary bonds move to a new location or rearrange (Li and Kr€ oger, 2012;Xu et al., 2010;Li and Kr€ oger, 2012;Yang et al., 2011).The rate of occurrence of these phenomena has a degree of randomness and could be the reason for the oscillatory behavior between subsequent cycles in Figure 7c.Because of this reason, the normalized hysteresis parameter (H N ) shown in Figure 7d also presents some oscillations among subsequent cycles, but the decreasing trend with the cycle number is marked.H N starts at $26% and decays rapidly with each cycle, until leveling off at $7%, above cycle 230.During cycling tensile loading/unloading, the CNT network comprising the CNTY changes through detachingattaching and zipping/unzipping mechanisms (Zhao et al., 2018;Yang et al., 2011;Zhao et al., 2015), which may contribute to the energy dissipation observed in H N : The process of unzipping between CNTs consume energy to overcome the van der Waals bonds (Xu et al., 2010;Li and Kr€ oger, 2012), while the zipping process does not require energy consumption (Xu et al., 2010).Unlike the zipping/unzipping mechanism, the detaching-attaching mechanism irreversibly changes the CNT network morphology during each loading/unloading cycle (Li and Kr€ oger, 2012;Yang et al., 2011;Zhao et al., 2015).Thus, the decrease in H N with the number of cycles could be due to the decrease in the number of fibrils that are rearranging (irreversible changes) in the CNTY.Therefore, the fact that the hysteresis remains constant above cycle 230 suggests that there are no longer irreversible changes in the internal structure of the CNTY, leaving only the energy dissipation by friction.Further cycling experiments were conducted by increasing the strain applied after each cycle, and they are discussed in Section S1 of the supplementary information.The increase in H N observed in the CNTYs during the incremental strain cyclic tensile testing (see Figure S1 of the supplementary information) indicates that there is an increase of the irreversible changes in the internal structure (rearrangement of fibrils and CNT bundles) of the CNTY.This supports the hypothesis that the drop in H N in the cycles with constant deformation is due to the decrease in the irreversible changes in the internal structure of the CNTY, and the remaining hysteresis is due to the energy dissipation by friction.
In terms of the dissipated energy density (H=q), the CNTYs reach a value of 0.06 J/cm 3 during the first cycle (H ¼ 0.24 J/g; N/tex ¼ kJ/g), and above cycle 230, H=q remains constant around 0.03 J/cm 3 (H ¼ 0.12 J/g).Table 3 compares the dissipated energy per unit weight of the CNTYs to that of other materials.Such compassion suggests that CNTYs are materials with extraordinary energy dissipation capabilities, whose specific energy dissipation capability (per unit weight) is comparable to that of carbon steel.

Constrained thermoresistive response
The constrained (inside the DMA tensile test rig) thermoresistive response of a representative CNTY is shown in Figure 8.It is observed that the fractional change of electrical resistance (DR=R 0 ) of the CNTY decreases linearly with temperature (T) up to $100 C during heating.Around this temperature (100-160 C), there is a marked step upward in DR=R 0 towards positive values (which were present in all replicates tested) that continue until $160 C. The thermoresistive curve continues with a similar negative slope thereafter.According to the thermogravimetric analysis (TGA) reported in Section S3 of the supplementary information, the CNTY loses up to $13% in weight during heating from 30 to 400 C. Important mass losses ($5%) occur between 100 C and 160 C which is the temperature interval where a marked step in DR=R 0 is observed in Figure 8. Mass losses at temperatures below 400 C are attributed to the evaporation of the densifier (acetone) used in the CNTY synthesis and adsorbed moisture (Avil es et al., 2009), and probably to the degradation of amorphous carbon and other carbonaceous forms (Avil es et al., 2009;Sch€ onfelder et al., 2012;Santangelo et al., 2013).The evaporation of functional groups and byproducts (mass loss) of the CNTY likely causes the rearranging of their CNT bundles and changes in porosity, leading to a sudden change of the electrical resistance.A decrease in the electrical resistance of the CNTYs with temperature is a negative thermoresistive response, which has been previously observed for unconstrained CNTYs (Balam et al., 2020;Aliev et al., 2007).As temperature increases, the density and mobility of electric charge carriers also increase, resulting in a quasilinear drop in the electrical resistance of individual CNTs that make up the CNTY (Miao et al., 2020;Balam et al., 2020).This mechanism has also been proposed as the governing thermoresistive mechanism of pitch-based carbon fibers (Day and Cauich-Rodrigez, 1998;Abu-Obaid et al., 2015;Yang et al., 2011).In order to quantify the thermoresistive sensitivity of the CNTYs, the temperature coefficient of resistance (b i ) was obtained from the slope of the linear fit of the DR=R 0 vs. temperature changes (DT ¼ T À T 0 , where T 0 ¼ 30 C) curve in the three temperature intervals indicated in Figure 8.In the 0 DT 50 K range, the CNTYs present a value of b 1 ¼ À8.63 Â 10 À4 ± 0.79 Â 10 À4 K À1 .After the step in DR=R 0 , b 2 ¼ À7.78 Â 10 À4 ± 1.80 Â 10 À4 K À1 is measured in the 130 K DT 250 K interval.This means that the thermoresistive response is indeed very similar in the full temperature range examined, except in the interval between 100 C and 160 C, where the sudden step is observed.Additionally, the cooling curve exhibits a linear behavior throughout the entire range, with b 3 ¼ À7.26 Â 10 À4 ± 0.05 Â 10 À4 K À1 .The proximity of the numerical results for b 1 , b 2 and b 3 reinforces the observation of a quasi-linear thermoresistive behavior of the CNTY for the full temperature range, upon heating and cooling.The average values of the temperature coefficient of resistance, b 1 , b 2 and b 3 , are comparable (a bit smaller) to those reported for similar CNTYs under unconstrained (free) conditions, for example, À9.46 Â 10 À4 K À1 (Balam et al., 2020) and À12 Â 10 À4 K À1 (Dau et al., 2016).The difference may be ascribed to the boundary conditions (constrained or not) used in each experiment.

Electrical response during dynamic mechanical analysis
The coupled DMA and electrical response of representative pristine (Figure 9(a)) and preheated (Figure 9(b)) CNTYs during DMA testing are shown in Figure 9 as a function of temperature (T).From Figure 9(a), it is observed that the storage modulus (E 0 ) of the pristine CNTYs decreases up to $70% with increasing temperature from 30 to $170 C, while the loss modulus (E 00 ) decreases up to $80% with increasing temperature from 30 to $90 C. From 90 C, E 00 decreases more gradually and tends to level off.On the other hand, the loss modulus of the preheated CNTY (Figure 9b) remains almost constant with increasing temperature, while the storage modulus decreases up to $40% at 120 C.After $120 C, E 0 remains almost constant.E 0 of the pristine CNTY is $2.80 times higher than that of the preheated yarns at 30 C, but the pristine CNTYs broke at $170 C, while preheated CNTYs did not break during DMA testing.The explanation of the DMA response of the CNTY may be assisted by the TGA thermogram of Section S3 of the supplementary information.They indicated that, when they are not previously heated, the CNTYs  continuously lose mass during heating up to 350 C. As the temperature increases, the thermal transformation of the hydrocarbon functional groups and byproducts in the CNTY may further affect the structural changes at the bundle level and porosity of the yarn, which facilitates the CNT bundle relaxation through sliding (Zhao et al., 2015).This may explain why the preheated CNTYs have a lower E 0 at the beginning of the DMA.At the same time, the loading/ unloading cycles gradually change the structure of the CNTY (twist angle, porosity, diameter, etc.), and this dynamic/friction effect convoluted with the thermal degradation of synthesis byproducts may yield the faster decrease in E 0 during the low-temperature range of the DMA test.
The analysis of tensile hysteresis (see Section 3.3) and the in situ Raman spectroscopy during tensile testing showed that the CNTYs undergo irreversible structural changes during the loading/unloading cycles.In this sense, the thermal degradation of hydrocarbons could be contributing to an increase in the hysteresis of the CNTYs, and therefore, causing its breakdown at lower temperatures.
During DMA testing, the fractional change of electrical resistance (DR=R 0 ) of the CNTY decreases as the temperature increases, following the trend of E 0 (see Figure 9(a)).Notice that DR=R 0 exhibits small oscillatory variations in the readings, which were observed in all DMA tests but not observed in the thermoresistive response (see Figure 8).Thus, it is believed that these oscillations in DR=R 0 are related to experimental noise associated with the oscillatory character of the DMA test, and the large electrical sensitivity of the CNTY (see Figure 6).However, further dedicated tests indicated that the frequency of oscillations of DR=R 0 does not match the DMA frequency.
To assess the contribution of thermoresistivity to the electrical response of the CNTY during DMA testing, the temperature coefficients of resistance were also obtained from the DR=R 0 vs. DT curves of the DMA in Figure 9.The temperature coefficient of resistance of the pristine CNTYs obtained during DMA testing in the 0 DT 50 K range, was b DMA Pri ¼ À9.83 Â 10 À4 ± 3.25 Â 10 À4 K À1 .The temperature coefficient of resistance of the preheated CNTYs during DMA testing was obtained in the full temperature range as b DMA Ph ¼ À6.92 Â 10 À4 ± 0.55 Â 10 À4 K À1 .Both are numerically similar to b 1 (-8.63Â 10 À4 ± 0.79 Â 10 À4 K À1 ) and b 3 (-7.26Â 10 À4 ± 0.05 Â 10 À4 K À1 ) calculated from the thermoresistive response of Figure 8.This fact indicates that the decrease in the electrical response of CNTY during DMA testing is governed by its negative thermoresistivity, with small oscillations in DR=R 0 superimposed given the oscillatory nature of the DMA test.

Conclusions
Carbon nanotube yarns (CNTYs) are smart hierarchical materials with outstanding specific mechanical properties and very high energy dissipation capabilities.Through Raman spectroscopy, it was proved that the tensile loading mechanisms of CNTYs are governed by structural changes of their fibrils and bundles, rather than stretching of the chemical bonds between the carbon atoms of the CNTs.The contribution of material (C-C) deformation to the stretching of the CNTY only arises above 0.5% strain, evidenced by a small Raman shift strain factor of À0.30 cm À1 /%.This structural behavior yields a material with very high energy-dissipation capabilities, which are reflected in its tensile hysteretic response under cycling loading and high loss modulus in DMA testing.The normalized hysteresis of these CNTYs decreases with each cycle until leveling off around cycle 230.This indicates that, in addition to the energy dissipated by friction, there is a contribution of the dissipated energy due to irreversible changes in the structure of the CNTYs.The specific (per unit weight) dissipated energy of these CNTYs is comparable to that of carbon steels.An average piezoresistive gage factor of 0.36 was measured for the CNTYs.The thermoresistive response under end-clamped conditions (i.e., mechanically constrained) was nearly linear and similar for heating and cooling, with an average coefficient of thermoresistive sensitivity of À7.26 Â 10 À4 K À1 .Both, the thermoresistive response and the DMA temperature scans are affected by the initial condition of the CNTY.This means that the thermal decomposition of hydrocarbons and possibly other volatile species remaining from the synthesis may affect the DMA thermal and thermoresistive responses.The electrical response of the CNTYs during DMA testing is governed by their thermoresistive response.The fact that CNTYs have an electrical response that is concurrently sensitive to monotonic and dynamic strain/stress, temperature, and evaporation of volatile byproducts, provides further opportunities for the development of new multifunctional smart materials.

Figure 2
Figure 2 shows CNTY specimens for uniaxial cyclic tensile testing (Figure 2(a)), electromechanical characterization (Figure 2(b)), constrained thermoresistive characterization (Figure 2(c)) and DMA (Figure 2(c)).The CNTYs were prestretched with a very small load ($0.73 mN), in order to minimize variability in the initial state of the CNTY.A 75 mg mass (fragment of a needle) was attached to one end of the yarn and subsequently bonded onto a cardboard or polyimide (Kapton) frame with a rectangular internal window, as shown in Figure2.The ends of the CNTYs were fixed to the frame using a cyanoacrylate adhesive and the

Figure 2 .
Figure 2. CNTY specimens.Arrows indicate the direction of the applied load (F).(a) Specimen for uniaxial cyclic tensile tests, (b) specimen for electromechanical tests, and (c) specimen for constrained thermoresistive characterization and DMA.

Figure 3 .
Figure 3. Schematic of the test set up for in situ Raman spectroscopy during tensile testing.

Figure 4 .
Figure 4. Schematic of the hysteresis parameters used for the uniaxial cyclic tensile test.

Figure 5 .
Figure 5. D band peak position of the CNTYs as a function of strain (e).

Figure 6 .
Figure 6.Electromechanical response of a representative CNTY under uniaxial tensile loading.

Figure 7 .
Figure 7. Mechanical hysteresis of a representative CNTY under cyclic tensile loading up to e ¼ 1%.(a) Tensile response of all 300 cycles, (b) tensile response of selected cycles, (c) residual specific stress (r Res =q), and (d) normalized hysteresis (H N ).

Table 1 .
Summary of the morphological and structural properties of the CNTYs.