a Illustration showing the sequential inkjet printing of CNT and reactive silver ink at the same location on the PLL-coated cotton fabric. b Optical images of the printed pattern for each process. The pattern color changed from black to silver with the precipitation of the AgNPs. The scale bars are 1 cm. c Schematic of the deposited materials in each fabrication process. CNT and reactive silver ink were sequentially printed and thermally reduced to precipitate AgNPs. d Chemical structures of PLL, CNT, and reactive silver. e Illustration of the hydrogen bonding between cotton, PLL, and CNT. Positively charged PLL acts as an adhesion promotor by bonding with negatively charged cotton fabric and CNTs. f–h Optical images showing the flexibility, washability, and heat resistance of the fabricated devices. i Wearable applications, namely tactile socks and heating gloves, of the proposed device.

The chemical structures of the materials is illustrated in Fig. 1d. PLL is a positively charged material with amine groups that enhance adhesion to other functional groups, such as hydroxyl, carbonyl, and carboxyl groups^46,47. The CNTs in the ink are functionalized with negatively charged hydroxyl groups, which improve their dispersibility in water⁴⁸. The Ag ions in the reactive silver ink exist as diamminesilver(I), which receive electrons from formic acid and are reduced to AgNPs upon heating. In this study, PLL served as the adhesive layer to improve the adhesion between cotton fabric and CNTs, and also, CNTs acted as heterogeneous nucleation sites of AgNPs, resulting in a better interfacial adhesion between precipitated AgNPs and printed CNTs, enhancing the durability of e-textiles⁴⁹. Figure 1e shows a detailed diagram of the chemical interactions in which the amine groups of the PLL bond with the [–OH] groups of the cotton fabric and hydroxyl groups of the CNTs.

Bending, washing, and heat-resistance tests were conducted on the fabricated cotton-based e-textiles to confirm their suitability for wearable applications. The printed e-textile exhibits high flexibility, demonstrating the applicability to dynamic environments that involve bending and twisting (Fig. 1f). Furthermore, it can be washed and ironed, similar to clothing (Fig. 1g, h). Based on these advantages, we fabricated tactile socks that can monitor body movements (walking and running) in real-time and thermal gloves that provide constant thermal sensation regardless of hand movement (Fig. 1i). These applications demonstrate the application of our cotton-based e-textiles produced by simple inkjet printing in various smart clothing fields.

Characterization of printed electrode

The interdigitated electrode patterns were printed with a total width of 6 mm and length of 8 mm on a cotton fabric in a 3 × 6 array (Fig. 2a). The particle-less ink facilitated printing possible without nozzle clogging. The silver color of the pattern in the optical image depicts the successful metallization of Ag ions without being trapped by the hydroxyl groups on the surface of the cotton fabric. Scanning electron microscopy (SEM) confirmed the precipitation of a large amount of AgNPs, completely covering the woven surface of the cotton fabric. Upon further magnification, the AgNPs (0D material) filled the empty spaces between the CNTs (1D material). The surface topology of the cotton-fabric-based electrode was examined by confocal microscopy (Fig. 2b). The confocal image shows a surface profile similar to that of the woven structure of the cotton fabric. Cross-section profiles at the intersections of the warp and weft yarn were measured along the A-A′ and B-B′ lines, indicating the rough surface owing to the weaving of the yarns. Energy-dispersive spectroscopy (EDS) measurements indicate the successful precipitation of the AgNPs throughout the surface of the cotton fabric by printing reactive silver ink over the CNTs (Fig. 2c). However, it can be seen that Ag was not detected in some of the warp and weft intersection areas. The reason is that, as can be seen in the confocal image, the depth of the gaps at those intersections was very deep (about 60 μm), and the amount of inkjet-printed droplets was minimal (about tens of picoliters), so the CNTs might not fill the gaps^30,50,51. Since the gaps were not filled with CNTs, AgNPs were not precipitated, and uncoated areas were generated. These areas were formed only at the intersections of the woven fiber bundles, and their length can be determined by the yarn thickness of the fabric used. Therefore, if the printed line width was narrower than the length of the uncoated areas or the yarn thickness, the conductive path would not be formed.

Fig. 2: Characterization of printed electrodes.

a Optical and SEM images of the 6 mm × 9 mm electrodes printed in a 3 × 6 array. AgNPs were precipitated over CNTs to fill gaps. b Confocal image of the electrode with a hierarchical structure. Cross-section profiles measured along two dashed lines (A-A′ and B-B′). c EDS spectra of the electrode showing AgNPs precipitated throughout the surface of the cotton fabric. Some uncoated areas without AgNPs were formed. d Electrical conductivity of electrodes printed with several materials. CNT/reactive silver had the highest conductivity, which is over 10³ times that of CNT. Data in (d) is presented as means ± standard deviations (n = 5). Source data are provided as a Source Data file.

The electrical conductivity of the patterned electrode was evaluated by printing various inks composed of single or multiple materials (Fig. 2d). The AgNP-based inks that can be printed without clogging the nozzle cannot form conductive lines on rough and porous cotton fabrics because of their insufficient concentrations. Additionally, the reactive silver ink alone could not form an electrode because of the [–OH] groups on the surface of the cotton fabric. The electrode printed using CNTs alone exhibited a low electrical conductivity, and the conductivity that could be achieved even by increasing the number of prints was limited. The conductivity was 0.033 S m⁻¹ during the first printing (Supplementary Fig. 3). After printing five times, the conductivity increased to 52.21 S m⁻¹, and no significant difference was noted after further printing. The conductivity that could be achieved with the CNTs alone was not sufficient for use as an electrode, and the conductivity changed depending on the pressure or mechanical deformation. Combining the CNTs with the AgNPs did not produce a significant improvement over CNT-only prints owing to the insufficient AgNP quantity. However, when using CNTs with reactive silver ink, AgNPs were precipitated at the CNTs, forming an interconnected Ag network. This network resulted in a high conductivity of 1.25 × 10⁵ S m⁻¹, which was over 2 × 10³ times higher than that of CNT-only electrodes, making them viable for electrode use^36,37. The printing number of the reactive silver ink also affected the electrical conductivity. The electrode had the conductivity of 1.18 × 10³ S m⁻¹ with single-ink printing, increasing to 1.25 × 10⁵ S m⁻¹ after printing twice, with no significant resistance change upon further printing (Supplementary Fig. 4).

In the case where only zero-dimensional (0D) materials, such as AgNPs, are incorporated, mechanical deformations, including bending and twisting, tend to induce crack formation, resulting in a deterioration of electrical conductivity. In contrast, the fabricated electrode comprises both 0D AgNPs and one-dimensional (1D) carbon nanotubes (CNTs), where the CNTs serve as conductive bridges that effectively preserve electrical conductivity under mechanical deformations⁵². The fabricated electrode exhibited negligible changes in resistance even after 1000 bending cycles and showed no significant resistance variation upon twisting up to 720°, confirming its potential for flexible electronic applications (Supplementary Figs. 5, 6).

The minimum line width achievable in this process, that is, the print resolution, was determined by printing 1 cm long lines with various widths, and the conductivity was measured. The line width was printed with a 10 μm step decrease from 250 to 120 μm (Supplementary Fig. 7). The actual width of the printed line was thicker because of ink spread owing to the hydrophilicity of the cotton fabric (Supplementary Fig. 8 and Supplementary Table 1). Only the conductivity of the lines with widths of 324.93 to 209.47 μm decreased in the range of 6.18 × 10⁴ S m⁻¹ to 3.68 × 10⁴ S m⁻¹, whereas no conductive path was formed for the other lines (Supplementary Fig. 9). In this study, a cotton twill fabric with a yarn count of Ne 40 was used. Here, Ne is a unit of yarn count, defined as the amount of yarn measured in hanks (840 yards) needed to make one pound of weight⁵³. That is, the higher the number, the thinner the yarn. In the case of the fabric we used, the yarn thickness was about 200 μm, and the length of the uncoated area was also almost the same. Therefore, the minimum conductive pattern width that we could obtain on the cotton fabric using this process was 209.47 μm. However, using fabrics with a higher yarn count would allow for thinner threads, resulting in a smaller uncoated area, and consequently, a narrower conductive pattern width was achievable (Supplementary Table 2). An increase in yarn count also reduces the surface roughness, which may negatively affect the sensitivity of the tactile sensor fabricated on such fabrics⁵⁴. In this study, Ne 40 fabric was selected because of the high surface roughness of Ne 20 fabric, making it difficult to form a highly conductive electrode with a small amount of ink discharged by inkjet printing. In addition, the contact area between the upper and lower layers varies depending on the weave structure of the fabric, with a twill weave in particular providing the maximum output⁵⁵. Therefore, Ne 40 fabric with a twill weave structure was used to fabricate the e-textiles.

Pressure-sensing characteristics of tactile sensor

We demonstrated a wearable tactile sensor as proof of the feasibility of the proposed inkjet printing process. A contact-resistive-type tactile sensor was fabricated, consisting of a sensing layer printed only with CNT ink and two electrodes printed on a single fabric with CNT/reactive silver ink (Fig. 3a). The sensing layer was printed in a square pattern with a size of 1 cm × 1 cm, and electrodes were printed in two rectangular patterns with a size of 1.0 cm × 0.4 cm and 0.3 cm spacing. The rough and hierarchical structure of the cotton fabric surface increased the contact area between the electrode and sensing layer with an applied pressure, decreasing the resistance (Fig. 3b). In addition, the increased contact between the CNTs in the sensing layer also led to a decrease in resistance owing to the porous structure of the cotton fabric. In the electrode, the contact between the CNTs increased as pressure is applied, but the resulting change in resistance was negligible because of the high electrical conductivity of the AgNPs; therefore, there was no resistance change in the electrode.

Fig. 3: Characterization of the resistive tactile sensor.

a Schematic of the fabricated tactile sensor composed of a sensing layer only printed with CNT ink and an electrode layer printed with CNT/reactive silver ink. b Sensing mechanism of the tactile sensor. The contact area between the layers and contact points between the CNTs in the sensing layer increased with pressure, resulting in a resistance drop. c Current–voltage curves of the sensor at various pressures (1–500 kPa). d Relative current change (ΔI/I₀) of the sensor under varying pressure. Pressure sensing performance after 1000 cycles of repeated tests. e Transient curve of the sensor after loading and unloading 10 Pa. f Pressure detection limit of the sensor with 10 Pa loaded sequentially from the unloaded state. g Real-time response of the sensor to repeated loading and unloading of 300 kPa over 3000 cycles at a rate of one cycle per second. Data in (d) is presented as means ± standard deviations (n = 5). Source data are provided as a Source Data file.

The tactile sensor’s current (I)–voltage (V) curves under various pressures ranging from 1 to 500 kPa demonstrate linearity (Fig. 3c and Supplementary Fig. 10). Initially, the sensor had a high resistance of 4.73 MΩ because of the rough surface and high porosity of the fabric. As the pressure increased, the number of contacts between the CNTs in the sensing layer and AgNPs on the electrode increased rapidly, and the resistance decreased to 1.38 kΩ at a pressure of 500 kPa. The sensitivity of the tactile sensor was investigated by measuring the change in current upon pressure application. The relative current change of the sensor with respect to the pressure applied presents two linear regions: 6.02 kPa⁻¹ at pressures below 40 kPa and 1.31 kPa⁻¹ above 40 kPa (Fig. 3d). At lower pressures, the contact area between the sensing layer and electrode increased rapidly because of the cotton’s hierarchical structure, formed by the fiber bundles and the weaving of yarns. The contact area, according to the pressure, greatly varied compared to the surface contact area, showing great sensitivity³⁵. However, as the pressure continued to increase, this effect saturated. In particular, as the pressure increased, the number of contact points between CNTs in the sensing layer increased, which reduced the rate of change in resistance, thereby reducing the relative current change and lowering the sensitivity. Nevertheless, the sensitivity of over 1 kPa⁻¹ up to 500 kPa outperforms previous textile-based tactile sensor studies^20,23,56,57. The fabricated tactile sensor was confirmed to exhibit high repeatability by maintaining the same performance after 1000 loading-unloading cycles at a pressure of 300 kPa. The reproducibility was validated by simultaneously printing the electrodes and sensing devices as an array, and four sensors were randomly selected to test the uniformity. All four sensors exhibited identical performances, demonstrating the feasibility of uniform performance of the wearable devices fabricated by the proposed inkjet printing process (Supplementary Fig. 11).

Tactile sensors require fast response and relaxation times to monitor physical activities in real-time. The sensor exhibited similar response and relaxation times of 90 and 110 ms, respectively, at a pressure of 100 Pa with a slightly longer relaxation time (Fig. 3e). Additionally, the limits of pressure detection were significantly lower at 10 Pa owing to the rough and hierarchical surface structure (Fig. 3f). This feature enables the sensing of minute movements, such as breathing and pulse, when used for healthcare monitoring^58,59,60. Figure 3g shows that the relative current change was almost unchanged when a pressure of 300 kPa was loaded and unloaded for 3000 cycles (1 cycle per second), confirming the durability of the sensor. Therefore, the fabricated tactile sensor can be applied as a wearable device in various pressurized environments.

Stability after washing/ironing and motion sensing demonstration

Printed e-textiles integrated into clothing must maintain their performance under various environmental conditions to which clothing may be exposed (e.g., varying temperatures, washing, and ironing). The fabricated tactile sensor was tested on a Peltier element with an applied pressure and varying operating temperature from 30 to 60 °C (Fig. 4a). In the relevant temperature range, the temperature coefficient of the sensing layer is 0.9% °C^−1 61, and that of the electrode layer is lower than 0.1% °C⁻¹, which is smaller than the change in resistance due to pressure (Supplementary Fig. 12). Thus, there was no significant difference in the relative current change measured under varying pressures within the temperature range.

Fig. 4: Consistent sensing performance under various conditions.

a Relative current change under applied pressure with varying working temperature by a Peltier device. b Relative current change after multiple ironing cycles (~200 °C). c Sensing performance after 1000 bending cycles with bending radius of 1 mm. d Comparison of sensing performance before and after washing. Although degradation occurred, the performance was maintained after the second cycle. e Long-term storage of 5 months without environmental control. f Walking and running motion monitoring with a tactile sock. Data in (a–e) are presented as means ± standard deviations (n = 5). Source data are provided as a Source Data file.

The heat resistance of the sensor was tested by ironing nearly at 200 °C for 1 min five times. No degradation in the pressure-sensing performance was observed (Fig. 4b). Cotton fabric is intrinsically a highly heat-resistant fabric with a low thermal expansion coefficient of 25 × 10⁻⁵ °C⁻¹ below 200 °C⁶². The implementation of the proposed inkjet printing process facilitated the sensor’s high heat resistance by preserving the inherent properties of the cotton fabric.

For practical use in clothing, wearable electronics should have flexibility and bendability to match body movements, such as joint bending. The bendability of the sensor was tested by the implementation of 1000 bending cycles with a bending radius of 1 mm (Supplementary Fig. 13). During this repeated bending, the pressure sensing was performed at 500 kPa every 250 cycles. The variation in the relative current change measured during repeated bending cycles is 1.40%, confirming reliable stability (Fig. 4c).

Textile-based wearable sensors should exhibit washability because clothes can be contaminated or can absorb sweat, necessitating washing for maintenance. Accordingly, the sensor’s washability was tested through up to 10 wash cycles with 20 g of detergent and a total load of 2 kg, according to ISO 6330:2021, the international standard for textile washing in commercial drum washing machines⁶³ (Supplementary Table 3). The relative current change slightly decreased after the first wash but stabilized thereafter (Fig. 4d), maintaining until the 6^th wash. After additional washing, the relative current change gradually decreased from an average of 10.7% to 23.2% across the entire pressure range compared to before washing. The initial reduction occurred because of the detachment of some of the deposited nanomaterials during washing. However, the adhesion between the cotton and nanomaterials, enhanced by PLL, prevented further desorption, and the sensing performance stabilized. From the 7^th wash onwards, repeated abrasion by washing reduced the adhesive strength, which resulted in decreased pressure sensing performance upon further washing, and after the 10^th wash, the measured relative current change decreased by up to 38% compared to before washing. To further verify the effect of PLL treatment on improving washability, the pressure sensing performance after washing was compared with that when PLL treatment was not applied. The measured relative current change decreased by more than 50% after the first wash when PLL was not coated because some CNTs separated from the surface without a PLL coating after washing, whereas the CNTs remained intact with the PLL coating (Supplementary Figs. 14, 15). This demonstrated the use of PLL as an adhesion promoter to ensure the washability of e-textiles.

Figure 4e shows the consistent performance of the tactile sensor after being left on a shelf for five months without any special environmental control. This indicates that e-textiles integrated into clothing can be stably reusable after long-term storage. The fabricated tactile sensor was attached to a sock to monitor human walking behavior. Walking and running were distinguished based on the peak and width of the current measured by the sensor. In particular, walking has a low peak and wide width, whereas running has a high peak and narrow width (Fig. 4f).

Heater and heating glove demonstration

E-textiles produced by the proposed inkjet-printing process can be utilized as Joule heaters to provide warmth or heat therapy. The surface temperature increased monotonically up to a maximum value of 134 °C as the power was increased from 0 to 2148 mW (Fig. 5a). The heating and cooling speeds of the heater were assessed by applying varying powers to 130, 200, 310, 360, 470, 600, and 670 mW, prompting the temperature to rise of 30, 35, 40, 45, 50, 55, and 60 °C, respectively. The temperature quickly converged and returned to its initial value when the power supply was removed (Fig. 5b). This ensured that the heater could be controlled accurately and rapidly to reach a set temperature. We also investigated the temperature and resistance of the heater during bending. Figure 5c shows the temperature and resistance changes in the heater at an applied power of 470 mW while bending the heater from the flat state to the bending radius of 1 mm. The temperature was maintained at an average value of 49.98 °C with a variation of 2.38%. Moreover, the heater showed a consistent resistance of 168.77 Ω with a variation of 2.15%.

Fig. 5: Characterization of fabricated Joule heater.

a Measured heater temperature by an IR camera at various applied power. b Transient temperature change of the heater at different power levels (130, 200, 310, 360, 470, 600, and 670 mW). c Constant temperature and resistance of the heater with varying bending radius at an applied power of 470 mW. d, e Optical and SEM images of the heater at flat and bent (r = 1 mm) state, respectively. f Consistent temperature and power of a heater glove during hand clenching and unclenching. The inserted IR images show the temperatures of the heater according to the hand postures. The scale bars are 3 cm. Data in (c, f) are presented as means ± standard deviations (n = 5). Source data are provided as a Source Data file.

Full size image

The heater stability was elucidated by analyzing the SEM images of the electrode in both flat and bent (r = 1 mm) states (Fig. 5d, e). In the flat state, the AgNPs were coated on the entire surface of the cotton fabric, and the CNTs were distributed between them, forming an overall conductive path. When bent, even if the cotton yarns are spaced further apart than in the flat state, the delamination of the AgNPs did not occur, and their contact with the CNTs was maintained. Therefore, the CNTs sustained a conductive path with the AgNPs even under mechanical deformation, ensuring resistance stability and device suitability for wearable applications. The temperature change of the heater according to the bending radius was captured using an IR camera (Supplementary Fig. 16). Almost no temperature change was noted when comparing the device performance at its flat state and with the bending radius of 4 mm and 1 mm. Also, the temperature was consistent during repeated bending cycles (Supplementary Movie 1).

The heater was attached to a glove to demonstrate its applicability in wearable heat therapy devices. The heater electrodes were fabricated using the same printing process, and the heater was attached to a glove by sewing (Supplementary Fig. 17). The fabricated heater glove is shown in Supplementary Fig. 18, where the heating part of the fabricated heater covered the entire palm with an area of 5.5 cm × 4.5 cm. The temperature and power of the heater were measured while repeatedly clenching and unclenching the gloved hand (Fig. 5f and Supplementary Movie 2). The temperature was uniform along all paths, as shown in the insets of the IR images. The average temperature was 50.10 °C with a variation of 1.93%, and the average power of 498.02 mW was maintained with a variation of 2.36%. During clenching, the heater folds (r < 1 mm) along the palm creases. However, the adequate bendability of the electrode achieved the same continuous thermal sensation.