Wearable bioelectronics is an emerging field of science that revolves around the interface of physics1, electronics2, material science3, chemistry4, medicine5, and engineering. The development of the field has already contributed to general healthcare6, sport medicine7, chronic pulmonary monitoring8, diagnosis of heart and cardiovascular disorders9, diabetes6, cystic fibrosis10, etc. Along these lines, we can observe the evolution of biomedical instruments: the hard, rigid metals and silicon-based electronic elements are being slowly but persistently replaced by novel, softer, and lightweight skin-like materials with equal functional load11,12,13. Reducing the complexity of biomedical tools, making them easily accessible and ultimately wearable, helps to develop an ambulatory health-monitoring network. Moreover, day-to-day health monitoring in asymptomatic subjects in a long-term systematic study will help gather big data that can be used to further build disease prediction algorithms14,15.

One particularly promising material for wearable biomedical application is graphene, due to its electrical, mechanical, and optical qualities16,17,18. Graphene electronic tattoos were first reported in 2017, pioneering the field of atomically thin wearable bioelectronics elements19,20. The previously reported GETs feature transparency of ~85%, stretchability of more than 40%, and the ability to self-adhere to the skin via van der Waals forces19. The ability of GETs to sustain stretching up to 40% allows us to operate them on the skin without significant consequences from motion-induced noise and strain-induced performance artifacts19,21. However, these GETs have been made on research-grade quality graphene supply at a considerably small scale, and the major properties such as the interface impedance, sheet resistance, and water vapor transmission rate (WVTR) have not been explicitly optimized. These previously reported GETs suffers from a few major drawbacks. First, the monolayer graphene tattoos had rather average electronic performance, with impedance higher than commercial gel electrodes. While monolayer CVD-grown graphene is a good charge carrier conductor, its properties usually degrade when the graphene sizes are up-scaled to a few ~cm, like in our GET technology. Second, minor inconsistencies such as grain boundaries or large cracks and folds appearing during the graphene transfer drastically reduce the overall performance. Additionally, these inconsistencies contribute to the large performance distribution of the monolayer graphene tattoos: while one GET might have superior properties, the neighboring tattoos can be severely damaged due to a small crack in the monolayer graphene surface. Moreover, the properties vary largely on the graphene supplier20. Finally, the previously used GETs were non-transparent to gas and water vapors, therefore limiting the natural perspiration processes when placed on the skin.

In this work, we report on the GETs 2.0: the improved multilayer graphene tattoos with superior electrical properties and efficient sweat perspiration capabilities. The findings reported in this work are the results of a comprehensive analysis of electrical (i.e., sheet resistance) and bioelectronic (i.e., skin impedance, sweat susceptibility) properties of the multilayer GETs. The first major influential parameter is the number of graphene monolayers per tattoo. The large distribution of monolayer (1 L)-GET electrical properties is associated with the high density of microcracks and defects present in the monolayer graphene due to imperfect growth and transfer conditions. Hence, alongside the 1L-GETs, we created multilayer GETs, namely bilayer (2 L)-GETs, and trilayer (3 L)-GETs. The 3L-GETs show 2.5-fold improvement in the interface impedance and 3.5-fold improvement in sheet resistance compared to the 1L-GETs. Most importantly, we show that the 2L-GETs and 3L-GETs performance deviation is much lower in contrast to the 1L-GETs. Besides, we show that even 1L-GETs develop enhanced performance (lower sheet resistance, decreased skin impedance, and lowed distribution of the values) with the self-assembly of graphene nanoscrolls (GNS) into their structure. Microsized holes, embossed in the topological structure of the GETs 2.0, cover the major drawback of the old GETs: their sweat impermeability. The holey GETs 2.0 are exceptional in WVTR up to ~3200 g/m2 per day without degradation in their electronic properties. Finally, we show that microholes allow for advanced multi-planar contact with graphene tattoos, opening up the roads for next-generation hybrid wearable systems such as skin-wearable electronic heaters. The GETs 2.0 are robust, electrically superior, lightweight, and breathable (see comparison Supplementary Table 1 and examples of the GETs placed on human skin and on the flower, see Supplementary Figs. 12).


One primary parameter of the passive wearable bioelectrodes is the electrical interface with the human body22. The interface, commonly known as the electrode-skin impedance or tattoo-to-skin impedance, is the complex interpretation of the charge transfer occurring at the contact. This contact interface impedance is frequency-dependent and, essentially, the consequence of parallel capacitances at the electrode-skin interface. An infinitely low electrode-skin impedance is desired, yet impossible. At high frequencies (>100 kHz), the contact impedances are typically low; hence little signal attenuation happens. At lower frequencies (<10 kHz), the contact impedance plays a significant role in the signal recording fidelity. The impedance of the dermal and subdermal layers is typically a few orders of magnitude lower than the electrode-skin impedance; hence it can be discarded from consideration. In this regard, it is essential to develop a wearable bioelectrode system with the lowest interface impedance attainable23,24,25,26,27,28. The modern medical-grade bioelectrodes are typically made of sticky gels and hydrogel electrolytes since they afford conformal contact with the skin’s entire surface. The gels penetrate the skin’s microscopic structure and conformably coat the whole surface, see Supplementary Fig. 3. In contrast to the gel electrodes, the rigid, thick, metal-based electrodes feature high electrical conductivity but yield a non-perfect contact with the skin. Trapped air between the electrode and skin (in the skin’s grooves and ridges) generally reduces the quality of electrical contact and increases device-to-device property variability.

On the other hand, atomically thin electronic tattoos are essentially unique structures since they form an imperceptibly superior contact with the skin’s microstructure and follow the skin’s grooves and ridges (see Supplementary Figs. 13). When comparing the three types of electrodes (bulk conductor, gel, and atomically thin tattoo), one can notice that in the first two components, the material conductivity does not play a substantial role in charge transport since the charge distributes through the bulk material evenly before being injected in/out of the tissue. However, in atomically thin conductors, such as graphene, the charge carriers must pass through the whole in-plane atomic lattice of the graphene prior to being injected into the tissue. Hence, we hypothesize that in-plane conductivity plays a significant role in understanding the performance of atomically thin graphene e-tattoos. It has also come to our attention that if the backside graphene (from the as-grown graphene-on-copper source foil) is not burned away by oxygen plasma, then during the wet etching of copper, the unsupported backside graphene rolls onto itself and self-assembles into the so-called graphene nanoscrolls (GNS), as coined by Liu et al.29 These GNS then attach to the PMMA-supported graphene electrostatically and form additional features that contribute to the tattoos’ electrical and mechanical properties. Thus, we hypothesized that GNS would improve the electrical properties of the tattoos and help reduce the device-device variability.

Multilayer GETs

With the major quest of improving the performance of atomically thin GETs, we architected monolayer (1 L), bilayer (2 L), and trilayer (3 L) graphene tattoo structures. The 1L-GETs were made with and without GNS, and 2L-GETs tattoos were fabricated in four kinds: with GNS present at both graphene monolayers, with GNS present at either (x2) of the graphene planes, and without GNS (See Fig. 1c). The 3L-GETs were made only with GNSx3 configuration since the time and efforts required to fabricate 2 L and 3 L GETs increase exponentially. The graphene tattoos were cut into desired shapes and transferred to the skin for impedance monitoring, as thoroughly explained in the protocol manuscript20 and the Methods section. Shortly, for the impedance monitoring, the GETs were all cut into dumbbell-doughnut-shaped structures with a skin-intact area of ~25 mm2 to provide a just comparison (see Supplementary Fig. 4). The multilayer graphene structures are made by multi-stacking monolayer CVD-grown graphene during the wet etching process (see Methods). The highest-quality CVD-grown graphene is now an affordable material with prices down to 0.5 $/cm2 (see Supplementary Table 2) and is expected to further decrease significantly. The 1 L, 2 L, and 3 L graphene structures were also transferred on SiO2/Si by etching off the PMMA after transfer. These samples were studied via scanning electron microscopy (SEM), atomic force microscopy (AFM), and Raman spectroscopy. Raman spectroscopy (see Supplementary Fig. 5) of the multilayer graphene structures has revealed no specific bilayer stacking (no broadening of the 2D peak has been found), supporting the claim of having two individual CVD-grown monolayers of graphene stacked one on top of another. Since the Raman spectra were taken from the “clean” areas without GNS, the I2D/IG peak ratio is in the range of 1.2–2.0, suggesting high-quality monolayer graphene30. The ID/IG peak ratio is usually well below 0.04 (see Supplementary Table 3 and Supplementary Fig. 5), suggesting a low density of defects. Interestingly, when imaging the bilayer graphene structures, we stumbled upon certain areas where IG/I2D peak ratios would rise above 10. The phenomenon can be explained by the ‘magic angle’ when two graphene monolayers form a resonant state at ~12° angle31. Since the CVD-grown graphene is polycrystalline, there is a finite chance that two heterostructured graphene monolayers would have ~12° angle between their planes in unique physical properties, which is known to cause such Raman response31. It might be interesting to study the “magic-angled” bilayer graphene for wearable applications; however, it would require large-scale single-crystal stacked graphene.

Fig. 1: Graphene layers and effect of nanoscrolls on GETs 2.0 performance.
figure 1

a Statistical average of tattoo-skin Impedance for 7 combinations of GETs and comparison with gold and Ag/AgCl gel electrodes. b Statistical average of GET’s sheet resistance for seven combinations of GETs and comparison with gold. 1L-GETs marked in ice blue, 2L-GETs in light green, and 3L-GETs in red. All GETs are 25 mm2 in area. The box represents 25% and 75% with a mean, whiskers are ±SD, and stars represent the best examples. c The schematic illustration of the GET combinations: 1 L, 2 L, and 3 L with and without GNS. d The performance benchmarking of 1L-GETs (combined), 2L-GETs (combined), and 3L-GETs in terms of sheet resistance and GET-skin impedance.

As shown in Fig. 1a, b, there is a clear impact of GNS presence on the 1L-GET performance in terms of skin impedance and sheet resistance. As hypothesized, the presence of the GNS helps to reinforce the electrical qualities of the monolayer graphene and helps to mitigate the performance variability arising from the monolayer graphene’s imperfections, cracks, folds, or grain boundaries. As we have found, the presence of GNS improves the skin impedance of the 1L-GET by ~20% on average and improves the sheet resistance by ~45%. In opposite, GNS introduces some minor disorder to the system, sightly increasing the SD of the sheet resistance. However, the important factor is that the average sheet resistance drops almost twice (from 2686 Ohm/sq. to 1456l Ohm/sq.), contributing to the reduction in the interface impedance (see Supplementary Table 3 for details). The GNS were examined with SEM and AFM, see Fig. 2. The SEM images confirm that the nanoscrolls are often tangled up and knotted. Length estimation is somewhat complex: some GNS are 1–20 µm long, but often they form a largely intercalated network that converges for 100’s µm (See Fig. 2 and Supplementary Fig. 6). We then used AFM to estimate the average diameter of the GNS to be ~32 ± 11 nm. We found that the average GNS diameter does not depend on the number of GNS layers in the GET stack (see Supplementary Fig. 7). These findings are well aligned with the previously reported height range of 10–100 nm29.

Fig. 2: Structural and topological characterization of the graphene nanoscrolls.
figure 2

a, b SEM and AFM images of the clean 2 L graphene. c, d SEM and AFM images of the 2 L graphene with GNS.

We carefully evaluated the four kinds of 2L-GET structures with the purpose of finding an optimal or superior combination of graphene layer and graphene nanoscrolls. However, as one can notice from Fig. 1a, b, and Supplementary Table 3, there is no distinct difference between these four 2L-graphene tattoo types, including the one without GNS, or comprised of two clean graphene monolayers. This means that the 2L-GET structure performance is saturated by the two graphene layers and does not get drastically improved with the further addition of graphene nanoscrolls. In these bilayer structures, the two graphene monolayers work in synergy and conduct the electrical potentials in a unified manner: when one monolayer has a defect, the charge carriers jump into the other monolayer, and vice versa.

The 3L-GET structures were made in the only combination: with three layers of GNS to reduce the efforts required for such 3L-GET fabrication. As one can see from Fig. 1a, b, the 3L-GET devices exhibit the lowest interface impedance (6.8 ± 0.6 kOhm @ 10 kHz) and the lowest sheet resistance (~0.55 ± 0.25' kOhm/sq.). Interface impedance at 10 kHz frequency is taken for benchmarking since it is a common carrier frequency for tissue bioimpedance characterization and has been widely used to describe electrode’s performance32,33,34. As one can notice from the benchmarking Fig. 1d, there is a clear trend in the electronic properties of the GETs with respect to the # of graphene monolayers present. However, the trend is not even linear: the performance enhancement between 3L-GETs and 2L-GETs is only ~17%, while the 2 L GETs are almost two times better than 1L-GETs. Hence, we suggest that 2L-GETs are the most suitable as functioning components of future graphene-based electronic tattoos in terms of performance, device-to-device variability, and efforts required for device fabrication. Furthermore, one essential qualitative attribute of graphene tattoos is transferred fidelity20. We noticed experimentally that 1L-GETs, especially the ones without GNSs are much more difficult to transfer from the temporary paper onto the skin. The GETs made from 2 L and 3 L graphene are significantly easier to slide off the tattoo paper, resulting in a faster and more reliable transfer with almost ~100% yield. We hypothesize that it can be associated with higher Van der Waals adhesion forces between atomically flat graphene layer and the temporary tattoo paper, compared to 1L-GET with nanoscrolls and multilayer 2 L and 3 L graphene structures.

Additionally, 2L-GETs with GNS have been made with different channel areas, 3 × 3mm2, 5 × 5 mm2, 7 × 7 mm2, and 10 × 10 mm2 (see Supplementary Fig. 14). As anticipated from the performance of thin-film conductors, the average channel resistance depends on the geometrical aspect ratio of the device rather than its dimension. Such, the average channel sheet resistance for these samples is in the range of 500 ± 100 Ohm/sq. Interestingly, we have found that contact resistance, RC has a significant trend to reduce with larger channel width (from ~350 Ohm for 3 mm wide samples to ~260 Ohm for 5 mm wide sample and down to ~190 Ohm for 10 mm wide samples, see Supplementary Table 6).

The cumulative and averaged impedance and phase Bode diagrams of the seven GET kinds are shown in Supplementary Fig. 8. While the magnitude of the impedance at 10 kHz was taken as the figure-of-merit for the above discussion, there is another noteworthy distinction that is apparent. As one can see from phase diagrams in Supplementary Fig. 8, the 1L-GETs feature a phase increase beyond 45 degrees at high frequencies, which is not visible in the case of 2 L and 3 L GETs. Qualitatively, this difference in the phase behavior leads us to conclude that the contact interface between the 2L-GETs and 3L-GETs GETs can be considered more nonpolarizable and much more efficient than the 1L-GETs22. It is important to note that GETs were made with graphene received from a single supplier. Furthermore, the data used in this manuscript were collected from at least three different batches of graphene from the supplier and found no performance variations between the graphene batch and its quality.


The essential qualities of graphene electronic tattoos are hindered in their nature: being atomically thin, the tattoos are insensible when worn by the subjects and imperceptible to sight and touch. Typically, the previously made electronic tattoos covered the skin entirely and did not allow the area under the tattoos to breathe and perspire normally. However, here we introduce graphene tattoos 2.0 with holey morphology. The micro-holey GET structures allow for advanced skin breathability and sweat excretion without a drastic reduction in electronic properties. The holes within the GET structure were made via mechanical perforation by the same mechanical plotter used to carve the graphene tattoos. The array (~1 mm pitch) of triangular-shaped holes (with 50–100 µm dimensions, see Supplementary Fig. 9) is directly embossed into the GET structure while being supported by the tattoo paper. First of all, it was essential to evaluate whether the introduction of these microholes decreases the electronic properties of the tattoos; hence we fabricated the same 1 L, 2 L, and 3 L GET structures with holes. As one can observe from Fig. 2c and Supplementary Table 5, there is only a minor to zero increase in the GET impedance associated with the holes. The effect is more prominent in the case of 1L-GET (17%), less visible for 2L-GETs (6%), and non-existent for 3L-GETs (no increase).

To evaluate the gas and water vapor transparencies of the GETs 2.0, we transferred the holey (N = 7), and normal (N = 7) 1L-GETs on top of small plastic vials containing water. When transferred, the GETs establish a firm seal around the opening, capping the vial. We monitored water loss within those vials for 2 weeks (see Supplementary Fig. 10) and compared its dynamics to fully open vials (without any cap, N = 3). Qualitatively, as one can see from Fig. 3a, b, when the vials are heated to ~50 °C for an extended period, water condensates inside the vials capped with normal GETs. The condensation does not happen with holey GETs, suggesting that the water vapors transport unobstructed through the holes (see Supplementary Figs. 1011 for more representative photographs). Quantitatively, as shown in Fig. 3d, the holey GETs are almost as efficient in water evaporation as entirely open vials, with WVTR of 2770 ± 494 g/m2/day against 3600 ± 600 g/m2/day. On the other hand, the normal GETs feature a twice smaller yet surprisingly tangible WVTR rate, 1400 ± 350 g/m2/day, which can be explained by large-area graphene having micro-defects, cracks, and holes. Supplementary Table 1 provides a comparison of the values to other state-of-the-art wearables.

Fig. 3: Comparison between normal GETs and holey GETs.
figure 3

a Photograph of a vial capped with normal GET, impeding normal evaporation of the water trapped inside. b Photograph of a vial capped with holey GET, without water condensation on the surface, suggesting that it is efficiently evaporated. c The change in GET performance (tattoo-skin impedance) of the 1 L (ice blue), 2 L (light green), and 3 L (red) GETs with and without holes. The box represents 25% and 75% with a mean, whiskers are ±SD, and stars represent the best examples. d The water vapor evaporation rate difference between a closed vial (with cap), capped with a normal 1L-GET (blue), capped with a holey 1L-GET (dotted blue), and an entirely open one (gray). The vials and their water content were monitored for 7 days, and the average data per 24 h is shown. See Supplementary Fig. 10 for complete water loss dynamics. The box represents 25% and 75% with a mean, whiskers are ±SD.

One peculiar observation that we consider essential for the future usage of graphene tattoos, has risen from a failed experiment of placing one GET on top of another GET. Ideally, we hypothesized the PMMA support to act as the dielectric between two graphene layers. Surprisingly, however, we found that when two GETs, both having graphene facing down, are placed on top of each other, there is a finite electrical contact. We further explored the phenomena by placing the GETs in a reversed configuration on top of the 4-point transfer length measurement (TLM) structures. The TLM structures consist of a glass slide covered with an Ecoflex and four conductive gold feedlines (see Supplementary Fig. 13 for details). One can see the sketches of a “normal” GET placement on top of the 4-probe structure in Fig. 4a, and the according to TLM measurement result in Fig. 4b. Figure 4c shows the schematic of the reversed GET placement. We hypothesize that this effect is caused by the folding of the tattoos at the edges of its structure. Furthermore, using holey GETs enhances the effect. Since the holes are formed by means of mechanical embossing, the “embossed” area is not permanently removed, but rather ‘flaps’ back and attaches to the backside of the graphene tattoo, opening additional density of graphene exposed to electrical contacts (see Supplementary Fig. 12). As visible from Fig. 4d, the electrical contact is satisfactory, and the sheet resistance is on par with the normal GET placement. This effect is essential for providing operational contact to tattoos. Previously, in order to contact the GETs, we have been using conductive gold adhesive tape that had to be placed before the tattoo. Even if the gold tape is ultrathin (~10 µm in our case), it imposes a rather steep slope for a tattoo to bend upon. We believe this step was among the major reasons for a reduced lifetime of the system (typically < 24 h of continuous wear), and we considered the tattoo-tape connection the weakest link in the overall tattoo configuration. Currently, with the understanding that graphene tattoos can simply be biased through the opposite side, it will be much easier to establish more robust, long-term, and hybrid connectors to the tattoos. We envision that the GETs 2.0 can now be integrated with complex and non-flexible systems such as smartwatches by establishing intermittent backside contact tapping without affecting the structural integrity of the tattoos.

Fig. 4: Sheet resistance and electrical contacts to Normal GETs and Reversed GETs.
figure 4

a, b An example of the 4-probe TLM measurement results with standard graphene tattoo placement (graphene down). c The opposite situation: the GET is placed upside-down with graphene up. However, the tattoo is not perfectly flat, and the edges and corners of the GETs typically fold during transfer. Hence, even when GET is placed with graphene up, a few graphene areas face down, establishing adequate electrical contact with the underlying gold. d The TLM measurement for one of such devices yields the Rs of 638 Ohm/sq. The average Rs for all such devices (N = 2) is 555 ± 116 Ohm/sq.

Skin-like heaters

Besides using the GETs for electrophysiological sensing, here we thought of employing the GETs towards bidirectional communication with a human body, and most precisely – towards skin-conformable heating. Human organisms are incredibly susceptible to minor changes in the surrounding environment, and even a small reduction of the temperature might lead to cell death, organ dysfunction, and even death. On occasions when wearing additional layers of clothes is not feasible or counterproductive (e.g., military or athletics), or perhaps when only certain areas of skin must be actively heated up, graphene tattoos will be highly suitable. This is essentially feasible for applications on the face (e.g., during skiing or just in extreme cold weather conditions), since graphene is almost invisible to the naked eye. Besides skin-wearable heaters, the GET-based heaters can be used in a variety of applications, including automotive (e.g., windshield heating), outer space thermal management systems, and clinical applications (e.g., cardiac electrotherapy). In this work, we tested the applicability of the 1 L, 2 L, and 3 L GETs as skin-conformable heating elements. For the experiments, graphene tattoo stripes of approximately 5 × 8 mm2 placed on the 4-probe conductive gold tape were used (see Supplementary Fig. 13), and constant DC voltage (5 V, 10 V, 15 V, 20 V, and occasionally 30 V) was applied. The temperature was read using the desktop optical thermal camera (FLIR ETS320). The voltages were applied for the duration of 1 min with 1 min “Off” for cooling down. An example of a temperature trace is shown in Fig. 5a, b featuring the temperature rise and fall profiles upon the gradual heating of the tattoo when DC voltage is applied. As mentioned earlier, the majority of the tattoos have been exposed to DC potentials ranging from 5 V to 20 V. Occasionally, when using 1 L or 2 L GETs, we had to apply up to 30 V of the potentials in order to induce heating above 50 °C. Exemplary static images are shown in Supplementary Fig. 15, and the full sequence video profiles can be seen in the Supplementary Movie 1. As one can clearly observe from Fig. 5c, for each applied potential, the temperature difference is much higher with an increased number of graphene monolayers in the tattoo structure. Figure 5c features the effective difference in temperature caused by the applied DC potential. For the actual maximum temperature reached during the heating, see Supplementary Fig. 17. Interestingly, while the 3L-GETs have features the highest temperature difference per applied voltage (up to ~4 °C/V). However, the heating efficiency of the tattoos was evaluated as the power (in Watts) required to raise the temperature for 1 °C and takes into account the channel resistance. Surprisingly, the average heating efficiency is not drastically different for 1 L (6.2 ± 1.2 mW/°C), 2 L (6.2 ± 1.3 mW/°C), 2 L reversed (6.5 ± 1.0 mW/°C), and 3 L (5.6 ± 1.0 mW/°C) GETs, which means the effect is primarily due to Joule heating (see Fig. 5c–e and Supplementary Table 7).

Fig. 5: GETs wearable heater efficiency.
figure 5

a Real time-temperature profile of a 2L-GET heated by applying 5 V, 10 V, 15 V, 20 V, and 30 V consequently. b Real temperature measured for a 2L-GET, overlapped. c The overall heating efficiency for 1L-GETs (blue, N = 11), 2L-GETs (green, N = 17), and 3L-GETs (red, N = 19), with y scale representing the effective raise in temperature rather than the real temperature. d Direct comparison of the heating efficiency between the 2L-GETs and 2L-reversed GETs. The box represents 25% and 75% with a mean, whiskers are ±SD. e The heating efficiency of the 1L-GETs (blue hexagons), 2L-GETs (dark green upwards triangles), 2L-reversed GETs (light green downwards triangles), and 3L-GETs (red circles) at different applied DC potential.

Finally, we tested the heating efficiency of 2L-GET channels with different sizes, namely 3 × 3 mm2 (N = 4), 5 × 5 mm2 (N = 4), 7 × 7 mm2 (N = 4), and 10 × 10 mm2 (N = 4). Since the devices have, on average, equal resistance, we anticipate almost equal heating efficiency. As one can see from Supplementary Fig. 16 and Supplementary Table 8, our hypothesis holds with a few exceptions. First, the absolute energy per device seems to have a downwards trend with the decreased area (3 mW/°C for 9 mm2 device vs. ~13 mW/°C for 100 mm2 device). However, when normalized per area, the trend is reversed. The small channel devices have on average a rather high heating efficiency of 36 ± 3 mW/°C/cm2, while it saturates for larger devices (25–100 mm2) at the range of 13–17 mW/°C/cm2, which means that heating can be done at larger sample dimensions in the future if required per application.

3L-GETs were not heated with 30 V, as such potentials caused excessive (>30 mA) currents through graphene lattice, which are destructive, and the devices often “burned” during the process. As observed through the thermal camera, the devices are more susceptible to such ‘burning’ when the electrical contact is not uniform, but a few hotspots exist. The hotspots cause excessive current density flowing through certain GET areas, consequently destroying the GETs at higher applied potentials. Remarkably, the temperature rise and fall profiles are substantially indifferent from the tattoo structure nor from the applied potentials (see Supplementary Fig. 18). On average, the temperature rise and fall time constants are within ~15–25 s range. This means that the final application of GETs as heaters can be made by short DC or AC voltage pulses. Together with placing GETs with PMMA side down to the skin make the system safe for on-skin application. The tattoo heating up and cooling down profiles are only substantially different (longer) in the case of reversed GETs (see Supplementary Fig. 19).

Furthermore, we show negligible, and even beneficial effects of sweat and motion on the GET performance. As one can see from Supplementary Fig. 21, a long (1 h) sweat-inducing outdoor walk (motion) at ~32 °C temperature results in slight (~25%) reduction (improvement) of interface impedance, suggesting a more superior interface due to hydrated tissue or build-up of conductive sweat at the graphene-skin interface.

Temperature sensing

At last, the GETs have been tested for their temperature sensing capabilities. In the COVID-19-affected pandemic world, it became obvious that on-demand ambulatory monitoring of human body temperature is an essential requirement for personal health status monitoring. While the temperature measured from the skin is not entirely representative of inner body temperature, a clear link between these two has been established35. Moreover, a long-term on-demand monitoring of the person’s skin temperature dynamics is needed for disease prediction and prevention. The temperature sensing performance of the seven GET kinds is displayed in Fig. 6. In contrast to the above-presented applications, no clear trend in temperature sensing has been found. The temperature electrical coefficient (TEC, α) was evaluated as the slope of the relative graphene resistance change (ΔR/R0) relative to the change in temperature (ΔT, °C). Examples of the actual relative resistance increase and decrease profiles with temperature are shown in Supplementary Fig. 20. The graphene samples were fixed on the skin replica and placed on a hotplate with a thermocouple connected for correct temperature control. 50 mV DC voltage was applied to the GET stripe, and the resistance was monitored continuously with a sampling rate of ~1 Hz. Figure 6a shows the TEC coefficients for all seven GET types when heated up and cooled down. Clearly, in all cases, the “cooling down” profile has a lesser slope, perhaps with the only difference of 3L-GETs, when the difference is not significant. The TEC for the graphene tattoos in the heating up experiments range from 1–3 × 10–3, while the same samples experience the TEC in the range of 0.5–2.0 × 10–3 when cooling down. We believe that the difference can be caused by the current-induced “annealing” of the graphene36, which is a constant process leading to the increased graphene resistance in time, contributing to (1) an increase in the heating coefficient slope, and (2) a decrease in the cooling profile. The other major and important observation is that 1L-GETs and 2L-GETs have a rather large SD of their performance (above 0.5 × 10–3), whilst 3L-GETs result in a much more uniform data distribution. Hence, in order to avoid complications with temperature sensing by means of the GETs, we believe the 3L-GETs are the most suitable candidates for the job. Supplementary Table 1 provides a comprehensive comparison of the GETs as wearable heaters, sensors, and electrodes.

Fig. 6: Temperature sensing performance of the GETs.
figure 6

a average values of the TEC alpha coefficient for seven combinations of the 1 L, 2 L, and 3L-GETs with temperature rise up (blue) and cool down (orange). b Combined TEC alpha coefficients for GETs when ramping temperature upwards. c Combined TEC alpha coefficients for GETs when cooling GETs down. The box represents 25% and 75% with a mean, whiskers are ±SD.


In conclusion, we present in this work the major improvement in the graphene tattoo technology, namely multipurpose and robust GETs 2.0: the lightweight and unobtrusive wearable tattoo sensors with superior electrical properties and permeability to sweat (see Supplementary Table 1). We show that the introduction of graphene nanoscrolls into the GET 2.0 structure makes the monolayer tattoos more robust, electrically conductive, and resilient. The presence of GNS on 1L-GET results in substantially (i) decreased sheet resistance, (ii) decreased skin impedance, and (iii) lower standard deviation of these values. We stress here that the addition of GNS evidently improves the reproducibility of the results. The largest drawback of 1L-GETs is their susceptibility to minor cracks or disorders. These disorders might drastically reduce the performance of bulk graphene tattoos, resulting in a large SD. As shown in this work, GNS or an additional layer of graphene help enhance the graphene tattoos’ electrical properties by forming an intercalated network, bridging any possible imperfections of the bulk graphene layer, consequently improving the overall performance of the e-tattoos. Furthermore, we performed a thorough study of multilayer graphene tattoos, and show that 2 L and 3 L GETs are generally more robust and electrically favorable for the majority of wearable electrophysiological recordings such as brain waves (EEG), muscle (EMG), heart (ECG), or ocular (EOG) activities monitoring due to their superior electrical properties19,33,37,38. The 2L-GETs show an almost twofold improvement in the tattoo-skin impedance compared to the 1L-GETs. However, since 3L-GET assembly requires exponentially longer time, without substantial (just ~17%) improvement in the electrical properties compared to the 2L-GETs, we suggest that 2L-GETs (with or without GNS) should power up the next-generation graphene tattoos. Holey microstructures, introduced here in tattoo 2.0, cover the major drawback of the previously developed GETs: their sweat permeability. The holey GETs 2.0 are superior in water vapor transmission (WVTR up to 3300 g/m2/day) without major degradation in their electronic properties (especially in the case of 2 L and 3 L graphene structures). Furthermore, we show that microholes allow for advanced multi-planar contact with graphene tattoos. Formerly, we used to drive an electrical contact to the GET only from the graphene plane of the graphene/PMMA planar GET structure with graphene facing down. Now, we show that microfolds, edges, and especially the mechanically embossed holes allow for efficient electrical contact with the tattoo from the PMMA plane. This opens up the roads for next-generation hybrid wearable systems using GETs 2.0 as skin-interfaced elements, temporary use hybrid electronic connectors, and rigid contemporary electronic systems such as smartwatches that hold the battery source, measurement circuitry, and wireless communication systems. Finally, we show that multilayer tattoo structures can be used as efficient skin-wearable electronic heaters, exhibiting heating efficiency in the range of ~6 mW/°C.


Graphene growth normal GET preparation

Single-layer graphene-on copper (Cu) substrate grown by the chemical vapor deposition was purchased from Grolltex. To make graphene tattoos, the PMMA (950 A4) was spin-coated on top of the Cu/graphene substrate at 2500 rpm for 1 min and baked on a hotplate at 200°C for 20 min. The Cu/graphene/PMMA was then placed into ammonium persulfate solution ((NH4)2S2O8, 0.1 M) for 12 h to etch away Cu foil. The graphene/PMMA film was then transferred into the DI water three times to remove Cu etchant residuals. Once ready, the PMMA/graphene is finally transferred onto the temporary tattoo paper20. The paper has a special resin-based coating, making the paper slippery and anti-adhesive whenever wet yet adhesive when dry. The graphene/PMMA/paper stack is then dried overnight and loaded into a mechanical to dice any shape of the final GET. Typically, the entire GET area in the experiments is ~25 mm2 (see Supplementary Fig. 4).

Multilayer GET preparation

To yield multilayer graphene tattoos, we still rely upon three monolayers of CVD-grown graphene. The process starts with preparation. To make a 2 L GET, for example, it requires preparing one piece of PMMA/graphene/copper (‘piece A’) and etch the copper away as explained above. When the etching is complete, ‘piece A’ is transferred into clean DI water. Another piece of monolayer CVD graphene-on copper foil (without PMMA, ‘piece B’) is used to fish out ‘piece A’. It was then dried overnight. The (A)-(B) bilayer structure is then annealed at 200 °C for 10 min to reflow the PMMA and establish a good adhesion. Next, the second copper foil is etched, and the transfer on graphene tattoo is performed as explained above. To make a 3L-GET, one ‘piece A’ and two ‘pieces B’ are required, and the transfer/etch/transfer cycle is repeated multiple times. The final two-layer (A)-(B) or three-layer (A)-(B)-(B) stack, however, is morphologically distinct from directly CVD-grown multilayer graphene that has an extra interlayer stacking.

GETs with or without GNS

During CVD growth, graphene is also formed at the backside of the copper foil (usually with lower quality). If the backside graphene is not explicitly eliminated from the copper foil, the backside graphene rolls into itself during the copper etching and creates the so-called nanoscrolls29. To fabricate monolayer graphene with GNS, no additional preparation is required. To make a monolayer GET without GNS, the backside of copper must be exposed to oxygen plasma to etch the backside graphene off. The pattern of presence or absence of GNS in the multilayered structure can be architected by etching or not etching the backside graphene from the pieces ‘A’ and ‘B’.

Holey graphene preparation

Holey GETs were made from the normal GETs after they are fixed with the temporary tattoo paper. The stipple pattern feature of the Silhouette Cameo plotter was used for the step. As confirmed optically, each hole has a triangular shape (resulting from the pyramidal shape silhouette cameo plotter’s blade), the hole-hole pitch is ~1 mm, and each hole is ~3000 µm2 in area, see Supplementary Fig. 9.

On-skin measurements

In order to yield the thinnest possible electrical contacts that at the same time adhere to the skin, we utilized a 10 µm thick adhesive tape (Iwatani, ISR-BSMK10G) and evaporated a metal stack (Ni/Au, 10/90 nm) on top of it. The tape is supported by the PET backing layer, which is delaminated from PET right before placing the tape onto the skin. The conductive adhesive tape is transferred onto the skin, followed by the graphene transfer. The pre-structured GETs are soaked in DI water for ~20–30 min before proceeding with the transfer. The soaked GETs brought in tight contact with skin, followed by slight pressure application and sliding the tattoo paper aside20. The human subject measurements were performed under the approval of the Institutional Review Board of the University of Texas at Austin (IRB no. 2018-06-0058) with written informed consent. Whenever required, the GETs can be removed by using adhesive tape (see Supplementary Movie 2).

Evaporation experiments

N = 14 pieces of bilayer graphene tattoos (N = 7 normal GETs and N = 7 holey GETs), each approximately 10 mm by 10 mm in size, have been prepared as explained above. In addition, N = 3 additional vials had no cap, and one vial had a cap on. Upon etching copper and the series of water cleaning, the graphene pieces were carefully transferred onto plastic vials filled with DI water. The transfer was complete such that the graphene tattoos covered the entire vial opening (see Supplementary Fig. 1011). The vials were left for 1 day at room temperature, then placed on a hotplate at 50 °C for 7 days to enhance evaporation rate, followed by another 5 days at room temperature. The full temperature profile is shown in Supplementary Fig. 10b. The water evaporation rate was determined by daily monitoring the weight of the water inside the vials.

Skin Impedance

The electrode-skin impedance is measured via the Hioki LCR meter IM3536, with the frequency sweep range between 10 Hz and 1 MHz. The measurements were performed without DC bias in the constant voltage mode with 50 mV AC amplitude. Each data point is measured four times and averaged instrumentally. The frequency sweep is collected three times consecutively. The measured values correspond to the electrode-skin impedance since the tissue bio-impedance is much smaller when compared to the typical electrode-skin impedance33. Since the Impedance measurements are performed between a GET pair when reporting the figure-of-merit impedance values, typically at 10 kHz frequency, the measured impedance is divided in half to represent Z per tattoo.

Skin-mock samples

To fabricate the skin-resembling surface for resistance measurements, glass slides were coated with a thick (2–4 mm) layer of Ecoflex 00–30, which was cured at room temperature for 48 h. Ecoflex is used here to reproduce skin-like texture and softness.

Sheet resistance

The sheet resistance was measured by placing a strip of the GET onto the skin-mock samples with four soft conductive gold electrodes. The 4-point TLM was performed to estimate sheet resistance. The values reported are averages from at least 2–3 strips of each kind. The above-mentioned ultrathin conductive adhesive tape is used to establish electrical contacts with the graphene.

Raman spectroscopy

Raman characterization was performed in a Renishaw inVia micro-Raman system. A 532 nm excitation wavelength laser with ~1 mW incident beam power and exposure time of 10 s were used. A 3000 l/mm grating is used for < 5 cm−1 resolution.

GET heaters

The GET heating experiments were performed with the same 4-terminal devices that were used for TLM-based RS measurements. The constant DC voltage of 5 V, 10 V, 15 V, and (optionally 30 V) were applied consecutively to the graphene tattoos, while the surface temperature was constantly monitored by means of the optical thermal camera (FLIR ETS320).

GET temperature sensing

The same TLM structures were used for temperature sensing. The samples were placed on a hotplate, and the temperature was gradually raised from room temperature to ~60 °C, then slowly cooled down. The actual surface temperature was monitored with a thermocouple.