Graphene field effect transistors (GFETs) are promising devices for biochemical sensing. Integrating GFETs onto complex non-planar surfaces could uncap their potential in emerging areas of wearable electronics, such as smart contact lenses and microneedle sensing. However, the fabrication of GFETs on non-planar surfaces is challenging using conventional lithography approaches. Here, we develop a combined spray-coating and photolithography setup for the scalable fabrication of GFETs on non-planar surfaces and demonstrate their application as integrated GFETs on microneedles. We optimize the setup to pattern 67 μm long GFET channels across the microneedle tips. Graphene is deposited between photo-patterned electrodes by spray-coating a liquid-phase exfoliated graphene ink while monitoring the channel resistance to achieve the required conductivity. The formation of the GFET channels is confirmed by SEM and EDX mapping, and the GFETs are shown to modulate in solution. This demonstrates an approach for manufacturing graphene electronic devices on complex non-planar surfaces like microneedles and opens possibilities for wearable GFET microneedle sensors for real-time monitoring of biomarkers.
Graphene field effect transistors (GFETs) are an emerging class of devices that have found applications in vast sectors of printed and wearable electronics1 and optoelectronics2 due to their ambipolar behavior,3 high charge carrier mobility,4 and sensitivity.5 GFET sensors leverage graphene's high sensitivity to detect and quantify biomolecules5 including proteins and nucleic acids,6 gases like NO2,7 small organic molecules,8 and pH.9 GFET sensors can exhibit very low detection limits, for instance, detecting SARS-CoV-2 spike protein in the 10−18 g mL−1 range10 and a fast response time, on the order of seconds.11,12
Graphene's sensitivity to external electric fields and doping alters the conductivity of a GFET channel and enables sensing for analytes.13 A GFET sensor requires at least three electrodes: source and drain electrodes—which are the contacts to the graphene channel—and the gate electrode modulating the electric field applied to the graphene channel (Fig. 1). The gate must be separated from the graphene channel by a thin dielectric film (few nanometers to hundreds of nanometers), forming a dielectrically gated GFET, or by a solution, creating a solution-gated GFET (SG-GFET).14
The type of substrate onto which the GFET is manufactured is mainly driven by the end application. For instance, for scalable printed electronics, GFETs have been ink-jet printed onto Si/SiO2 substrates.15,16 Furthermore, GFETs have been fabricated on flexible and stretchable polymers17 and textiles,18 for wearable sensing applications. Similarly, for scalable and portable testing, GFETs have been manufactured on printed circuit boards.12 All of these substrates are planar, with various roughness values, but without major 3D features.
Integrating GFET sensors onto complex non-planar surfaces such as contact lenses or minimally invasive microneedles could enable continuous biochemical measurement in tears or interstitial fluid (ISF) for personalized, wearable health monitoring.19,20 The ISF contains diverse biomarkers and metabolic products that often correlate with blood serum levels.21,22 This enables applications such as glucose monitoring for diabetes management,23 antibiotic level monitoring,24 and reproductive health monitoring.25,26 Moreover, compared to hypodermic needles, microneedles are minimally invasive and less painful,27 only penetrating the epidermis and allowing the multiplexing of several sensors.28 Similarly, tears contain a wide range of analytes, including glucose, multiple sclerosis biomarkers, or eye condition biomarkers,29 and tear sensing can provide real-time and noninvasive health information.
Among various non-planar surfaces, microneedles represent a typical complex surface requiring advanced deposition and patterning techniques for integrating GFET sensors. However, the manufacturing of graphene channels for GFET sensors on 3D microneedle surfaces presents substantial fabrication challenges incompatible with standard lithography techniques. Here, we demonstrate a sequential spray-coating and lithography process (Fig. 2) to pattern graphene GFET channels on non-planar surfaces and demonstrate their potential to fabricate scalable microneedle GFET sensor arrays operating in SG-GFET configuration. A customized combined spray-coating and lithography system based on a low-cost 3D printer has been developed for this purpose (Fig. 3).
Figure 2 shows the fabrication steps for SG-GFET microneedle sensors. The process starts with microneedle substrates containing four arrays of 16 microneedles each, consisting of square-based ( mm) pyramids with a height of 1 mm. The substrate is manufactured through injection molding, via a process described in Ref. 30.
First, the microneedles were metalized by sputter-coating a gold film (100 nm, 5 nm Ti adhesion layer), achieving a sheet resistance of 0.54 sq. If desired, the microneedles can also be metalized by spray-coating metal nanoparticle inks, as shown in supplementary material S2. The deposited gold film was patterned through photolithography to define source and drain electrodes. To determine the optimal photoresist thickness, films with photoresist thickness ranging between 2 and 7 μm were spray-coated (Fig. S4 in the supplementary material). The roughness was found to be strongly dependent on the photoresist thickness, where the average roughness was at 2 μm thickness (Fig. S4). However, once a thicker layer was formed (4–7 μm), the roughness dropped dramatically (Ra: 0.01–0.02 μm, Fig. S4), likely due to the drops coalescing. The maximum achievable resolution decreases with increasing photoresist thickness,31 and therefore 5 μm photoresist thickness allows to obtain a smooth photoresist film without compromising the resolution.
After photoresist spray-coating, the resist film was soft-baked (10 min, 110 °C). The patterning was performed with a laser, and the details of the patterning process are discussed in subsequent paragraphs. Following photopatterning, the resist was developed in AZ400K developer (Microchemicals, 1:3 dilution) for 60 s. Once parts of the gold film were selectively exposed by patterning and development, they were etched away by a mixture of iodine and potassium iodide. This etchant selectively dissolves gold at room temperature but does not attack the polycarbonate substrate.32 Parts of the electrodes that are not intended to be in contact with the measured samples, such as the traces on the microneedle bases, were insulated with a UV-curable varnish applied with a manual brush applicator.
To deposit graphene onto the electrodes, a graphene ink was prepared by liquid-phase exfoliation (LPE) following reference.12 Graphite powder (Ceylon Graphite K1) was dispersed in 2-propanol and stabilized by 0.4 mg/mL of polyvinylpyrrolidone (PVP). Following 9 h of bath sonication and removal of the unexfoliated graphite by centrifugation at 13 000 g (mean), a stable dispersion with a concentration 0.82 mg mL−1 was obtained. The UV–vis spectrum of this ink is shown in Fig. 4(a). Full characterization of the graphene ink (rheometry, Raman spectroscopy, atomic force microscopy, and stability data) is reported in Ref. 12.
The graphene ink was sprayed to form the GFET active channel. The spray deposition aimed at obtaining transistors with channel resitance (Rch) similar to previously published working SG-GFET devices with .12 The resistance was continuously monitored during the spray-coating process by connecting the drain and source electrodes to a potentiostat, applying a voltage between the electrodes, , and measuring the current between the drain and source electrodes, . Figure 4(b) shows the channel resistance as a function of time. Initially, the channel is open (non-conductive), but after 16 s of spray-coating, percolation gives rise to an onset in the conductivity of the channel, and the resistance rapidly drops by 5 orders of magnitude to 165 . Finally, intense pulsed light (IPL) annealing was used33 to remove the PVP stabilizer. This annealing method allows for selective heating of the graphene film (degrading PVP) while leaving the reflective electroplated gold film and the transparent polycarbonate substrate unaffected. Furthermore, based on our previous Raman measurements,12 the annealing method also does not affect the graphene flakes themselves. Figure 4(b) shows that the IPL pulse reduced Rch from 165 to 85.7 . However, we note that only the first IPL pulse has such a strong effect, and the second pulse only decreases it by a further 3% to 83.3 . The thickness (t) of the deposited graphene channel in this step was approximately 25 nm thick (Fig. S5 of the supplementary material), based on profilometric measurements of an equivalent graphene film deposited on a Si/SiO2 substrate.
A photolithographic process on a non-planar object, such as the microneedles, is highly challenging. The formation of electrodes on non-planar surfaces has previously been addressed by using holographic photolithography,34,35 using flexible photomasks36 or by forming the electrodes on planar substrates and then transferring them onto the non-planar object.37 However, the holographic approach requires complex calculations of the phase mask and its manufacture.34,35 Similarly, flexible photomasks or electrode transfer are unsuitable for the microneedles due to their sharpness and geometry. Furthermore, the distance between the focusing lens and the surface of the microneedle onto which the laser should be focused (focal distance) varies with the microneedle height (1 mm in our case), and unless special measures are taken, such as continuously adjusting the focus, the patterning light would be blurred on parts of the substrate. Additionally, any optics suitable for this work must have a working distance greater than the microneedle height (>1 mm) to prevent physical collisions.
Due to the specific requirements, a customized laser lithography patterning setup was developed in this work (Fig. 3). The exposure is performed with a laser module (405 nm, 1 mW, Picotronic 70115505) that includes a built-in aspheric lens focusing the laser beam. The setup is placed in a 3D-printed body and mounted on the spray-coater setup, enabling movement along three axes. The patternable area is limited by the 3D printer to cm (fitting up to 50 substrates), but a larger deposition area, useful when scaling up this process, could be achieved by using a larger 3D printer. The challenges around focal distance variation and working distance requirements were addressed by using a long focal distance of 150 mm and a small numerical aperture (NA) of 0.014 (calculated as where n is the index of refraction of air, 1, and the laser beam divergence angle, ). This results in a large depth of focus, defined as the range of substrate–lens distance over which linewidths are maintained within specifications.38 As shown in Fig. S6 of the supplementary material, the linewidth change as a function of the focal distance is small, going from 50 to 72 m across the whole microneedle height range.
With the focal distance and laser power fixed, we optimized the effective exposure dose by varying the laser movement speed between 10 and 1624 mm min−1 (see optical micrograph in Fig. S7 of the supplementary material). At high laser speed (1624 mm min−1, ), the lines do not get fully patterned, whereas at low speed (10 mm min−1, ), they get over-exposed and substantially broadened, as can be seen in Fig. S7. The smallest lines obtained by our setup were as narrow as 41 μm (1624 mm min−1 writing speed, Fig. S7). The linewidth is negatively impacted by the low numerical aperture, which causes substantial diffraction-related broadening;38 at the same time, this low NA permits long working distances and large depth of focus required on non-planar surfaces. When the 1624 mm min−1 speed was used for microneedle patterning, the frequency of the defects increased (especially conductive bridges between electrodes, see Fig. S8 of the supplementary material), caused primarily by the roughness of the photoresist film and the microneedle substrate. Any variation in the photoresist thickness changes the optimal exposure dose (which is dependent on the thickness).38 Therefore, a much slower speed (100 mm min−1, ) was used, resulting in broader lines and also substrate formation without bridging defects.
Precise alignment of the microneedle substrate during laser patterning is crucial. The substrates were aligned by directing the integrated laser beam onto a dichroic beam splitter (Thorlabs DMLP490T, Fig. 3) and then onto the substrate. The laser point on the substrate can then be observed by a camera (Raspberry Pi V2.1) through the same beam splitter, and the exact position of the microneedles relative to the laser spot can be determined. We used the microneedle tips as the substrate position references. After the relative positions of the four corner microneedle tips were determined, a homographic transformation matrix [Eq. (S1) of the supplementary material] was computed to correct for the substrate rotation and offset.
An SEM micrograph of the graphene channels patterned on microneedles is shown in Fig. 4(c), demonstrating a 63 μm graphene channel across the microneedles. Previous works have reported that this channel length is sufficient to achieve functional GFET sensors.12 The channel runs across 16 microneedles for a total width of 43.8 mm. At across the channel, the channel conductivity was , where R is the resistance of the channel and W its width, L its length, and t the thickness of the graphene film. The conductivity of the metal electrodes is substantially higher than that of graphene, at . This is consistent with previously reported GFET channel resistances.12 The overall yield for the optimized parameters, defined as the percentage of substrates from a batch yielding GFETs showing modulation, was . While the patterns in Fig. 4(c) are primarily straight lines, the laser patterning setup would be also able to produce L-shaped or curved patterns, as this is supported by the 3D printer's firmware.
Energy-dispersive x-ray spectroscopy (EDX) was used to conduct elemental mapping across a microneedle shown Fig. 4(d). Figure 4(e) shows the EDX map of gold in the area of a microneedle, demonstrating the presence of gold electrodes (yellow color) surrounding the graphene channel, identified by the presence of carbon (green color). Due to the 3D shape of the microneedles, the EDX measurement is affected by shadows—as the EDX detector is offset to the side, the x rays from the bottom right part of the images cannot effectively reach it.
To verify the GFET function, the substrate was immersed in a phosphate buffer ( ), a drain voltage was applied ( ), and the gate voltage swept between −0.4 and +0.4 V. Figure 4(f) shows the characteristic curve of an SG-GFET (blue curve), where the drain current ( ) is clearly modulated by the gate. The gate current ( ) is almost 3 orders of magnitude smaller than , at 3 μA (mean) vs 2476 μA (mean). Outside of the range shown (−0.4 to +0.4 V), the gate current increased substantially due to electrolytic effects. The GFET modulation observation is consistent with the results for previously reported sprayed solution-gated GFETs on planar substrates.12 At the same time, unlike planar devices, the GFETs on microneedles can directly sample the interstitial fluid, unlocking wearable GFET biosensing.
In this work, we have fabricated and demonstrated graphene field effect transistors on a complex non-planar surface, microneedle tips, which behave differently from graphene-based coatings on AFM tips.39 Graphene channels were manufactured with a typical length of and typical channel resistance of . The process that was developed creates a sensing platform, bringing GFET sensing to wearable microneedles. This technology can potentially interrogate a single GFET microneedle, paving the way toward multi-analyte microneedle biosensor arrays. This process can also be adapted for non-planar substrates other than microneedles, such as manufacturing GFET sensors on contact lenses (for sensing in tears) or fibers (for wearable sensing).
SUPPLEMENTARY MATERIAL
See the supplementary material for further information on the deposition methods and the patterning setup developed here, for data on focus and patterning speed optimization as well as for the results obtained from using spray-coated silver instead of gold as the metallization layer.
The authors acknowledge funding from the European Union (NextGeneration EU), through the MUR-PNRR project SAMOTHRACE (ECS00000022). F.T., M.H., and B.F.S. acknowledge funding from EPSRC via Grant Nos. EP/P02534X/2, EP/T005106/1, EP/R511547/1, and EP/X026876/1. M.H. acknowledges funding from the Imperial College President's Ph.D. Scholarship (EPSRC EP/T51780X/1) and support from the EPSRC Centre for Doctoral Training in Advanced Characterisation of Materials (EPSRC EP/S023259/1). The authors would like to thank Professor Nikolaj Gadegaard for the microneedle substrates, Dr. Nicola Gasparini for assistance with microneedle metallization, and Dr. Mahmoud Ardakani for his help with SEM measurements.
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
Martin Holicky: Conceptualization (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). B. Fenech-Salerno: Investigation (equal); Writing – review & editing (equal). A. E. G. Cass: Funding acquisition (equal); Supervision (equal); Writing – review & editing (equal). F. Torrisi: Funding acquisition (equal); Methodology (equal); Project administration (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).
DATA AVAILABILITY
The data supporting this study's findings are available from the corresponding author upon reasonable request.