Polylactic Acid/Carbon Nanoparticle Composite Filaments for Sensing

Polylactic acid (PLA) is a bio-based, biodegradable polymer that presents high potential for biomedical and sensing applications. Ongoing works reported in the literature concern mainly applications based on 3D printing, while textile applications are hindered by the limited flexibility of PLA and its composite filaments. In the present work, PLA/multiwall carbon nanotube (MWCNT) composite filaments were produced with enhanced flexibility and electrical conductivity, which may be applied on a textile structure. A biodegradable plasticizer was incorporated in the nanocomposites, aiming at improving MWCNT dispersion and increasing the flexibility of the filaments. Filaments were produced with a range of compositions and their morphology was characterized as well as their thermal, thermomechanical, and electrical properties. Selected compositions were tested for sensing activity using saturated acetone vapor, demonstrating a suitable response and potential for the application in fabrics with sensing capacity.


Introduction
Poly(lactic acid) (PLA) is a biocompatible, biodegradable, melt-processable polymer with good mechanical properties, wicking, breathability, and ability for adsorption of organic compounds, which has been extensively researched as a replacement for noncompostable polymers. PLA and PLA composites have been developed for applications in garment, medical textiles, and disposables [1,2]. Other potential application areas include wearable devices, patch sensors for motion [3], 3D-printed lithium-ion batteries [4], flexible pressure sensors [5], functional coatings for textiles [6], and piezoresistive sensors [7]. The brittleness and low mechanical properties of PLA are, however, a limitation for application in wearables, motivating research on plasticization [8] and reinforcement with nanoparticles [9] and seeking enhanced mechanical performance and flexibility.
A plasticizer for PLA should be nontoxic, biocompatible, and should present good miscibility with the polymer, leading to a decrease in the glass transition temperature (T g ) [10]. A major problem concerns the immiscibility and tendency for migration of plasticizers to the surface of PLA with consequent loss. This effect is most important for low molecular weight (M W ) plasticizers such as lactide monomer and ester-like plasticizers and thus the preference for higher M W plasticizers. Poly(ethylene) glycol (PEG) and  [15] Triacetin Flax fiber Melt extrusion [21] Poly(butylene adipate-co-terephthalate) Rami fiber Melt extrusion and compression molding [22] Epoxidized linseed oil Hazelnut shell flour Melt extrusion [16] PEG (M W 6000 g·mol −1 ) CaCO 3 Halloysite nanotubes LAK 301 Micro-sized talc Melt extrusion [23] Glycerol Bamboo (different forms) Batch mixing [24] PEG (M W 1500 g·mol −1 ) ZnO Clove essential oil Solvent cast [25] Acetyl triethyl citrate Montmorillonite layered silicate Melt extrusion [26] Diglycerine tetraacetate and PEG (M W 1000 g·mol −1 )

Sodium montmorillonite and organoclays
Milling and injection molding [27] PEG (M W 6000 g·mol − [33] PLA nanocomposites produced with carbon nanotubes (CNT) showed improved dispersion of CNT upon addition of a plasticizer, enabling the preparation of electrically conductive, flexible multifunctional nanostructured composites with lower T g compared to PLA, finding applications as shape-memory polymers [14,15].
In the present work, we report the melt extrusion of PLA with multiwall CNT (MWCNT), comparing composites with the neat polymer and PPG as a plasticizer. The influence of PPG on the dispersion of MWCNT in PLA and its plasticizing effect is studied. Electrical conductivity and sensing ability under acetone vapor are studied for composites with different MWCNT content. Filaments were produced with different diameters foreseeing an application in textile wearable devices.

Coating of PLA Pellets
Semicrystalline PLA pellets (Luminy ® L175, Total Corbion PLA, 4203 NS Gorinchem, The Netherlands) were coated with MWCNT dispersions in PPG by stirring with a common kitchen mixer until EtOH evaporation. Coated pellets were further dried for 2 h at 60 • C and then for 8 h at 100 • C prior to melt extrusion.

Melt Extrusion of PLA/MWCNT/PPG Composites
Melt compounding of PLA nanocomposites was carried out in a co-rotating intermeshing twin-screw extruder (Rondol Microlab, France) with 10 mm diameter screws composed of three mixing sections ( Figure 1) and a length-to-diameter ratio (L/D) of 25. A temperature profile was defined for the feeding zone (FZ), the three different zones in the barrel (Z1, Z2, and Z3) and die, as follows: 120 • C/160 • C/175 • C/180 • C/175 • C. A volumetric feeder (Piovan MDP1, Italy) supplied coated and noncoated PLA pellets at FZ. For composites without prior coating with PPG and MWCNT, MWCNTs were added manually at FZ. The pellets were dried for 2 h at 60 • C and then for 8 h at 100 • C before processing. The extruded rod presented a diameter of approximately 1.5 mm after the die exit. A filament puller with controlled speed was used to collect the extruded filament and set the diameter at approximately 0.5 mm. Filaments were collected with larger diameter (collected directly from the die exit) and with smaller diameter (collected with diameter control). Extrusion layout: the screws are composed of an initial transport zone at feeding zone (FZ), followed by the first set of mixing blocks at the beginning of Z1, a transport zone, plus an intensive mixing zone at Z2; at Z3, there is a small mixing zone in between transport ones.  1. Extrusion layout: the screws are composed of an initial transport zone at feeding zone (FZ), followed by the first set of mixing blocks at the beginning of Z1, a transport zone, plus an intensive mixing zone at Z2; at Z3, there is a small mixing zone in between transport ones.
The compositions prepared are listed in Table 2. Optical microscopy (OM) images were obtained with a Leica DM2500 PLA microscope equipped with a digital camera Leica DMC2900. Transversal sections of 1.5-2 µm thickness were cut with an ultramicrotome (Leica UC6). LAS X software was used to record the images and ImageJ software to analyze total and particle areas. Contrast differences were employed and the agglomerate areas larger than 5 µm 2 were measured. The analysis was performed on micrographs collected from at least three different sections per composite.

Scanning Electron Microscopy
Scanning electron microscopy (SEM) images were obtained with a Nano SEM-FEI Nova 200 (FEG/SEM) (FEI Europe Company, Hillsboro, OR, USA) at high vacuum (resolution of 1.0 nm at 15 kV). Samples were cryofractured in liquid nitrogen, and the resulting surfaces were coated with Au/Pd to observe the transversal section of filaments. The images were obtained in secondary electron mode.
Differential scanning calorimetry (DSC) scans were performed in a Netzsch DSC 200 F3 Maia ® (Netzsch-Gerätebau GmbH, Selb, Germany), equipped with a cooling system Netzsch Intracooler 40 (Netzsch-Gerätebau GmbH, Selb, Germany) at a heating rate of 10 • C·min −1 , from −10 • C to 190 • C, under nitrogen atmosphere. An error of 0.5 • C is associated with all temperature measurements. Peak areas and relevant transition temperatures were determined with Netzsch Proteus ® software.

Raman Spectroscopy
Raman spectra were obtained using a HORIBA LabRAM HR Evolution confocal Raman microscope spectrometer (Horiba Scientific, Piscataway, NJ, USA) using Horiba Scientific's Labspec 6 Spectroscopy Suite Software for instrument control, data acquisition, and processing. The Raman spectrometer was equipped with a 532 nm laser, a Syncerity CCD Camera detector (Horiba) and a microscope equipped with a 10× and 100× magnifi-cation objectives. A grating of 600 gr·mm −1 was used. For each composite, three spectra were collected from distinct zones of the filament and averaged, and peak positions were determined by applying a baseline (in LabSpec 6) and fitting with a Lorentzian function (in OriginPro 2016).

Dynamic Mechanical Analysis
Dynamic mechanical analysis (DMA) was carried out in a TT DMA (Triton Technology, Grantham, UK) equipped with a 1L Dewar and a TTDMA AutoCryo Type 2 for cooling with liquid nitrogen. The analysis was performed in tension mode with a 10 mm gauge length. The measurements were carried out at a frequency of 1 Hz with a displacement amplitude of 0.001 mm, from −50 • C to 120 • C at a heating rate of 2 • C·min −1 . Storage and loss moduli (E and E", respectively) reported at 0 • C were obtained by computing a linear fit in the temperature range from −40 C to 10 • C.

Resistivity Measurements
Electrical resistivity measurements were carried out on a Keithley SMU 2635B SourceMeter ® (Keithley Instruments Inc., Cleveland, OH, USA). Intensity/voltage (I/V) curves were acquired under direct current (DC) on a source range of 20 V (from −10 V to 10 V with 0.5 V steps) at room temperature. A linear fit was applied, in Microsoft Excel, to calculate the resistance (R) as the inverse of the I/V curve slope. Bulk conductivity (σ DC ) was calculated according to Equation (1), where L is the length of the conductive path (distance between electrodes) and A is the sample cross-section area.

Vapor Sensing Measurements
The vapor sensing ability of the composite filaments was investigated by in situ monitoring of their electrical resistance response upon alternating exposure to acetone vapor and dry air for four successive cycles. The exposure time to organic vapor and the drying time in dry air were 100 s and 150 s, respectively. Prior to the exposure to the saturated acetone vapor, the filaments were set in the measuring chamber under dry-air flow for 50 s to reach a stable initial resistance (R i ). The filaments for the vapor sensing test were coated with silver ink at the end sides to ensure good contact between the specimen and the electrodes. All samples were tested at 25 • C. The set-up of the sensing device is a self-made equipment described in previous works [34]. The sample resistance R was measured using a Keithley 2001 digital multimeter (Keithley Instruments Inc., Cleveland, OH, USA), and the experimental results are presented as relative resistance (R rel ). R rel is the ratio of the resistance variation (∆R = R − R i ) to the initial resistance. A bubbler evaporation system was used to deliver the desired organic vapor to the detection chamber using dry air as carrier. In the drying run, pure dry air was passed through the chamber. The total gas flow rate was set to 30 L·h −1 using a mass flow controller.

Morphology of Melt-Extruded Filaments
OM images of the composite filament cross sections with different loads of MWCNT, with and without PPG, are displayed in Figure 2. OM analysis provides a 2D description of the MWCNT fraction that did not disperse in the polymer during melt mixing, remaining as agglomerates. The micrographs show that filaments with 3 wt.% MWCNT present a larger number of agglomerates compared to 2 wt.% MWCNT. They also illustrate the effect of the predispersion of MWCNT in PPG, resulting in improved dispersion of MWCNT with consequent decrease in the number of agglomerates present as well as formation of smaller agglomerates.   Statistical analysis confirms that melt-extruded filaments of neat PLA and MWCNT present agglomerates with larger area (Table 3) with medians ranging from 18 to 23 µm 2 . Nanocomposites incorporating PPG present agglomerates with smaller median areas, within a narrow range from 6 to 7 µm 2 . OM images depict prominent agglomerates of MWCNT for composites without PPG and smaller MWCNT agglomerates for composites prepared with PPG. Density plots (Figure 3) and the cumulative distribution ( Figure S1) of agglomerate areas show that composites with PPG present smaller agglomerates with a narrower area range (75% of agglomerates < 10 µm 2 , Table 3), whereas composites without PPG present a wider dispersion in agglomerate areas. A Kruskal-Wallis statistic for alpha = 5% shows that both the addition of PPG (for H (1) = 5.39, p-value = 0.02) and MWCNT concentration (for H (1) = 7.79, p-value = 0.005) have a significant effect in the agglomerate area ratio. It may be concluded that (i) the agglomerate area ratio is significantly lower for composites containing PPG and that (ii) increasing MWCNT concentration increases the agglomerates area ratio. The latter conclusion is expected and reported in Table 3. Comparison of filaments with similar composition shows that the composites without PPG display the highest agglomerate area ratio values, with a larger area covered by MWCNT agglomerates.

Raman Spectroscopy
Raman spectra of neat-PLA were recorded and averaged, and the spectrum is presented in the supplementary materials ( Figure S2) with highlighted bands corresponding to the vibrational modes of each bond. Bands of the vibrational modes of C = O, C-O-C, and C-H bonds are common to the Raman spectra of all composites. The asymmetrical stretch of CH 3 bond, with a Raman shift~1450 cm −1 , is observed between the D and G bands of MWCNT (~1338 cm −1 and~1577 cm −1 , respectively) as presented in Figure 5. The slight upshift of D, G, and 2D (~2672 cm −1 ) bands for the composites relative to the MWCNT in powder form ( Figure 5) may be attributed to the dispersion and disentanglement of MWCNT and a slight compression effect of the surrounding polymer [41,42].

Thermal Characterization
The thermal characterization of the extruded filaments was performed by TGA and DSC analysis.
Thermal degradation curves, and the corresponding derivatives (DTG), of PLA and composites are presented in Figure 6. Neat PLA displays one degradation step with first derivative peak maximum occurring at 394 • C. Two weight loss steps are observed for PLA/20PPG: one with a first derivative peak maximum at 255 • C that may be attributed to PPG vaporization and the second near 396 • C, corresponding to PLA degradation ( Figure 6, Table 4).  The weight losses observed for the composites containing PPG in the temperature range of 200-350 • C are approximately 10 wt.% (Table 4), while the PLA/PPG composition shows a weight loss near 15 wt.% in the same temperature range. Considering this weight loss an indicator of the PPG content, it remains below the target composition of 20 wt.%, suggesting loss by vaporization during melt-extrusion.
A decrease in T g and melting temperature (T M ) is observed for PLA/20PPG as compared to neat PLA, confirming the plasticizing effect of PPG (Figure 7). Composites without plasticizer show T M and T g similar to PLA (Table 5, Figure 7). The plasticizing effect of PPG is also noticeable in MWCNT composites, which present lower T g and T M than the corresponding nonplasticized compositions.  After melt processing into filament form, the neat PLA shows a pronounced coldcrystallization peak in the DSC crystallization scan (Figure 7), i.e., upon heating, the PLA filaments can still rearrange and crystallize further when temperature rises above T g . Compositions containing PPG appear more stable upon heating, showing lower tendency for molecular rearrangement and depicting smaller cold crystallization enthalpy (∆H cc ). This observation is quite relevant, as it may indicate higher dimensional stability of the filaments prepared with plasticizer, and their possible application at higher service temperatures. The PLA and composite filaments containing PPG present a slight increase of the total melting enthalpy (∆H M ), thus indicating that the PLA crystallization ability was not hampered by the presence of PPG.

Dynamic Mechanical Properties
DMA analysis shows the effect of the addition of high modulus MWCNT to PLA with and without PPG. While the addition of MWCNT increases the composite storage modulus (E ), the plasticizer has the opposite effect, decreasing E and increasing the loss modulus (E") for PLA and its composites. The composites with MWCNT and PPG present lower E compared to neat PLA but higher than plasticized PLA, showing the relevance of the PPG plasticizing effect. The smaller-diameter nonplasticized composite filaments show increasing E with MWCNT load. The addition of PPG lowers the value of E continuously from −40 • C (the melting temperature of PPG) until approximately 40 • C, while the nonplasticized composites maintain their E value along the whole temperature range until reaching the T g of PLA (Figure 8). The plasticizing effect of PPG in the composite filaments is confirmed by the increase in E", therefore showing higher ability for energy dissipation compared to the corresponding nonplasticized composites (Figure 8e). The peak maximum of Tanδ (Table 6) indicates the temperature at which significant segmental motions of the polymer chains are enabled, corresponding to the polymer T g . Tan δ is observed at lower temperature for PLA and composites containing PPG compared to neat PLA and its composites (Table 5), consistent with the results obtained by DSC. Table 6. E , E", and tan δ peak for neat PLA and composites.

Sample
At 0 • C Temperature of tan δ Peak ( • C)

Conductivity and Sensing Ability of Composite Filaments
The results obtained for the DC volume conductivity (σ DC ) of the PLA composite filaments are presented in Figure 9. Measurements were performed on the small-diameter (~0.5 mm) composite filaments with and without PPG, and on the large-diameter filaments (~1.5 mm) with PPG, to monitor possible effects that may be induced during filament collection, which may affect the MWCNT distribution in the composite. All composite filaments produced without plasticizer show σDC in the order of 10 −7 S·cm −1 . The composite filaments with 3 wt.% MWCNT and PPG presented a DC in the order of 10 −3 S·cm −1 for both small and large diameter, and similar results were obtained for large-diameter 2 wt.% MWCNT/PPG filaments. The plasticizer played a major role aiding the dispersion of MWCNT, as observed by OM, and the enhanced MWCNT dispersion allowed the formation of electrically conductive filaments. However, the smalldiameter filaments with 2 wt.% MWCNT/PPG yielded lower DC, which may be due to the filament preparation procedure, namely the pulling action imposed on the melt dur- All composite filaments produced without plasticizer show σ DC in the order of 10 −7 S·cm −1 . The composite filaments with 3 wt.% MWCNT and PPG presented a σ DC in the order of 10 −3 S·cm −1 for both small and large diameter, and similar results were obtained for large-diameter 2 wt.% MWCNT/PPG filaments. The plasticizer played a major role aiding the dispersion of MWCNT, as observed by OM, and the enhanced MWCNT dispersion allowed the formation of electrically conductive filaments. However, the smalldiameter filaments with 2 wt.% MWCNT/PPG yielded lower σ DC , which may be due to the filament preparation procedure, namely the pulling action imposed on the melt during filament forming. This slight pulling of the polymer melt may induce an orientation effect, leading to a decrease of the electrical conductivity [43]. This marginal pulling effect was consequent for the electrical conductivity of the composites with lower MWCNT content but not for the filaments with higher MWCNT concentration. The latter presented similar average electrical conductivity for large-and small-diameter filaments, despite increasing its scatter along the filament's length.
The cyclic sensing response of PLA composite filaments was investigated using saturated acetone vapor. As observed in Figure 10, small-and large-diameter samples react to the immersion in acetone vapor by an increase in the relative resistance R rel and to the rinsing with dry air with a resistance decrease. Thereby, large differences are observed depending on the concentration of MWCNTs and the addition of PPG. The small-diameter composite filaments with 3 wt.% MWCNT without PPG shows the lowest response with R rel values of approximately 15% after the 150 s immersion time. Addition of PPG significantly increases R rel values up to about 150% in the large-diameter filaments and up to 3700% in the small-diameter filaments. Filaments with 2 wt.% MWCNT and PPG yielded lower signal than filaments with 3 wt.% MWCNT, with maximal R rel near 125% for the thick and near 250% for the thin filaments. Thereby, all the large-diameter filaments present lower signal noise but also a much lower reversibility in the drying process, retaining high R rel values even after 100 s exposure to dry air. The smaller-diameter filaments show higher signal noise, despite higher reversibility within the drying period with a small shift of the starting R rel values after each cycle. The small-diameter filaments with 3 wt.% MWCNT and 20 wt.% PPG are the best performing, showing high response and good reversibility. This demonstrates the suitability of these composite filaments for integration in fabrics for gas-sensing applications.

Conclusions
The present work describes the preparation and characterization of composite filaments based on PLA and MWCNT, and the evaluation of the effect of the addition of PPG as a plasticizer. The addition of PPG was observed to assist the dispersion of MWCNT within the PLA matrix, reducing the size and number of MWCNT agglomerates and thus increasing the electrical conductivity of the composites. The plasticized composites were thermally stable up to approximately 200 C. The plasticizing effect was confirmed by

Conclusions
The present work describes the preparation and characterization of composite filaments based on PLA and MWCNT, and the evaluation of the effect of the addition of PPG as a plasticizer. The addition of PPG was observed to assist the dispersion of MWCNT within the PLA matrix, reducing the size and number of MWCNT agglomerates and thus increasing the electrical conductivity of the composites. The plasticized composites were thermally stable up to approximately 200 • C. The plasticizing effect was confirmed by DSC and DMA, showing a consistent decrease by nearly 15 • C of the glass transition temperature for plasticized PLA and its composites. The tendency for cold crystallization of PLA observed above T g was considerably reduced by the incorporation of the plasticizer. DMA results showed larger E" values for composites containing PPG, demonstrating enhanced ability for energy dispersion as compared to neat PLA and its MWCNT composites.
The addition of PPG to the composites and the diameter of the filaments produced were observed to influence the sensing response toward acetone vapor. The electrical conductivity and sensing ability achieved for the smaller-diameter filaments (∅~500 µm) are promising; however, their textile application is so far limited to simple textile structures, considering the large filament diameter. The next steps include the production of composite filaments with smaller diameter and their characterization, aiming at the production of a wearable functional textile or textiles for leakage detection.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.