Transistors platform for rapid and parallel detection of multiple pathogens by nanoscale-localized multiplexed biological activation

The rise in antibiotic-resistant pathogens, highly infectious viruses, and chronic diseases has prompted the search for rapid and versatile medical tests that can be performed by the patient. An electronic biosensing platform based on eld-effect transistors (FETs) is particularly attractive due to sensitivity, fast turn-around, and compatibility with semiconductor manufacturing. However, the lack of methods for pathogen-specic functionalization of individual FETs prevents parallel detection of multiple pathogens. Indeed, so far functionalization of FET based biosensors is achieved by drop casting without any spatial selectivity. Here, we propose a paradigm shift in FET’s biofunctionalization. Speci�cally, we use thermal scanning probe lithography (tSPL) with a thermochemically sensitive polymer that can be spin-coated on any FET material. We demonstrate that this scalable, CMOS compatible methodology can be used to functionalize individual FETs with different bioreceptors on the same chip, at sub-20 nm resolution, paving the way for massively parallel FET detection of multiple pathogens. Antibody-and aptamer-modied FET sensors are then realized, achieving an ultra-sensitive detection of 5 aM of SARS-CoV-2 spike proteins and 10 human SARS-CoV-2 infectious live virus particles/ml, and selectivity against human in�uenza A (H1N1) live virus.


Introduction
A variety of global health threats, such as highly infectious viruses, chronic diseases, and increasing prevalence of antibiotics-resistant pathogens, require rapid and versatile tests which can be performed by the patient.Field-effect transistors (FETs) con gured as biosensors serve as a suitable platform, compatible with modern semiconductor manufacturing, for label-free rapid sensing by converting interactions between target analytes and surfaces into real-time electrical signals [1][2][3][4].
However, the application of FET-based sensors for highly parallel detection of multiple distinct pathogens or biomarkers remains elusive.At present, functionalization of FET based biosensors is achieved by drop casting without any spatial selectivity.Therefore a critical need is a CMOS-compatible and scalable surface chemistry fabrication strategy that can allow modi cation of each FET on a chip on-demand with distinct pathogen/biomarker-speci c bioreceptors, such as antibodies [5,6] and aptamers [7,8].
Overcoming this challenge will enable portable and wearable rapid diagnostic tests with unprecedented capabilities.
Furthermore, the detection limits of FET-based sensors have been pushed to ultralow concentrations (i.e., femto-molar or less) by adapting nanoscale materials such as nanowires [9,10], In 2 O 3 [7], and graphene [11][12][13].The progress in this direction promises new applications for FET-based sensor technologies in drug discovery [14,15] and clinical diagnostics [16][17][18][19].Nonetheless, existing surface functionalization strategies have several limitations.A primary limitation is the lack of a spatially selective methodology with nanoscale precision for realizing bioreceptor-modi ed FET sensors.The proposed approach so far to achieve multiplexing is to create physical macroscopic barriers between different FETs [7,12,20], using for example polydimethylsiloxane (PDMS) barriers and chambers measuring several millimeters in size (see Supplementary Table 1 for a review).However, modern semiconductor manufacturing can integrate billions of sub-100 nm size FETs into functional microchip products [21], therefore, a paradigm shift in biofunctionalization is required for con guring each individual nanoscale FET into a distinct pathogen/biomarker-speci c biosensor.
Another limitation pertains to the development of surface chemistry processes that are FET-material independent.For instance, attaching a given antibody or aptamer to a graphene lm will involve different processing steps compared to an oxide lm.Functionalizing graphene requires creating either ester or vinyl sulfone groups on the surface [22], which react with the primary amine groups of the receptorantibodies.On the other hand, a different approach is needed for functionalizing In 2 O 3 FETs sensors with aptamers, involving the covalent modi cation of In 2 O 3 with dopamine or serotonin aptamers using silane chemistry [7].These functionalization strategies depend on the FET material and lack the ability to allow local chemical patterning of each FET with a different bioreceptor, as the entire biosensor chip is functionalized identically.
To address the above-mentioned limitations, here, we introduce a scalable, CMOS compatible functionalization strategy, applicable to any FET material, permitting local chemical modi cation of individual nanoscale FETs on the same chip with different bioreceptors from antibodies to aptamers, at sub-20 nm resolution and 200 nm pitch, a distance comparable to the pitch of modern FETs array in CMOS chip.Our strategy involves the use of thermal scanning probe lithography (tSPL) with a thermochemically sensitive polymer.We demonstrate the capability of this strategy through multiplexed functionalization with different bioreceptors at nanoscale precision.In a typical FET biosensor, the detection occurs by recording the change in charge near the channel region, due to the trapping of a target analyte by a speci c bioreceptor.We show that changes in charge at the surface of the polymer covering the FET channel equally give rise to detectable electronic signals by capacitively coupling through the polymer lm.Furthermore, the ability to pattern individual FET allows for in-situ differential sensing, a key ingredient to ensure signal delity.The feasibility of this platform is established by testing antibody-and aptamer-modi ed FET sensors fabricated using this polymer-based FET biofunctionalization approach.The sensors achieve an ultra-sensitive detection of 5 aM of SARS-CoV-2 spike proteins and 10 human SARS-CoV-2 infectious virus particles/ml, which caused the COVID-19 pandemic, and selectivity against human in uenza A (H1N1) live virus.

Nanoscale thermal biofunctionalization: the NanoBioFET platform
To implement a scalable strategy capable of locally functionalizing individual FETs at sub-100 nm resolution with desired bioreceptors (from antibodies to aptamers), and applicable to any channel material (e.g., graphene, oxides or silicon), we use thermal scanning probe lithography (tSPL) [23,24].In our approach, tSPL uses a hot nanotip to expose amine groups with nanoscale resolution on a thermally sensitive biocompatible polymer resist [25][26][27][28][29], which is spin coated on a fully-fabricated array of FETs (see Fig. 1b).This spatially selective activation process enables subsequent modi cation of individual or a group of FETs with a desired bioreceptor, resulting in an array of FET-based biosensors con gured for simultaneously detecting various target analytes (see Fig. 1a).We name this platform NanoBioFET.
The implementation of the NanoBioFET platform begins by fabricating a FET array.As shown in Fig. 1b, we adopt monolayer graphene for realizing the FET-based sensors (see Methods), due to its potential for ultrahigh sensitivity [30,31].We then cover the fully fabricated graphene FETs (gFETs) by spin coating two thermally sensitive polymer resist lms (see Methods).The stack comprises a rst lm of about 70nm-thick polymethacrylate-carbamate-cinnamate copolymer (PMCC) [26][27][28], which provides amine groups on demand in designated FET regions upon local heating by tSPL.PMCC can be locally patterned through local heat-induced deprotection of amine groups from tetrahydropyranyl carbamates in the carbamate block of PMCC [25][26][27][28][29] (see Methods).The second lm is poly(phthalaldehyde) (PPA) [23,32] which serves as a top layer resist (10-20 nm thick) to reduce non-speci c binding outside the FET channel region (see Supplementary Note 1).
After forming the bilayer polymer resist stack, we proceed with the localized functionalization of individual FETs with desired bioreceptors.Employing tSPL, we modify the bilayer polymer resist stack atop the channel region of each target gFET with amine groups.This process utilizes a hot nano-tip in a commercial tSPL system (see Methods), which removes PPA completely above the channel region, while simultaneously applying heat to the PMCC polymer surface above the amine deprotection temperature (~ 120 C), exposing amine groups in the desired area (Fig. 1b).For differential sensing, we pattern only the FET used as a sensor, while leaving un-patterned the FET used as a control.Ensuring signal delity is a key challenge for commercialization of FET-based biosensors.Hence, providing a platform that allows for in-situ differential sensing is a key milestone.Using standard conjugation strategies, the amine groups in the channel region can subsequentially be functionalized with ad-hoc bioreceptors such as antibodies or aptamers (see Methods for details).At this point, the sensor/control FETs are ready for biodetection of a speci c target, such as SARS-CoV-2 virus.As discussed later, when the FET sensor is exposed to a speci c target, the target is immobilized by the bioreceptor near the polymer surface, giving rise to a change in the electronic signal of FET-based sensors.
In Fig. 1c, we demonstrate the resolution of the NanoBioFET fabrication method.Speci cally, Fig. 1c shows an in-situ tSPL topographical image (see Methods) of an amine pattern produced by tSPL in a PPA/PMCC bilayer polymer resist stack deposited on a SiO 2 /Si wafer (without graphene).The pattern consists of a 10 x 10 matrix of 15 nm-diameter amine circles with 40 nm pitch.This image shows a resolution compatible with the 5 nm-node silicon FET technology and below.Importantly, this image is taken in-situ during the tSPL patterning process, since thermal probes can also allow local nanoscale imaging (Methods).This feature enables localization of target regions on a chip with nanoscale precision without requiring sophisticated and costly pattern alignment procedures.
To demonstrate the ability of the NanoBioFET fabrication method to pattern relevant biomarkers with high reproducibility and precision, Figs.1d and 1e present two uorescence optical microscopy images of one-hundred squares (with 5 µm and 500 nm sides, respectively) of biotinylated anti-SARS-CoV-2 aptamers, which have been uorescently tagged with a red dye (see Methods for details on the aptamers used here).These aptamer patterns have been produced rst by exposing amine groups by tSPL on the surface of the PPA/PMCC polymer resist, and then by conjugating the amine groups to NHS-biotin, followed by streptavidin, and biotinylated aptamers (the details of the chemical functionalization steps are reported in the Methods).The PPA/PMCC double polymer stack is here spin-coated on a silicon oxide/silicon wafer.
In Figs.1f and g, we show how the NanoBioFET fabrication process is implemented on a gFET array chip.
For this purpose, we fabricate an array of four gFETs with a channel length of 1 µm, where each sensor FET is adjacent to a control FET.The PPA and PMCC resists are then spin-coated on the gFETs, and tSPL is used to remove PPA and pattern amine groups in the channel region of the two gFET sensors, leaving the FET controls unpatterned.In Figs.1f and g we show, respectively, an in-situ tSPL topographical image, and an ex-situ friction atomic force microscopy (AFM) image of the resulting gFET sensors and controls after tSPL patterning.The graphene region underneath the polymer is indicated with a dashed red rectangle.In Fig. 1f, the tSPL topography image shows that the patterns above the channel region of the gFET sensors are at a depth of approximately 10 nm, corresponding to the thickness of the PPA lm, which is sublimated during tSPL patterning.It is important to note that since tSPL allows in-situ imaging of the FET underneath the dual polymer stack, we are able to locate the channel region of each FET and pattern where required without the need of markers.The AFM friction image in Fig. 1g is particularly revealing since it shows clear friction contrast in the sensor area due to the change in hydrophilicity upon the thermal deprotection of the amine moiety.This agrees with a more hydrophilic surface where the amine groups have been exposed, and the consequent presence of larger friction forces at the nanoscale due to larger capillary forces [24,33].

The NanoBioFET platform for parallel biochemical sensing
To fully exploit the potential of having a massive number of FET biosensors in a microchip, it is necessary to functionalize each FET with distinct bioreceptors so as to allow detection of different types of target biomolecules, such as different viruses.In Fig. 2, we demonstrate the capability of the NanoBioFET fabrication process to pattern each FET with an independent bioreceptor to permit parallel sensing on the same chip.Figure 2a shows a schematic representing the steps required for patterning the different FETs with distinct bioreceptors.Initially, we fabricate an array of FETs and we spin-coat a PPA/PMCC polymer stack on them.Then a rst round of NanoBioFET fabrication, as depicted in Fig. 1b, is performed on FET-1 to attach bioreceptor-1 (red square), followed by a second round to attach bioreceptor-2 (green square) to FET-2, and so on until all FETs are functionalized up to the desired nnumber of bioreceptors.Each round of NanoBioFET fabrication includes: rst, in-situ tSPL topographical imaging of the surface to locate where the pattern needs to be made (e.g., the desired FET); second, tSPL local patterning of an individual FET channel; third, a series of biochemical functionalization steps, and selective attachment of the desired bioreceptor to the required FET as described in the Methods part.
In Fig. 2b we present a uorescence optical microscopy image of biochemical patterns generated by four sequential rounds of NanoBioFET fabrication as depicted in Fig. 2a.In particular, Fig. 2b shows the uorescence of four types of NHS-esters terminated dyes (red, yellow, green and sky-blue) attached directly to the amine moieties exposed during the tSPL process on the surface of the PPA/PMCC polymer resist deposited on a silicon oxide/Si wafer.Figure 2c shows the implementation of the multiplexed chemical patterning in an array of gFETs.Speci cally, four representative gFETs channel regions have been functionalized with four different types of NHS-esters terminated dyes (red, yellow, green, and skyblue) attached directly to the amine moieties exposed by tSPL on the surface of a PPA/PMCC polymer resist spin-coated on a gFET chip.
Figures 2d and e demonstrate the capability of the NanoBioFET fabrication method to produce independent patterns of different types of bioreceptors sensitive to different types of target molecules, e.g., different viruses.In particular, here we show a uorescence image of patterns of two types of uorescently labelled aptamers, in uenza A anti-hemagglutinin (HA) aptamer tagged with a green uorophore, and anti-SARS-CoV-2 aptamer (tagged with a red uorophore).The patterns are created by exposing amine groups by tSPL, and subsequently by attaching NHS-ester biotin to the amine patterns.
Biotin patterns are then exposed to streptavidin, and nally to the biotinylated HA aptamer (green).A second identical round of NanoBioFET fabrication is performed to attach the biotinylated CoV-2 aptamer (red) to the desired area on the surface.The details are reported in the Methods.
Figures 2e and f demonstrate the high spatial resolution with which patterns/bits of different bioreceptors can be fabricated on the channel of parallel FETs with this approach.In particular, following the same process as in Figs.2d and e, we produce patterns of two types of aptamers (HA and CoV-2 aptamers).The uorescence image in Fig. 2e shows green-aptamer and red-aptamer patterns with a "bit" dimension of 500 nm.Because of the limited resolution offered by optical microscopy, we also fabricate two-aptamer circle/dash patterns with a 20 nm width and a minimum 200 nm pitch, i.e., minimum distance between circle (HA aptamer) and dash (CoV-2 aptamer) pattern centers.We then image them insitu by tSPL imaging (Fig. 2f).We perform three measurements on the PPA/PMCC surface.First, we image pattern type-circle after a rst round of tSPL.Second, we functionalize pattern type-circle with NHS-biotin/streptavidin/HA aptamer and image the pattern region after a second round of tSPL to produce pattern type-dash.Third, we image the polymer surface after functionalization of pattern typedash with NHS-biotin/streptavidin/CoV-2 aptamer.The cross-sections show the high registry and robustness of the fabrication process and the change in depth of the patterns after functionalization due to the lling of each pattern with the NHS-biotin/streptavidin/aptamer molecules (approximately 10-15 nm).The two types of patterns have been produced with different shapes to add clarity to the image and show that tSPL can also produce patterns at different depths and shape.See also supplementary Figure S8.

Electronic sensing using the NanoBioFET platform
Having established the versatility and nanoscale spatial precision of the NanoBioFET fabrication strategy, we next examine its feasibility in electronic detection of target analytes.As a proof-of-concept, we implement electronic sensors based on gFETs.Our sensing experiments are tailored for the detection of the SARS-CoV-2 virus, chosen as an example target species.This choice is motivated by the commercial availability of bioreceptors for this virus, including different antibodies and aptamers.
gFETs are promising biosensor candidates due to their potential for high sensitivity and ease of fabrication [13,22,34].Like other FET-based biosensors, a bioreceptor-modi ed gFET translates the pathogen-bioreceptor interaction near the surface to a detectable electronic signal.Figure 3a shows the schematic illustration of an antibody-modi ed gFET.The sensor is solution-gated, where a xed bias gate voltage (V gs ) is applied to the solution (300 mV in our experiments) using a Ag/AgCl reference electrode.
A small bias voltage (V ds =50 mV) is simultaneously applied between the source and drain electrodes, generating a current ow (I ds ) in the gFET, which is monitored in real time.The amplitude of I ds changes upon conjugation of target analytes (e.g., spike protein or virus) with bioreceptors, due to a change in electronic charge on the surface of the biosensor.Furthermore, the recorded I ds signal can be used to quantify the concentration of target analytes through creation of a sensitivity calibration curve, as we explain later.
However, a key experimental challenge in FET-based biosensing is the electronic screening of charges with increasing distance from the surface, characterized by the Debye length λ Debye .A common approach for overcoming this limitation is to adjust the ionic strength of the buffer environment [35,36], thereby increasing λ Debye .In the experiments below, we adopt a similar strategy, modifying the buffer solution and its ionic strength based on the choice of the bioreceptor (i.e., antibody or aptamer).

Sensing using antibody-modi ed NanoBioFET
All electronic sensing experiments begin by employing the fabrication and biofunctionalization protocols described in Figs. 1 and 2, which involve anchoring NHS-biotin-streptavidin chains onto the thermochemically activated amine groups on the PMCC polymer, above the channel region of the gFETs.
Our initial investigations focus on the use of biotinylated antibodies as a bioreceptor, which attach to the NHS-biotin-streptavidin chains.In this con guration, detecting an electronic signal induced by the antibody-analyte interaction requires λ Debye of at least ~ 10 nm.To accommodate this requirement, we use a 1 mM HEPES buffer solution as the sensing environment (see Methods) [9].Whereas a lower ionic strength enhances λ Debye , the reduction of ion concentration in the sensing environment may negatively affect the e cacy of the antibody-analyte interaction.
We select biotinylated anti-SARS-CoV-2 spike RBD neutralizing antibody (see Methods) for these electronic sensing experiments.The feasibility of its binding a nity to SARS CoV-2 spike protein at 1mM HEPES buffer is con rmed by surface plasmon resonance (SPR) measurements (Fig. 3b).In these SPR experiments, the concentration of spike proteins is limited to a low nanomolar range, which is the typical sensitivity of this type of measurements [37,38].Following this con rmation, we proceed to perform the electronic sensing measurements with the NanoBioFET platform.In all sensing experiments, the bioreceptor-modi ed gFETs are placed in a micro uidic chamber (see Methods).We then perform an initial screening of a gFET sensor quality by measuring its transfer characteristics (I ds vs. V gs ) using a buffer solution gate.In Fig. 3c, we show the typical transfer characteristics of an antibody-modi ed gFET, obtained by sweeping the solution gate voltage, V gs , and recording I ds at a xed V ds of 50 mV.V gs modulates the gFET charge carrier concentration and carrier type, from holes in the p-branch, to electrons in the n-branch, passing through the charge neutrality point (at V CNP ).For simplicity the data are centered around V CNP .When an analyte is captured by the bioreceptor in the channel region, the local change in charge produces a shift of the I ds vs.V gs characteristics and therefore a change in I ds is measured at xed V g .
We next monitor the real-time electronic response of a SARS-CoV-2 antibody-modi ed gFET sensor to different concentrations of spike protein.The experiment involves recording I ds continuously in time at a xed V gs of 300 mV while injecting analytes at different times into the micro uidic chamber.The objective of this experiment is to quantify the sensitivity of this antibody-modi ed gFET and evaluate its limit of detection.In Fig. 3d, we show the corresponding transient response of DI ds /I 0 , where I 0 is the initial I ds (i.e., at t = 0) and DI ds is evaluated by subtracting I 0 from the subsequently recorded I ds .We initially perform repeated injections of the buffer solution and monitor the gFET's DI ds response.The purpose of these injections is to record possible artifacts which might contribute to a false sensor response.The sensor response due to three buffer injections are marked with yellow shading in Fig. 3d.
Although each injection of the buffer solution generates a small detectable response, they are consistent among the three injections and, more critically, do not cause a permanent shift in the DI ds /I 0 baseline.
These observations give con dence that the artifacts of the injection process are negligible and temporary, and thus do not contribute to the steady-state sensor response due to the antibody-spike protein interactions.We then monitor the transient sensor response due to injections of the spike protein (see Methods for discussion on the concentrations).Upon each spike injection, marked with red arrows in Fig. 3d, a small bump is initially observable in the transient curve of the sensor response, associated with the artifact of the injection.However, this response is accompanied with a strong decrease in DI ds /I 0 , that follows an apparent exponential behavior (marked with dashed exponential ts).The subsequent spike injection occurs once DI ds /I 0 establishes a new steady-state baseline.Each of the spike protein injections generates a qualitatively similar response, with increasing magnitude, as expected for increasing concentration of spike protein in the buffer solution.
Figure 3e shows the sensitivity plot of the antibody-modi ed gFET, plotting the sensor response (the amplitude of the exponential decay tting function) against its corresponding spike concentration.Since the sensor response must be null when the target analyte concentration is zero, a linear t to the data must go through the origin and we are hence able to make a meaningful t, obtaining a sensitivity of 0.59 ± 0.04% aM.Analysis of the electronic noise in these measurements reveals a 36 nA rms input-referred noise (0.3% of I 0 ), corresponding to an estimated limit of detection of 1.5 aM (see Supplementary Note 2).
The excellent sensitivity of these gFETs, which could be further optimized in terms of electrical characteristics, reinforces the prospects of NanoBioFETs as biosensors in experiments involving ultralow concentrations of analytes.

Sensing using aptamer-modi ed NanoBioFET
We next examine the versatility of the NanoBioFET platform in adapting aptamers as bioreceptors for detecting spike proteins.In recent years, aptamers have become increasingly appealing as capturing probes due to their cost-effectiveness and durability [7].More critically, modifying FET-based sensors with aptamers has proven to be an effective strategy in signi cantly relaxing the requirements on λ Debye and, consequently, the ionic strength of the sensing environment [39].Recent demonstrations have revealed the utility of aptamer-modi ed FET sensors in detecting analytes at physiological ionic strength [8], which has a λ Debye of 0.7 nm.The sensing mechanism of an aptamer-modi ed FET-based sensor is attributed to the fact that the analyte-induced conformational changes of the aptamer alter the surface charge within λ Debye , generating a detectable electronic signal.In our experiments, explained below, we demonstrate the successful electronic detection of spike proteins in buffers solutions having a λ Debye of ~ 2 nm.
We implement aptamer-modi ed gFETs following an identical tSPL fabrication and biofunctionalization procedure as in the experiments in Figs. 1 and 2, and we use a biotinylated anti-SARS-CoV-2 aptamer as bioreceptor (see Methods for details).Before the electrical sensing experiments, we employ uorescent microscopy to con rm the aptamer attachment in the channel region of the gFETs designated as biosensors.Indeed, the strong uorescence in the channel region of an aptamer-modi ed gFET (see left panel in Fig. 4b) con rms the effectiveness of our procedure in locally attaching aptamers.In this experiment, we also demonstrate the versatility of this approach in implementing control gFETs on the same chip by leveraging the spatial-selective biofunctionalization capability of tSPL.Control devices are produced simply by skipping the tSPL step in the channel region of a few select gFETs, while subjecting these gFETs to the same subsequent surface chemistry treatments as sensor gFETs.The uorescent microscopy results for the control gFET (see right panel in Fig. 4b) con rm the chemical inactivity of the channel region in the control gFET with minimal non-speci c binding.
In Fig. 4c, we present SPR experiments con rming the binding between the anti-SARS-CoV-2 aptamer and spike protein in 0.1X PBS, corresponding to a λ Debye of approximately 2 nm.The subsequent electronic sensing experiments in a micro uidic chamber demonstrate the ability of aptamer-modi ed gFETs in generating detectable electronic signals in response to spike protein injections in the same buffer concentrations (see Fig. 4d).The simultaneous monitoring of an adjacent control device in this experiment provides con dence about the delity of the electronic signals generated by the gFET biosensor.This key feature of the NanoBioFET platform in enabling differential detection and simultaneous monitoring of non-speci c binding is an important step toward achieving robust analyte detection using FET-based biosensors.
A closer look at the results in Fig. 4d reveals two important observations.The rst observation is the signi cantly weaker signal amplitude of the aptamer-modi ed gFET compared to its antibody-modi ed counterpart in Fig. 3d.Whereas the antibody-modi ed gFET generates a DI ds /I 0 of 2.5% in response to 5 aM of spike concentration, the aptamer-modi ed counterpart produces an order of magnitude weaker signal amplitude when exposed to 500 aM of spike protein.We attribute this characteristic primarily to the signi cantly smaller λ Debye at 0.1X PBS.We expect that ongoing research in the eld, focused on increasing the λ Debye at a given ionic strength [36] and developing alternative chemical conjugation strategies beyond biotin-streptavidin will directly bene t future experiments utilizing aptamer-modi ed gFETs.The second observation pertains to the relationship between the signal amplitude and the spike protein concentration.The data indicate a nearly diminishing response with the subsequent spike protein injections at higher concentrations.This observation suggests that the capturing aptamers on the gFET are reaching full occupancy, wherein the captured surface proteins block the interactions of incoming proteins with the surface probes.

Evaluation of NanoBioFET platform using live human viral particles
Following the success of spike protein detection with bioreceptor-modi ed gFETs, we perform a nal test with human live SARS-CoV-2 viral particles and the SARS-CoV-2 antibody as bioreceptor from earlier, evaluating sensitivity to low viral loads and selectivity of the response to speci c viruses.Additionally, the experiments in Fig. 3 con rm the high sensitivity of antibody-modi ed gFETs in detecting low concentrations of spike proteins.Therefore, we employ the antibody-modi ed NanoBioFET platform in experiments involving live viral particle detection.
The measurement setup and procedure are nearly identical to that for the antibody-spike protein measurements earlier, with a 1 mM HEPES buffer (see Fig. 5a and Methods for details).We monitor the temporal changes in I ds during the course of the sensing experiment.The experimental sensing procedure involves injection of the blank virus medium at the beginning of the experiments, followed by a few alternating injections of SARS-CoV-2 virus and the human H1N1 in uenza virus (see Methods for details on these viruses).Figure 5b shows the transient response of the antibody-modi ed gFET with live virus injections in 1mM HEPES.Injections of virus medium (black circle), SARS-CoV-2 virus (red diamond), and H1N1 virus (green star) are marked in Fig. 5b.The data demonstrate the sensitive and selective detection by the SARS-CoV-2 antibody-modi ed gFET.The SARS-CoV-2 virus (red diamond) and H1N1 virus (green star) in Fig. 5 show that the SARS-CoV-2 virus repeatedly produces an exponential-like response in the sensor, whereas the H1N1 virus (which we use as a negative control) or buffer produce no response.The rst four virus injections are at a concentration of 20 TCID 50 /ml and the last at 200 TCID 50 /ml, estimated to be about 10 infectious virus particles per ml, demonstrating ultra-sensitive detection of the live virus.
The insensitivity of the sensor to H1N1 injections highlights the selectivity of the platform.

Conclusions
In conclusion, the results presented here establish a paradigm shift in functionalization of FETs for biosensing.While so far functionalization of FET based biosensors is achieved by drop casting without any spatial selectivity, here we show a scalable CMOS compatible surface functionalization nanofabrication strategy that allows modi cation of individual FETs on the same chip with distinct pathogen-speci c bioreceptors, with nanoscale resolution.We call this biosensor chip NanoBioFET platform.The NanoBioFET platform is implemented through the combination of thermal scanning probe lithography, thermochemically sensitive polymers which are spin-coated on the FETs chip, and in-situ thermal imaging.We demonstrate that this methodology can be used to chemically functionalize individual FETs with different bioreceptors at sub-20 nm resolution and 200 nm pitch, a distance comparable to the pitch of modern FETs array in CMOS chip.Functionalization of target regions with submicron pitch is a crucial ingredient to achieve massively parallel FET detection of multiple target pathogens.The ability to pattern individual FET also allows for in-situ differential sensing, a key feature to ensure signal delity.
The versatility of NanoBioFET is demonstrated by modifying gFET chips with antibody and aptamer bioreceptors and subsequently employing them in electronic detection of spike protein and human SARS-Cov-2 live virus.The polymer lm covering the FET allows nano-functionalization, and it functions as a coupling capacitor between the surface-anchored bioreceptors and the buried FET sensors, allowing the electronic detection of interactions between target analytes and speci c bioreceptors.The electronic sensing experiments reveal ultrasensitive and selective sensing performance of the NanoBioFET platform.We achieve robust detection for 5 aM of SARS-CoV-2 spike proteins and 10 human SARS-CoV-2 infectious live virus particles/mL, as well as selectivity against human in uenza A (H1N1) live virus.
Lastly, we remark on the nanomanufacturing prospects of the proposed NanoBioFET platform.A key asset of this functionalization strategy is its generalizability to various FET materials from silicon to graphene.Indeed, the polymer stack is spin coated atop the fully fabricated chips, a process scalable to commercial silicon and other types of substrates with up to 300-mm diameter.Therefore, the NanoBioFET platform is compatible with commercial semiconductor manufacturing for producing CMOS chips that can contain many FETs (thousands to millions), where each FET or a group of FETs can be modi ed for detecting a speci c target pathogen.To this end, we note that the tSPL process could be parallelized with the use of hot probes arrays, while the scanner head could work in parallel with picoliter piezotype printing.Finally, tSPL is a exible and sustainable nanofabrication method that does not require vacuum or high energy or alignment marks.The multiplexed parallel sensing enabled by the NanoBioFET platform will offer unprecedent opportunities in the elds of in-home diagnostics, wearables, AI for health data, and e-health.

FET devices fabrication
The FETs are fabricated using a four-stage process.The rst stage involves the preparation of the substrate and graphene material.The substrate is prepared by covering SiO 2 -coated silicon substrates with a 10-nm aluminum oxide (Al 2 O 3 ) lm grown by atomic layer deposition at 270°C, followed by the densi cation of Al 2 O 3 at 500°C for 1 hour in an oxygen ambient.Commercially available single layer graphene lms (ACS Material), grown by chemical vapor deposition (CVD) on copper metal substrates, are then transferred on the Al 2 O 3 -coated substrates using a standard PMMA-based transfer method [40] (details in Supplementary Note 3).In the second stage, graphene islands constituting the channel region of FETs are patterned using a combination of electron beam lithography (EBL) and plasma cleaning (details in Supplementary Note 3).In the third stage, fabrication of gFET array is complete through the formation of source and drain metal electrodes using a combination of EBL patterning, e-beam evaporation of 5 nm Cr/ 20 nm Ti/ 25 nm Au metal stack, and lift-off in acetone, followed by covering the metal electrodes with an EBL-patterned SU8 isolation lm.Stable connections are made from the contact pads to external test circuit using spring loaded pogo pins.The contact pads are designed to lie outside the micro uidic chamber during electrical measurements and allow space for the pogo pins.  1b).More information are reported in [25][26][27][28][29].The purity is determined via 1H nuclear magnetic resonance spectroscopy.

Two-polymer resist stack
The two-polymer resist stack is spin coated on the desired chip in a two-step process.First, a PMCC lm (about 70 nm thick) is spin-coated on the chip.The PMCC lm is obtained by using a 15 mg/ml solution of PMCC dissolved in cyclohexanone [28,41].The chips are spun at 1000 RPM for 5 seconds and 1500 RPM for 15 seconds to disperse the PMCC solution, then 4000 RPM for 30 seconds to evenly apply a thin lm to the substrate.The chips are then treated with 302 nm UV light to crosslink the polymer and increase adhesion to the substrate.Second, a PPA lm (10-20 nm thick) is spin coated on PMCC.PMCC coated chips are spin-coated with a solution of 0.5% PPA in anisole to cover the PMCC surface.The spincoating of PPA is performed at 6000 RPM for 30 seconds.

Thermal Scanning Probe Lithography and Microscopy
Patterning of the PMCC/PPA polymeric stack is performed using a commercial tSPL system (NanoFrazor, Heidelberg Instruments, Germany), which utilizes a heated silicon probe [23].For patterning, the probe on the head of the thermal cantilever is heated up by a resistive micron-heater.The probe works as a separate thermal reading sensor for in-situ topography thermal imaging when the micro-heater is turned off.This is a key feature that allows imaging of the FETs underneath the polymer resist, permitting the local functionalization without the need of markers.During the patterning, the thermal reading sensor probes the topography of the patterned structure right after each patterning line when retracing back in contact mode, which leads to the simultaneous patterning and imaging capability of the NanoFrazor system as well as a closed feedback loop correction [27,28,41].The probe temperature is automatically calibrated through the system software according to the current-voltage characteristics of the Si tip [42].
The tSPL patterning parameters (such as write temperature, dwell time, and load) are adjusted to achieve the amine deprotection on the PMCC surface and PPA removal, while controlling the depth of the lithographic indentation.Details on the calibration of the optimal tSPL writing temperature, load and dwell time are reported in the Supplementary Note 5.

Nanoscale local biochemical functionalization
The biotin-streptavidin interaction is used as the bio-conjugation strategy to attach ad-hoc bioreceptors in designated regions of chips.Following the tSPL patterning to expose surface amine (-NH 2 ) groups, the chip is covered with a solution of 100 nM NHS-Biotin in dimethyl sulfoxide (DMSO) and incubated for 1 hour.The NHS ester groups react with the surface NH 2 groups, conjugating the biotin to the surface.The chip is then washed with a 1x phosphate buffered saline solution (PBS) and DI-H 2 O to remove any nonreacted material, and dried using compressed N 2 gas.The chip is then functionalized with 100 nM streptavidin in 1x PBS for 30 min.Following this step, the sample is functionalized with the desired biotinylated bioreceptor: 100 nM anti-SARS-CoV-2 antibody in 1x PBS, 100 nM anti-SARS-CoV-2 aptamer in DI water or 100 nM In uenza A anti-Hemagglutinin aptamer in DI water.
Polymer synthesis A solution of (E)-3-(4-(3-methoxy-3-oxoprop-1-en-1-yl)phenoxy)propyl methacrylate (Mcoum), 2-((((tetrahydro-2H-pyran-2-yl)oxy)carbonyl)amino)ethyl methacrylate (Mcarb), and azobisisobutyronitrile (AIBN) is prepared in THF in a glovebox (Mcoum:M carb:AIBN = 40:160:1).This solution is stirred at 70 ºC in the glovebox for 16 hours, after which, the solution is exposed to air and puri ed by adding dichloromethane, precipitating from cold hexanes, ltering, and drying under vacuum to isolate poly((tetrahydropyran-2-yl-N-(2-methacryloxyethyl)carbamate)-co-(methyl-4-(3methacryloyloxypropoxy)cinnamate)) (PMCC) as a white solid (see Fig. -20 nm resolution biochemical functionalization of FETs by tSPL.a 3D-Render of the NanoBioFET platform for FETs based multiplexed nanoscale biosensing.b Schematic showing the steps of fabrication in the NanoBioFET platform, from polymer deposition onto the FETs, to tSPL nanoscale activation of amine groups, and bioreceptor bioconjugation.c Thermal SPM topographical image of a tSPL nanoscale amine pattern fabricated in PPA/PMCC.In the inset, the depth pro le of the corresponding cross section AA', showing a FWHM of 15 nm for each sensor pixel.d and e Fluorescence microscopy image of biotinylated aptamers terminated with a red dye on a PPA/PMCC SiO 2 /Si chip.Speci cally, the image shows Biotinylated Anti-SARS-CoV-2 Aptamer.After tSPL patterning, the sample is covered with a solution of 100 nM NHS-Biotin in DMSO and incubated for 1 hour.The sample is then functionalized with 100 nM streptavidin in 1x PBS for 30 min.After washing and drying, the sample is functionalized with the red-dye tagged aptamer.f and g tSPL topography (f) and friction AFM (g) images of four PPA/PMCC coated GFETs with a channel width of 1 mm, where the areas above the GFET's channel (graphene) sensors have been patterned by tSPL to remove PPA and activate amine groups on PMCC.

Figure 4 Spike
Figure 4 Electrical measurementssamples SAR Cov-19 spike protein (Acro Biosystems) and live Covid particle (patient nasal swabs) with increasing concentrations through both channels.All experiments are run at 37°C.