Single and bundled carbon nanofibers as ultralightweight and flexible piezoresistive sensors

The study addresses the need for ultrathin, flexible, and wearable sensors capable of static and dynamic strain sensing. Traditional piezoelectric materials (e.g., PZT) are brittle and respond primarily to dynamic stimuli, limiting their use in static strain applications. While flexible piezoelectric polymers like PVDF have been widely used, piezoresistive flexible sensors based on conductive nanomaterial–polymer composites are emerging as promising alternatives. Carbon-based nanomaterials (graphene, CNTs, CNFs, carbon black) embedded in elastomers (e.g., PDMS, PU, Ecoflex) have been explored, but many approaches involve fragile structures or complex fabrication. Electrospun PAN-derived CNFs are attractive due to their flexibility, low weight, and potential for high piezoresistive sensitivity. However, the onset and mechanisms of piezoresistivity in electrospun CNFs, and the influence of pyrolyzation temperature on their morphology, structure, and electrical behavior, remain insufficiently understood. This work systematically investigates how pyrolyzation temperature (500–950 °C) affects CNF properties, elucidates the piezoresistive mechanism via a conductive graphitic domain discontinuity model, and demonstrates ultralight, flexible sensors using both CNF bundles and single CNFs for strain and airflow sensing.

• Prior wearable/flexible sensors have used piezoelectric materials (PZT, BaTiO3, ZnO), but issues include brittleness and lead content (for PZT). Flexible piezoelectric polymers (PVDF, PVDF-TrFE) enable dynamic sensing and energy harvesting but are not suited for static strain sensing.
• Piezoresistive sensors using nanocarbon–elastomer composites (graphene, CNTs, CNFs, carbon black with PDMS, PI, Ecoflex, PU) have shown promise, though often requiring complex processing or yielding fragile structures (e.g., graphene foams).
• Electrospun PAN-derived CNFs have been mechanically and structurally characterized (Zussman et al., 2005) and individual electrospun CNFs showed modest gauge factors (1.96–2.55; Cai et al., 2014), but their deployment in flexible sensors is underexplored.
• The conductive-domain/tunneling model has been proposed to explain CNF conductivity via sp2 graphitic domains within an sp3 amorphous matrix, where inter-domain tunneling is strain-sensitive (Cai et al., 2014).

Materials and electrospinning:

• PAN (MW 150,000 g/mol) dissolved in DMF to 9% (w/v); stirred at 30 °C for 120 min to obtain a homogeneous solution.
• Electrospinning with Inovenso setup using a stainless-steel rotating mandrel collector (10 cm diameter) covered with Al foil; spindle speed 1500 rpm.
• Syringe pump with 10 mL syringe, 18G needle; flow rate 1 mL/h; DC voltage 12 kV; tip-to-collector distance 15 cm; ambient 22 °C; spin time 30 min to obtain uniform nanofiber film.
• Aligned nanofiber bundles collected on Al foil; isolated single nanofibers electrospun across micro-trenches on MEMS substrates.

Stabilization and pyrolyzation:

• Transfer nanofiber bundles to ceramic crucible; stabilization in air at 245 °C for 60 min (ramp up/down 5 °C/min) to form oxide layer and induce cyclization/oxidation.
• Nitrogen purge for 30 min, then pyrolyzation in N2 at target temperatures: 500, 600, 700, 800, 950 °C; ramp 5 °C/min to setpoint, hold 1 h, then ramp down at 5 °C/min.

Characterization techniques:

• Morphology: FE-SEM (FEI Nova NanoSEM 230). Nanofiber diameter measured before and after pyrolysis; shrinkage quantified.
• Raman spectroscopy: Renishaw inVia (532 nm laser, 1800 l/mm grating), high confocal mode (spot ~0.5 µm), 1% of 34.1 mW power, static mode, averaged over 100 accumulations. D and G bands quantified; I_D/I_G (R) computed; crystallite size L_a estimated via wavelength-scaled formula.
• XPS: Thermo ESCALAB250Xi; survey and C1s deconvolution to quantify sp2, sp3, and C=O contributions; elemental composition (C, N, O) for as-spun, stabilized, and 950 °C pyrolyzed samples.
• TEM: Cross-sectional lamella prepared by FIB (FEI xT Nova Nanolab 200); samples sputtered with ~25 nm Au; ex situ lift-out to carbon film Cu grids; imaging with Philips CM200 FEG/TEM at 200 kV; bright-field, EDS, SAED. Sonicated fibers used to observe fractured edges and sheath layers.
• Electrical (bundle mats): 10 x 10 mm CNF mats (thickness ~10 µm) measured on a Cascade probe station to extract sheet resistance vs pyrolyzation temperature.

Sensor fabrication:

• CNF bundle strain sensor (OCA substrate): CNF strip (180 mm x 5 mm) transferred onto optically clear adhesive (OCA); seven copper tape electrodes placed 25 mm apart; silver conductive epoxy used for contact; sandwiched with a second OCA film; diced into six sensors.
• CNF bundle strain sensor (PDMS substrate): CNF rectangle (45 x 2 mm) fully embedded/encapsulated in PDMS for tensile strain tests.
• Single CNF sensor: Silicon wafer patterned with 100 µm-wide, ~100 µm-deep trenches (DRIE); 1 µm thermal SiO2 insulation; a single PAN fiber electrospun across trench; stabilized and pyrolyzed at 950 °C; ends secured with conductive silver epoxy; electrical connection via gold bonding wires to PCB.

Mechanical/electrical testing:

• Tensile strain (CNF-PDMS): Kammrath-Weiss tensile stage; quasi-static loading up to 50% tensile strain; three repeated runs with continuous resistance logging; gauge factor computed from linear fit of ΔR/R vs strain.
• Static bending (CNF-OCA): Sensor placed on 3D-printed arc brackets with radii of curvature (ROC) 8, 10, 12, 15, 18, 20 cm; biased with constant currents 0.5, 1.0, 1.5 mA; voltage across sensor recorded (HP 34401a). Baseline (flat) voltages: 0.948 V (0.5 mA), 1.894 V (1.0 mA), 2.83 V (1.5 mA).
• Long press (CNF-OCA): Fixed mass applied for 20 s; sensor configured as one arm of a balanced Wheatstone bridge; output logged before, during, and after load removal.
• Dynamic strain (CNF-OCA): Sinusoidal stimuli at 7 Hz and 35 Hz applied via an 8 mm sphere attached to a mini-shaker (Brüel & Kjær) through a 120 mm long, 2 mm dia steel rod; drive by function generator and B&K 2718 power amplifier. Bridge biased at 9 V DC; data acquired with NI USB-6009 at 2 kHz; unamplified outputs plotted.
• Single CNF flexibility: Hysitron PI-85 picoindenter inside FE-SEM used to stretch isolated CNF and image deformation.
• Wind tunnel test (single CNF): TECHQUIMENT AF1125 subsonic wind tunnel; voltage divider with DAQ 5 V output, R_ref ≈ 1.5 MΩ (~half the CNF resistance), buffer amplifier to NI USB-6009; velocities 0–35 m/s; each condition ≥60 s; sampled at 2 kHz; averaged in OriginPro.

• Pyrolyzation-induced diameter shrinkage: Average nanofiber diameter decreased by approximately 20%, 40%, 50%, 55%, and 60% for 500, 600, 700, 800, and 950 °C, respectively. Surface smoothing observed after carbonization.
• Raman evidence of graphitization and crystallite growth: I_D/I_G (R) decreased and L_a increased with temperature (Table 1): • 500 °C: R = 4.76, L_a = 1.04 nm • 600 °C: R = 2.90, L_a = 1.71 nm • 700 °C: R = 2.36, L_a = 2.10 nm • 800 °C: R = 2.07, L_a = 2.39 nm • 950 °C: R = 1.33, L_a = 3.73 nm Graphitization initiates below 700 °C with increasing ordering at higher temperatures.
• XPS composition and bonding: As-spun/stabilized/950 °C pyrolyzed bundles showed (C/N/O at.%): 80.9/16.9/1.9; 74.6/16.7/7.5; 90.2/2.78/7.0. C1s deconvolution at 950 °C: sp2 at 284.8 eV, sp3 at 285.6 eV, C=O at 289 eV; sp2 fraction ≈ 0.93, indicating dominant but discontinuous graphitic domains within an amorphous matrix.
• TEM: 600 °C samples showed no discernible crystalline domains in cross-section; 950 °C samples displayed layered planes parallel to the fiber skin and misoriented crystallites; core–shell structure with turbostratic stacking disorder; multiple sheath layers visible at fractured edges.
• Electrical conductivity vs temperature (bundles): Conductivity emerged at ≥600 °C; sheet resistance decreased exponentially with temperature. Resistance at 500 °C too high to measure; a 1 x 1 cm bundle at 700 °C had ~88.5 kΩ. Strong correlation with Raman (increasing G peak prominence).
• Mechanism: Conductive graphitic domains (sp2) embedded in sp3 amorphous carbon with inter-domain electron tunneling dominate conduction; tunneling distance sensitivity yields pronounced piezoresistivity.
• CNF bundle strain sensor (PDMS): Linear ΔR/R vs tensile strain up to 50% with average gauge factor 11.14 (n=3 runs).
• CNF-OCA bending: Under constant bias (0.5, 1.0, 1.5 mA), sensor voltage decreased linearly with increasing bending (smaller ROC), consistent with increased conductive path and reduced resistance.
• Long-press stability: Under a constant mass applied for 20 s, output remained stable and returned to baseline upon removal.
• Dynamic response: Accurately tracked cyclic strain at 7 Hz and 35 Hz with no noticeable delay.
• Ultralightweight: 1 cm² CNF bundle mass ~310 µg (~30x lighter than standard writing paper at ~9.56 mg/cm²).
• Single CNF NEMS flow sensor: Output voltage increased with airflow velocity from 5 to 35 m/s; consistent with resistance changes due to strain; circuit used R_ref ≈ 1.5 MΩ (about half the CNF resistance). SEM nanoindentation confirmed single CNF stretchability.

The study establishes that pyrolyzation temperature critically governs the evolution of graphitic order, composition, and electrical conductivity in electrospun PAN-derived CNFs. Raman/XPS/TEM collectively confirm the formation and growth of turbostratic graphitic domains embedded in an amorphous sp3 matrix as temperature increases, with measurable conductivity arising at and above 600–700 °C. The conductive domain discontinuity (inter-domain tunneling) model explains the observed piezoresistivity: mechanical strain alters tunneling distances between graphitic domains, modulating resistance. This mechanism accounts for the high strain sensitivity in CNF bundles and the measurable piezoresistive response in a single isolated CNF. Application-wise, CNF bundle sensors show linear, high-gauge-factor responses over large strains (0–50%) and robust performance in static bending, prolonged loading, and dynamic stimuli, while a single CNF NEMS device demonstrates airflow sensing over 5–35 m/s. The results indicate that electrospun CNFs offer a facile, ultralight, and flexible platform for static and dynamic sensing, potentially simplifying fabrication compared to more complex graphene/CNT systems while delivering competitive sensitivity.

This work elucidates the structure–property relationships of electrospun PAN-derived CNFs as a function of pyrolyzation temperature and demonstrates their effectiveness as ultralight, flexible piezoresistive sensing elements. Key contributions include: (i) comprehensive material characterization (Raman, XPS, TEM, SEM) linking increased graphitic ordering to enhanced conductivity; (ii) validation of the conductive graphitic domain discontinuity model as the origin of piezoresistivity in both single and bundled CNFs; (iii) fabrication of CNF bundle-based flexible strain sensors with linear response and an average gauge factor of 11.14 up to 50% strain; and (iv) demonstration of a single CNF NEMS airflow sensor operating from 5 to 35 m/s. These findings present CNFs as cost-effective, scalable, and high-performance alternatives for wearable and NEMS sensing. Future work could explore long-term durability, hysteresis and repeatability under cyclic loading, environmental stability (humidity/temperature), optimization of fiber alignment and bundle architecture, and integration into multifunctional wearable systems.