Liquid‐Gated Graphene Field‐Effect Transistors for Biosensing on Lipid Monolayers

This study presents a comprehensive fabrication process for high‐quality graphene field‐effect transistors (GFETs) and their characterization. It is demonstrated that the suggested cost‐efficient method of fabrication utilizing the direct‐laser writing (DLW) system is reliable and ensures lower contact resistance between graphene and metal compared to other works. This improvement results in minimal signal loss and distortion, providing significant benefits to the overall performance of the GFETs. The fabricated devices are functionalized with a lipid monolayer containing tris‐nitrilotriacetic acid fused to the lipids for the binding of histidine‐tagged proteins. Monomeric enhanced green fluorescent protein (GFP) is used as a model protein to explore the current–voltage characteristics response of the liquid‐gated GFETs. The results demonstrate that the devices are well suited for the electrical biosensing of proteins in physiological buffer conditions, paving the way for label‐free detection of protein–protein interactions at membranes.

This study presents a comprehensive fabrication process for high-quality graphene field-effect transistors (GFETs) and their characterization.It is demonstrated that the suggested cost-efficient method of fabrication utilizing the direct-laser writing (DLW) system is reliable and ensures lower contact resistance between graphene and metal compared to other works.This improvement results in minimal signal loss and distortion, providing significant benefits to the overall performance of the GFETs.The fabricated devices are functionalized with a lipid monolayer containing tris-nitrilotriacetic acid fused to the lipids for the binding of histidine-tagged proteins.Monomeric enhanced green fluorescent protein (GFP) is used as a model protein to explore the current-voltage characteristics response of the liquid-gated GFETs.The results demonstrate that the devices are well suited for the electrical biosensing of proteins in physiological buffer conditions, paving the way for label-free detection of protein-protein interactions at membranes.protein fused to an N-terminal hexahistidine tag (H6-mEGFP) was used as a model to explore the current-voltage response of the liquid-gated GFETs.The results demonstrate that our GFET devices are well suitable for electrical biosensing of proteins at physiological buffer conditions.They indicate toward defects in the lipid layer that were not visible in the optical characterization done previously.This shows that electrical measurements lead to a better understanding in the preparation of the membrane environments necessary for PPI detection.This is an important step toward label-free detection of biological processes at membrane interfaces.

Liquid-Gated GFETs Fabrication
The devices were fabricated onto 10 Â 10 mm Si/SiO 2 (200 nm) samples utilizing a commercially available direct laser writer (DLW) PICOMASTER 100 (Raith Laser Systems B.V., The Netherlands).For photolithography, the photoresist thickness was optimized to fall within a range of 250-300 nm by combining AZ1518 positive photoresist with AZ EBR solvent in a 1:2 ratios.The DLW system was operated using a 300 nm laser spot and 50 mJ cm À2 exposure dose.
After metallization, graphene (Easy Transfer Monolayer Graphene G/P-25-25 on Polymer Support from Graphenea, Spain) floating on MilliQ water was transferred to the substrate (Figure 1e).The samples were air-dried for 30 min before annealing at 150 °C for 1 h on a hot plate.After subsequent storage under vacuum for 24 h, the samples were dipped in warm (50 °C) acetone and isopropanol for 1 h each to remove the sacrificial polymer layer.
Graphene patterning was done with another photolithography followed by oxygen plasma exposure at 100 W for 2 min to remove the graphene that is not located between the contacts (Figure 1f ).
Then, the devices with the patterned graphene were washed in acetone and isopropanol (Figure 1g) and passivated for electrical measurements in liquid with a final photolithography step (Figure 1h).The passivation included high-temperature annealing in an argon atmosphere (1 SLPM, Heraeus tube furnace).The optimized process involved three steps with gradual temperature growth and slow cooling afterward, favoring higher mechanical and electrical stability of the passivation (Figure 2a).
An optical microscope image of a few fabricated devices is shown in Figure 2b with a zoom onto one of the TLM structures in the inset.Such TLM structures were used later on for the characterization of the devices.Once the chip was passivated, wire bonding on the chip carrier was performed, followed by encapsulation with polydimethylsiloxane (PDMS) resin and forming a liquid gate reservoir via a glued-in glass ring (Figure 2c).

GFETs Characterization
The electrical characterization was performed utilizing the MPS150 probe system equipped with four DPP210 CascadeMicrotech probes.
The TLM measurements were conducted with two probes applying a bias voltage of 100 mV to obtain the total resistance  R tot for each length.This R tot is dependent on many factors, such as channel resistance (R ch ), contact resistance (R c ), and the resistance associated with metal contacts. [29]As the graphene provided by Graphenea (Spain) is synthesized by chemical vapor deposition (CVD), the sheet resistance is expected to be uniformly distributed throughout the entire TLM structure; consequently, R tot depends linearly on the channel length L ch : [30] with the drain-source voltage (drain-source current) V DS (I DS ), the sheet resistance R sh , the width of the channel W ch , and the contact resistance R c .

GFETs Characterization
All functionalization steps and I-V measurements were performed in ambient conditions with HEPES-buffered saline (HBS) at physiological salts concentration (20 mM HEPES, 150 mM NaCl; pH 7.5).The electrical measurements were done utilizing Ag/AgCl reference electrode (Scientific Products GmbH) with wiring according to Figure 3.
In the first step, we form a lipid monolayer assembly on the graphene layer.The formation of a lipid monolayer on graphene due to its hydrophobicity was previously reported by Füllbrunn et al. [22] .In our study, we used a combination of lipids, namely, DOPC and DODA.Specifically, DODA lipids were covalently conjugated with tris-NTA.DOPC/tris-NTA-DODA mixture was dissolved in HBS at a ratio of 95/5 mol%, respectively, to a final concentration of 250 μM.Prepared lipids were introduced to the sample surface (500 μL) and incubated for 40 min (Figure 3a).
Next, the sample was washed 5 times with 500 μL HBS.Subsequently, 10 mM nickel(II) chloride (NiCl 2 , Sigma-Aldrich) in HBS was added.The nitrogen atoms of tris-NTA are able to coordinate with Ni(II) ions in a chelation complex, forming a stable interaction.This specificity allows for the selective binding of His-tagged proteins on the next step.The sample was incubated for 5 min, followed by a fivefold HBS wash to eliminate any residual NiCl 2 (Figure 3b).
Purified mEGFP fused to H6-mEGFP is immobilized by the tris-NTA-Ni 2þ complex as described by Lara Jorde et al. [20] H6-mEGFP was expressed in Escherichia coli and purified by standard protocols using metal ion chromatography and size exclusion chromatography.The sample was incubated for 10 min with 500 nM H6-mEGFP, followed by a final fivefold HBS wash to remove unbound proteins (Figure 3c).
This functionalization route follows prior works, where the successful mEGFP immobilization was confirmed and characterized by total internal reflection fluorescence spectroscopy and reflectance interference. [22]

Results and Discussion
The fabrication of GFETs was performed following the technological procedures outlined in Section 2.1.
[32][33] As a result, they are considered important figures of merit for determining the quality of GFETs.Besides, it is worth noting that the contact resistance at the graphene/metal interface depends on the width of the contact (W ), not its area (W • L). [30] Thus, apart from the contact resistance, also the contact resistivity ρ c is an important characterization parameter for the fabricated structures.
We measured 90 TLM structures, similar to the one illustrated in Figure 2b (inset), distributed over five chips to perform a comprehensive characterization of the GFETs.The results are presented in Figure 4.
Following Equation (1), a linear regression R tot versus L is used to determine the sheet resistance from the slope and the contact resistance from the y-axis intercept of the linear fit.This analysis  Furthermore, we extracted the contact resistivity, ρ c , to be approximately 384.4 Ω μm.This value exceeds the performance of most devices with similar conformation (see Table 1) and proves the high quality of the fabricated structures.
After the characterization of TLM structures, we utilized GFETs for the immobilization of mEGFP on the lipid monolayer.The detection was performed by monitoring the change in the voltage of the current neutrality point (V CNP ).
The V CNP is an indicator for the doping of the graphene layer caused by electrostatic gating.In the intrinsic graphene, it is expected to be found close to 0 V.However, the V CNP of the initial I-V transfer characteristic of our GFET (Figure 5a, black) is shifted to positive gate voltages and located at V GS = 0.245 V. We attribute this shift to light n-doping caused by adsorbates or adsorbates attached to silanol groups captured between graphene and silicon dioxide during the wet transfer. [34]fter introducing a mixture of DOPC and tris-NTA-DODA lipids, our hypothesis suggests that the formed lipid monolayer acts as a passivation layer, reducing unintentional doping caused by impurities from the fabrication stage.Initially, we observed a shift in the V CNP by ΔV GS = À30 mV.Upon performing the HBS wash to remove any unattached DOPC/tris-NTA-DODA, we find that the lipid monolayer causes a final V CNP shift by ΔV GS = À20 mV compared to the pristine device (Figure 5a, red line).This result indicates the lipids monolayer formation.
To create the Ni 2þ -tris-NTA complex, NiCl 2 is added.The nickel cations attachment to a tris-NTA binding sites induce an additional shift in the V CNP by ΔV GS = À70 mV (Figure 5b, blue).Then another buffer wash follows.The final shift caused by attached nickel ions is ΔV GS = À10 mV compared to the V CNP of the buffer-washed device with the formed lipids monolayer (Figure 5a, blue and Figure 5b, dashed blue).The observed significant initial shift in the V CNP upon the addition of NiCl 2 and the subsequent notable shift back after an HBS buffer wash indicate that the formation of the Ni 2þ -tris-NTA complex involves only a small fraction of the introduced nickel cations.Nevertheless, it is worth noting that the introduction of positively charged ions and the subsequent direction of the V CNP change align well with the electrostatic gating model. [35]n the last step, the mEGFP is immobilized.Typically, enhanced GFP carries seven negative net elementary charges. [36]Therefore, its attachment results in a positive shift in V CNP by ΔV GS = þ10 mV compared to the previous step (Figure 5b, yellow).However, after the buffer solution wash, the V CNP voltage shifted in the opposite direction by ΔV GS = -20 mV.
Previous studies showed a similar effect.It is typically attributed to a direct interaction between the bare graphene surface with biomolecules.This involves noncovalent forces such as π-π stacking and van der Waals interactions.[39] These interactions are nonelectrostatic and occur between the aromatic rings of graphene and the aromatic amino acids including not only those in mEGFP and other proteins but  also in the histidine tag.Following this argumentation, we conclude that defects in the lipid layer might allow for unspecific binding of the His-tagged mEGFP to our devices.

Conclusion
In summary, we successfully fabricated GFET biosensors and used them to detect mEGFP immobilization on a lipid monolayer.The characterization of the GFETs showed an average contact resistance of R c = 76.9AE 48 Ω and an average sheet resistance of R sh = 19.3AE 2.6 Ω, with a contact resistivity of ρ c % 384.4 Ω μm, outperforming most devices with similar conformation.We were able to monitor the V CNP changes at the graphene-lipid interface during functionalization.However, the results obtained from the immobilization of mEGFP indicate the presence of defects in the lipid monolayer that might imply conformational changes in mEGFP due to direct absorption onto the graphene surface, highlighting the need for further investigation and optimization of the lipid monolayer formation.Following optimization, the established procedures and insights obtained from this work can be adapted for carbon nanotube-based field-effect transistors.With their remarkable sensitivity and nanodimensional surface area, these devices hold great potential for conducting PPI studies on the singlemolecular level. [6]

Figure 2 .
Figure 2. a) Passivation annealing procedure (here, the red dashed line is a guide for the eyes); b) an optical image featuring four GFETs and one TLM structure with TLM structure close-up onto inset.Here, Ti/Au feedlines are colored in yellow, and passivation openings are colored in light green; c) a final view of the chip after wire-bonding and encapsulation.

Figure 4 .
Figure 4. Total resistance depending on the length of the TLM structure with a linear approximation for R c and R sh extraction (blue solid line).Here, red boxes show the median value for every length with the upper and lower quartiles, whiskers indicate one standard deviation.

Table 1 .
Comparison of contact resistivity of GFETs fabricated in this work to the GFETs contact resistivity reported in the other research papers.Stack layers [nm] Deposition technique Graphene ρ c [Ω μm] References