Flexible and Electrically Conductive 3D-Printed Ti₃C₂T_x MXene–Hydrogel Copolymers for the High-Precision Sensing of Biomechanical Processes

Abstract

The application of MXene–polymer composites to wearable and implantable medical devices requires the development of hydrophilic and biocompatible MXene–polymer hydrogel composites with high electromechanical response, flexibility, and durability. Here, we formulate low weight percentage MXene–hydrogel copolymer inks enabling the direct light processing (DLP) of Ti₃C₂T_x MXene–polyvinyl alcohol (PVA)–polyacrylic acid (PAA)–hydrogel composites. The low wt% MXene–PVA–PAA composites demonstrate high biocompatibility, mechanical flexibility, high sensitivity and high precision for sensing acute bending angles. The sub-millidegree angle resolution of these electromechanical sensors demonstrates their suitability for applications such as the highly precise tracking of joint movements. In addition, the synthesized MXene membranes show promise for applications in osmotic energy conversion, with a harvested electric power density of 6.79 Wm⁻².

1. Introduction

Lightweight and biocompatible MXene composite membranes and films have applications in numerous energy storage/generation and biomedical technologies, including osmotic energy conversion devices [1], tissue engineering [2], wearable devices [3], biodegradable microrobots [4] and conformable 3D electronic devices [5]. MXenes are a class of two-dimensional materials similar to graphene, which consist of atomically thin layers of transition metal carbides, nitrides or carbonitrides. The majority of M-X bonds in the two-dimensional sheets are exceptionally strong, while their surfaces can be functionalized with a wide variety of ligands, imparting them with high strength, flexibility, electrical conductivity and thermal stability. Since the discovery of the first MXene Ti₃C₂T_x in 2011, more than 30 other MXenes have been synthesized and extensively characterized [6].

Combining MXenes with polymer hydrogels such as polyacrylamide, polyvinyl alcohol (PVA) and polyacrylic acid (PAA) enables the development of composites with a unique combination of properties: electrical conductivity imparted by the MXene, and flexibility, biocompatibility and high cell adhesion imparted by the hydrophilic (co)polymer matrix. One of the particularly promising applications of MXene–polymer composites is in wearable and implantable medical devices powered solely by the biomechanical energy generated by physiological processes such as breathing, heart beats, muscle stretching, blood flow and the flow of ions in and out of cells via ion channels [7]. However, to withstand long-term in vivo use, implantable MXene devices also require durability under many cycles of repeated stress/strain and high electrical conductivity to enable the readout of electrical signals. These requirements can potentially be met by developing composites combining MXenes, such as Ti₃C₂T_x nanosheets, with hydrogel polymers such as PVA, PAA and PVA–PAA copolymer blends.

Extrusion-based 3D printing techniques such as direct ink writing (DIW) [8] and direct light processing (DLP) have recently emerged as promising methods for fabricating MXene composite films with a range of microstructures and morphologies. In DLP, the laser-printed polymers are cured with UV light to complete polymer cross-linking. These techniques require the formulation of inks that have high viscoelasticity and good shear-thinning effect when passing through a small nozzle. There are few commercial inks with these properties, hence development of new inks is an active area of research. Recent advances include the 3D printing of piezoresistive sensors using a supra-molecular system for DIW [9] the and real-time monitoring of hand and limb movements using pressure [10] and strain [11] sensors. Emulsion-based graphene oxide and MXene inks have also been developed for the 3D printing of foams with high electrical conductivity and electromagnetic shielding performance [12]. Capillary-driven DIW and fused deposition modelling methods have been employed for the 3D and 4D printing of MXene nanostructures into ordered arrays and micropatterns on flexible substrates [13]. The outstanding biocompatibility of MXene-based hydrogels has enabled development of an adhesive self-healing MXene hydrogel for epidermal electronics [14]. Further applications in bio-integrated electronics can be enabled by isolating the sensing layer from direct skin exposure by encapsulation within an ultrathin, flexible and biocompatible film [15]. In the present work, we fabricate MXene–polymer hydrogel composites using hydrophilic and biocompatible MXene inks suitable for 3D DLP and demonstrate their suitability to implantable biomedical devices for the accurate monitoring of biomechanical processes such as joint movements.

2. Materials and Methods

2.1. Chemicals and Agents

All chemicals were purchased from commercially available sources and used as received. Lithium fluoride (LiF), titanium aluminium carbide MAX phase (Ti₃AlC₂), n-butyllithium solution 2.5M in hexane, hydrochloride acid (37 v% HCl), poly(vinyl alcohol) (PVA, MW = 89,000−98,000 g mol⁻¹), acrylic acid (AA, 99 wt% in DI water), poly(ethylene glycol) diacrylate (PEGDA, average M_n = 575 g mol⁻¹), and lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) were all purchased from Sigma-Aldrich Australia.

2.2. Preparation of MXene (Ti₃C₂T_x) Nanosheets

Ti₃C₂T_x nanosheets were synthesized by etching the Ti₃AlC₂ phase with LiF/HCl as reported previously [16]. Briefly, 1g of LiF powder was dissolved in 20 mL of 9M HCl, and the solution was mixed thoroughly at room temperature with rigorous stirring. Then, 1 g Ti₃AlC₂ was added carefully into the mixed solution drop by drop to avoid overheating from the exothermic reaction. The reaction was allowed to proceed under continuous stirring for 24 h at 35 °C. The as-obtained precipitate was collected by centrifuging and then washed repeatedly using deionized water until almost neutral pH (°6). The solid sample was then dried into membranes at room temperature, as shown in Figure S1.

2.3. Preparation of MXene–Hydrogel Copolymer Inks for 3D Printing

MXene–PVA–PAA inks with MXene concentrations ranging from 0.07 to 0.28 wt% were prepared by solution blending [17] using water as the solvent. The compositions of each sample are listed in

Table 1. Composition of MXene–hydrogel copolymer inks used for 3D DLP.

Sample Name	Sample Composition	PVA	AA	PEGDA	MXene	LAP	H2O
PVA–PAA–0.28MX	10PVA25PAA-0.28MXene	3 g	7.5 g	75 mg	30 mg	60 mg	30 g
PVA–PAA–0.14MX	10PVA25PAA-0.14MXene	3 g	7.5 g	75 mg	15 mg	60 mg	30 g
PVA–PAA–0.07MX	10PVA25PAA-0.07MXene	3 g	7.5 g	75 mg	7.5 mg	60 mg	30 g
PVA–PAA	10PVA25PAA	3 g	7.5 g	75 mg	3.75 mg	60 mg	30 g

. The method of formulation of an ink containing 0.28 wt% MXene (denoted by 10PVA25PAA-0.28MXene) is selected as an example of the preparation process (Figure 1). First, 3 g of PVA powder was dissolved in 100 mL of deionized (DI) water at 90 °C for 5 h to create a 20 wt% PVA solution. Additionally, 30 mg of MXene powder was dispersed uniformly in 15 mL of DI water through sonication under a nitrogen atmosphere for 90 min, yielding a 2 mg mL⁻¹ MXene solution. Then, 15 mL of the PVA solution and 15 mL of the MXene solution were combined in a light-shielded container. Subsequently, 7.5 g of AA, 75 mg of PEGDA, and 60 mg of LAP were added. The mixture was stirred at 2000 rpm for 5 min to prepare the final 3D printing ink, referred to as 10PVA25PAA-0.28MXene in Table 1.

2.4. 3D Printing of MXene–Copolymer Hydrogels

Dogbone patterns of dimensions 100 × 6 × 5 mm following the ASTM (E8) subsize standard were modelled using Pro/ENGINEER software and exported as ‘obj’ files. These files were subsequently opened in Anycubic Photon Workshop 3D Slicer Software, where the following parameters were configured: z-axis movement speed of 1 mm s⁻¹, layer thickness of 50 μm, and exposure duration of 25 s per layer. For the bottom 10 layers, the exposure time was set to 30 s. The ‘dl2p’ format slicing file was then printed using a DLP 3D printer (Anycubic Photon D2) (University of Technology Sydney, Ultimo, Australia) equipped with a near-ultraviolet (λ = 405 nm) laser. After loading the photo-responsive ink into the printer’s resin tank, the defined printing program was initiated. Post-printing, the residual ink on the surface of the printed objects was gently washed off using DI water. Finally, the printed hydrogels were put into a UV curing machine (Anycubic) (University of Technology Sydney, Ultimo, Australia) for 30 min to ensure complete curing of any uncross-linked components.

2.5. Characterization of MXene Membranes

Scanning electron microscopy (SEM) (University of Technology Sydney, Ultimo, Australia) of MXene sheets was conducted on a Supra 55VP SEM operating at 5 kV. Morphological characterization was conducted on an FEI Tecnai T20 transmission electron microscope operating at 200 keV. Powder X-ray diffraction on a Bruker D8 Discover Diffractometer was used for the evaluation of the MXene structure. The X-ray source was Cu-Kα (λ = 1.5406 A), with a scanning range of 2–10° and step size of 0.02°. Fourier transform infrared spectroscopy (FTIR) was conducted on MXene composites before and after freeze-drying. In the freeze-drying process, MXene hydrogel composite samples were frozen at –21 °C for 24 h and then transferred to an Alpha 3-4 LSCbasic freeze dryer with a condenser temperature of –(95$\pm$5) °C and subjected to a vacuum pressure of 0.01 mbar for 3 days. A Nicolet FT-IR 6700 spectrometer was used to collect FTIR spectra in transmittance mode over the range of 4000–400 cm⁻¹ with a resolution of 4 cm⁻¹. Each spectrum was obtained by averaging 32 scans. Peaks were identified and assigned to different functional groups by comparison with reference spectra for PVA, PAA and fluorinated Ti₃C₂T_x MXene.

2.6. Osmotic Energy Conversion Evaluation

To test the ion transport performance, MXene membranes were mounted into a home-made electrochemical device between two electrolytes with different ion concentrations. Measurements were conducted in KCl solutions with concentrations ranging from 10⁻⁴ to 2 M. The current density and power density curves in Figure 2g were obtained by recording the current under different loaded resistance without external voltage. The generated current originated from the osmotic pressure. With C_MXene/C_silica = 0.5/0.01 M, the power density was calculated by P_R = I² × R_L where R_L is the electrical loading resistance.

2.7. Mechanical Property Measurement and Analysis

Tensile tests of the MXene hydrogel composites were conducted on a Shimadzu Autograph AGS-X series universal testing machine. Tensile tests were carried out at a speed of 50 mm/min, and the load and displacement data were recorded using Shimadzu Trapezium X software. Ultimate tensile strength and Young’s modulus were calculated by analysing three stress–strain curves collected from each sample.

2.8. Electrochemical Response Measurement

Electrochemical response measurements were conducted on the dogbone-shaped hydrogel composite samples by clamping the ends of each specimen with alligator clips and connecting them to a Bio-Logic SP-300 electrochemical workstation (software version EC-Lab V11.60), as shown in Figure S1. A constant 6 V DC bias was applied throughout the measurements. Each sample was bent along its longitudinal axis from 0° to 90° in 10° increments. During each bending step, current-time traces were sampled at 0.1 s intervals, and the steady-state current was determined at each angle. For each composition (PVA–PAA, PVA–PAA–0.07MX, PVA–PAA–0.14MX, and PVA–PAA–0.28MX), angle-resolved current readings were collected across nine discrete bending angles, with 50 to 200 valid data points per angle depending on sample stability and noise level.

2.9. Machine-Learning-Assisted Current–Angle Data Analysis

Current–angle data sets obtained from the electrochemical response measurements were analysed using Python (v3.10) with the pandas, numpy, and scikit-learn libraries. Data files were first cleaned by removing incomplete records and rounding angle values to the nearest 10°. An isotonic regression model was applied to fit a monotonically increasing current–angle curve, ensuring physical consistency in the directional response.

The sensitivity-to-noise index (SNI, units: 1/°) was defined at each bending angle $\mathrm{\theta }$ as follows:

$$\mathrm{SNI}\left(\theta \right)={\displaystyle \frac{\left|\frac{dI}{d\theta }\right|}{\sigma \left(\theta \right)}}$$

(1)

where $\left|{\displaystyle \frac{dI}{d\theta }}\right|$ is the absolute derivative of the current–angle curve and $\sigma \left(\theta \right)$ is the local noise estimated using the median absolute deviation (MAD) of residuals at angle $\theta$. The overall SNI was computed as the mean of SNI values across interior angles (10–80°), excluding endpoints to reduce boundary artefacts.

For inverse angle prediction, the mean absolute error (MAE) and its standard deviation (SD) were calculated across grouped cross-validation folds, where each bending angle group was held out once as the test set. For each fold k, the fold-wise MAE was defined as follows:

$${\mathrm{MAE}}_{\mathrm{k}}={\displaystyle \frac{1}{n_k}}\sum _{i=1}^{n_k}\left|{\widehat{\theta }}_i^{\left(k\right)}-{\theta }_i^{\left(k\right)}\right|$$

(2)

where ${\theta }_i^{\left(k\right)}$ and ${\widehat{\theta }}_i^{\left(k\right)}$ are the true and predicted bending angles of the $i$-th sample in fold $k$, and $n_k$ is the number of test samples in that fold.

The overall MAE and SD were then computed as follows:

$$\mathrm{MAE}={\displaystyle \frac{1}{K}}\sum _{k=1}^K{\mathrm{MAE}}_{\mathrm{k}}$$

(3)

At each angle, local noise was estimated using the median absolute deviation (MAD) of residuals. A sensitivity-to-noise index (SNI, units: 1/°) was computed as the absolute derivative |dI/dθ| divided by the corresponding noise σ(θ), and the overall SNI was reported as the mean across interior angles, excluding endpoints. To assess the accuracy of inverting current back to angle, the model was trained and evaluated using a grouped cross-validation strategy. The mean absolute error (MAE) and its standard deviation (SD) were used to quantify the model’s generalization performance.

3. Results and Discussion

3.1. Structure and Properties of MXene Membranes

Prior to ink preparation, the synthesized MXene nanosheets were characterized using SEM, TEM and XRD. SEM (Figure S2) and TEM (Figure 2a,b) images show that the MXene sheets are uniform in thickness, with typical lateral sizes in the 100–500 nm range. The structure of the MXene sheets was analysed by XRD (Figure 2c), which shows that the (002) peak is located at 5.83°, corresponding to an interlayer spacing (d-spacing) of 1.52 nm according to the Bragg equation.

3.2. Ion Transport Properties of MXene Membranes

MXene membranes are normally charged due to the F⁻, O²⁻ and OH⁻ surface functional groups, making them perfectly suited for use in ion-selective separation [18], with applications in water treatment, gas separation, and osmotic energy conversion. When present in sufficiently high concentrations on the surfaces of lamellar MXenes, such charged surface groups induce the formation of an electrical double layer (EDL) in the channels between the MXene sheets. Due to the confinement of the double layers within the nanoscale channel volumes between neighbouring sheets, counter-ion transport is accelerated while co-ion transport is decelerated. There are several mechanisms by which ion-selective separation can be achieved, including size selectivity (controlled by the membrane channel or pore size) [19], charge selectivity (controlled by the charge of the diffusing ion) [20], and adsorption selectivity (controlled by the type of adsorption sites and mechanism of adsorption) [21].

The ion transport properties of the MXene membranes were investigated using current–voltage (I–V) measurements (Figure 2d) using a home-made electrochemical cell. Measurements were conducted in KCl solutions with concentrations ranging from 10⁻⁴–2 M. These confirm that the membrane exhibited linear ohmic characteristics in a range of KCl concentrations (Figure 2e). To evaluate the ion selectivity, the MXene membrane was mounted between two electrolytes with different ion concentrations (Figure S3). In the absence of an external voltage (V = 0), both K⁺ and Cl⁻ diffuse from the high (1 M) to the low (10 µM) concentration region. A net current (short-circuit current I_sc) is observed only when there is a difference in the diffusion rates of the two ions. Because the MXene is negatively charged, cations (i.e., K⁺) are enriched while anions (i.e., Cl⁻) are excluded from the channel, resulting in the I_sc direction being consistent with the positive ion flow from a high to a low concentration. Reversing the concentration gradient configuration results in a reversed direction of I_sc (Figure S4). This result demonstrates that the Ti₃C₂T_x membranes exhibit a stable surface charge-controlled cation selectivity. When the ion transport from the bulk region decreased to the nanoconfined areas, the overlap of the electric double layer (EDL) was calculated according to the Debye length (λ_D). The Debye length is the effective distance of the charged surface and can be calculated according to the equation ${\mathrm{\lambda }}_D=\sqrt{{\displaystyle \frac{{\epsilon }_r{\epsilon }_{0}kT}{2C_{bulk}{z}^2{e}^2}}}$, where ${\epsilon }_r$, ${\epsilon }_0$, $e$ and $k$ are the relatively permittivity of the solution, the permittivity of a vacuum, the electron charge, and the Boltzmann constant, respectively. This is increased and can completely cover the pore size of the chiral layer when the concentration of the feed solution is under 0.01 M (Table S1). The calculated result agrees with the experiment shown in Figure 2f, where the ionic conductance of the MXene is deviated from the black dash line of the 0.01 M KCl solution, which further confirms that the free-standing MXene membrane has the ability to control ion transport via surface charge [21] Finally, the harvested electric power density (P_R) was assessed using the same configuration as above (Figure 2g). The power density can be calculated as P_R = I² × R_L with C_MXene/C_silica = 0.5/0.01 M, where R_L is the electrical loading resistance. As shown in Figure 2g, the MXene membranes can obtain 6.79 W · m⁻² of osmotic energy conversion. The results confirm that the prepared MXene membrane shows promising ability as an energy conversion-related device in energy storage or biomedical technology, when compared with other published works [18,19].

3.3. Molecular Structure and Mechanical Properties of PAA–PVA–MXene Composites

FTIR spectroscopy within the range 4000–400 cm⁻¹ was used to analyse the molecular structure and bonding of the MXene–hydrogel composites. Due to the low concentrations of MXene used in this work, and as observed in a prior study of low weight percent MXene composites [22], there was no difference in the FTIR peaks from the freshly prepared MXene–PVA–PAA composites when compared with the pure PVA–PAA sample. FTIR spectra of the freshly printed hydrogel composites (Figure 3a) clearly exhibit a broadened O–H stretching band at 3550–3200 cm⁻¹, indicating substantial water content as well as hydrogen bonding between PVA–PAA [23].

Two bands at 2952 cm⁻¹ and 2922 cm⁻¹ can be ascribed to C–H stretching in aliphatic CH₂ groups. Correspondingly, the C–H bending of acrylate is found at 1413 cm⁻¹ and the CH₂ bending is located at 1450 cm⁻¹ [24]. The characteristic C=O stretching peak of PAA presents at 1705 cm⁻¹ [25]. A prominent band at 1635 cm⁻¹ is present in the FTIR spectra of all four composites. We ascribe this band to H–O–H bending in the freshly printed hydrogel composites, as reported in previous works [25]. The bands at the low wavenumbers of 1254 cm⁻¹ and 1178 cm⁻¹ can be attributed to C–O stretching [26]. The bands located at 1090 cm⁻¹ and 1050 cm⁻¹ are ascribed to the P=O and P–O bonding in the light initiator LAP [27].

To verify the incorporation of MXene membranes in the MXene–PVA–PAA composites, FTIR was also measured after freeze drying (Figure 3b). The FTIR spectra of the composites exhibit sharper and more characteristic polymer bands after removing most of the free and adsorbed water. The O–H band narrows to 3400–3250 cm⁻¹, due to the removal of water and the enhancement of PVA–PAA hydrogen bonding. The O–H peak shifts from 3365 cm⁻¹ for the pure PVA–PAA to 3343 cm⁻¹ for the PVA–PAA–0.14MXene composite, indicating the presence of the MXene [28]. The peak at 1635 cm⁻¹ that is ascribed to H–O–H bending completely disappears and the peak at 2583 cm⁻¹ due to hydrogen bonding between PVA and PAA becomes more obvious. The C=O peak shifts from 1705 cm⁻¹ to 1699 cm⁻¹ in the MXene–PVA–PAA composites after freeze drying due to a reduction in hydrogen bonding with water and rearrangement of PAA/PEGDA carbonyl environments. The rest of the bands in the FTIR spectra of the freeze-dried MXene–PVA–PAA composites are the same as those in the freshly prepared samples.

Figure 3c displays the stress–strain curves and summarized mechanical properties of the freshly 3D printed hydrogel composites, PAA–PVA, PVA–PAA–0.07MX, PVA–PAA–0.14MX, and PVA–PAA–0.28MX. Three samples of each composite were tested using the same procedure. The introduction of MXene sheets into the PAA–PVA composites resulted in a progressive increase in ultimate tensile strength from 67 ± 1 KPa at 0 wt% MXene to 71 ± 7 KPa at 0.07 wt%, 78 ± 26 kPa at 0.14 wt% and 133 ± 50 KPa at 0.28 wt% as shown in Figure 3d. The increase in tensile strength may be due to the formation of strong hydrogen bonds between the MXene sheets and polyacrylic acid molecules in the PAA [29]. The Young’s modulus of the composites generally increase with the addition of MXene; however, the modulus of the composite with 0.07 wt% MXene is higher than at 0.14 wt%.

The mechanical properties of the freshly printed hydrogel composites are summarized in Figure 3d and are compared with previously reported values for similarly low wt% hydrogel composites in

Table 2. Mechanical properties of MXene–PAA–PVA hydrogels (this work) compared with prior work.

MXene Composites	Manufacturing Technique	MXene wt%	Young’s Modulus (KPa)	Tensile Strength (KPa)	Elongation (%)
PVA–PAA–MXene (This work)	DLP 3D printing	0	1.4 ± 0.1	67 ± 1	65.1 ± 6.7
		0.07	3.1 ± 0.1	71 ± 7	36.2 ± 5.2
		0.14	2.0 ± 0.1	78 ± 26	90.5 ± 9.7
		0.28	3.7 ± 0.1	133 ± 50	105.1 ± 15.3
GelMa–MXene [31]	Photoinitiating solution casting	0	53.2 ± 9.9	14 ± 2	26.3 ± 5.6
		0.025	45.5 ± 4.7	11 ± 4	26.4 ± 3.7
		0.05	35.0 ± 7.2	11 ± 1	31.4 ± 4.5
		0.125	26.01 ± 4.3	8 ± 2	37.3 ± 6.6
		0.25	35.2 ± 3.5	10 ± 2	25.3 ± 6.5
PAA–Starch–MXene [25]	Solution casting	0.24	211	340	1237
PAA–MXene [30]	Direct ink writing 3D printing	1	795.8	893	622

. The study by Zhao et al., showed that at 1 wt %, the tensile strength of 3D-printed PAA–MXene hydrogel was found to be 893 kPa, substantially higher than the present work [30]. This can be ascribed to both a lower water content and higher crosslinking density due to the synthesis method, in which acrylic acid (AA) monomers are pre-polymerized directly within MXene nanosheets and in which ammonium persulfate (APS) was used after 3D printing to produce a fully crosslinked hydrogel. A series of GelMa–MXene composites with similar weight percentages [31] presented higher Young’s modulus than our MXene–PVA–PAA composites, but much reduced strength and elongations. A starch-based PAA–MXene composite displayed much better mechanical properties [25]. These differences can be attributed to differing matrices, MXene contents, manufacturing techniques, and mechanical testing conditions, and will be investigated further in the future [30].

3.4. Electrochemical Properties of 3D-Printed MXene–PVA–PAA Composites

Following assessment of the osmotic energy conversion properties of the MXene membranes, MXene–PVA–PAA copolymer hydrogel inks and PVA–PAA copolymer hydrogel inks were subjected to angle-resolved electrochemical characterisation under a constant 6 V DC. Current–angle datasets were obtained across nine bending angles from 0° to 90° and these served as the foundation for quantitative performance analysis. The real-time response and recovery curves are presented in Figure S4. We can see that the current stabilities of all composites were excellent (Figure S4a). The current levels of the composites highly depended on the water content of each sample. The current variations of all composites were very clear and steady (Figure S4b) when they were bent from 0° to 90° with a step size of 10°. The relative values of current variations are plotted in Figure S4c. The responses of the MXene composites were much more stable compared with the pure PVA–PAA, especially at the recovery stage. Figure 4a shows the current–angle (I–θ) responses of the 3D-printed PVA–PAA and MXene–hydrogel copolymers with MXene loadings of 0.07, 0.14, and 0.28 wt%. All samples exhibited a clear increase in electrical current, with increasing bending angle across the 0–90° range. The pure PVA–PAA composite showed a quasi-linear current increase from 0.52 to 0.60 mA between 0° and 70°, followed by a sharp, non-linear increase to 0.74 mA at 90°. However, although the pure PVA–PAA hydrogel showed the largest absolute current variation, the MXene-containing composites demonstrated more uniform and linear responses throughout the full bending range. Specifically, the current increased from 0.31 to 0.37 mA for 0.07 wt%, 0.25 to 0.30 mA for 0.14 wt% and 0.06 to 0.07 mA for the 0.28 wt% MXene composite.

To quantitatively evaluate sensor performance and enable accurate angle inference from current readings, a machine-learning-assisted analysis method was implemented (Figure 4b), as described in the Section 2. Figure 4c compares the overall SNI, averaged over the interior angles, for the four samples. A higher SNI value indicates superior angle sensitivity relative to noise. The PVA–PAA–0.14 wt% MXene sample achieved the highest SNI of ~4.36, indicating that a 1° change in angle produces a current response approximately four times greater than the background noise. This corresponds to a minimum resolvable angle of approximately 1/SNI ≈ 0.23°. The SNI values for the 0.28 wt% and 0.07 wt% samples were ~3.90 and ~3.52, respectively, while the neat PVA–PAA sample had the lowest SNI of ~1.14. This result suggests that, while the pure PVA–PAA hydrogel has a larger absolute signal, its significantly higher noise level limits its effective sensitivity. We assessed the accuracy of the sensors by performing a current-to-angle inversion using five-fold cross-validation. This process simulates the real-world application of converting a measured current back into an angle. Figure 4d shows the mean absolute angle error (MAE) of this inversion. The PVA–PAA–0.14 wt% sample demonstrated a remarkably low MAE of ≈0.0002°, significantly outperforming neat PVA–PAA (≈0.0034°), 0.07 wt% (≈0.0011°), and 0.28 wt% (≈0.0016°). Furthermore, the SD of the inversion error was also the lowest for the 0.14 wt% sample (≈0.0003°), as shown in Figure 4e and

Table 3. Sensing performance of the PVA–PAA–MXene composites.

Name	Sensitivity to Noise (1/°)	Mean Absolute Angle Error (°)	Standard Deviation (°)
PVA–PAA	1.14	0.0034	0.003
PVA–PAA–0.07MX	3.52	0.0011	0.001
PVA–PAA–0.14MX	4.36	0.0002	0.0003
PVA–PAA–0.28MX	3.90	0.0016	0.002

. The combination of sub-millidegree resolution and sub-microradian stability underscores the sensor’s capability for highly reliable angle reconstruction and should enable applications in biomechanical monitoring such as high-precision tracking of joint movements.

Under a constant bias voltage (6 V), the bending-induced current change reflects the variation in the sample’s equivalent resistance ($I=V/R$). The pristine PVA–PAA hydrogel is very soft and its electrical conduction is dominated by aqueous/ionic pathways; therefore, at the same bending angle it is more prone to pronounced local compression and microstructural rearrangement. This leads to a shortened effective conductive path length, an increased effective conductive cross-sectional area, and improved connectivity of water/ion transport channels, resulting in a larger decrease in resistance and thus a larger increase in current. In contrast, after introducing MXene, the 2D nanosheets act as rigid fillers and “anchoring” sites that reinforce and stabilise the polymer network. These suppress chain slippage, pore collapse, and water-phase migration during bending, thereby improving structural recovery upon bending and release. As a result, the deformation-induced modulation of the equivalent resistance is reduced, the current response becomes more symmetric, and the overall response amplitude decreases, especially at higher MXene loadings.

4. Conclusions

We have demonstrated 3D DLP of highly flexible and biocompatible low weight % MXene–PVA–PAA composites suitable for high-precision biomechanical sensing. We first synthesized the MXene nanosheets by etching the Ti₃AlC₂ MAX phase with LiF/HCl. The free-standing MXene membranes have the ability to control ion transport via surface charge and electrochemical measurements of harvested electric power density show 6.79 W · m⁻² of osmotic energy conversion. These MXene sheets were then mixed with hydrophilic PVA–PAA polymers to formulate printable inks with MXene concentrations in the range of 0.07–0.28 wt%. DLP was used to fabricate dogbone-patterned MXene–hydrogel copolymer composites which were subjected to repeated bending tests and evaluated for their performance as biomechanical sensors. The low wt% MXene–PVA–PAA composites demonstrate high mechanical flexibility and high precision for sensing acute bending angles. Further work will focus on improving the mechanical properties (Young’s modulus and tensile strengths) while retaining the electromechanical sensing performance and evaluating performance in real-world biomechanical applications such as the highly precise tracking of joint movements.