Field‐Effect Transistor‐Based Biosensor for pH Sensing and Mapping

It is vital to acquire real‐time pH signals with high resolution as pH variation can reflect important information regarding health status and physiological environment. Field‐effect transistor (FET)‐based biosensors (bio‐FETs), a kind of potentiometric sensor, are being rapidly developed for pH detection due to their advantages of high sensitivity, low temperature dependence, and high portability. More importantly, the high scalability of bio‐FETs renders them applicable for achieving high spatial resolution in pH sensing. In this review paper, the design, operation principle, and critical characteristics of the FET‐based pH sensor are introduced. Then, the recent progress in pH mapping with FET sensor arrays, including static array where a sensor pixel is directly addressed by wiring, and active‐matrix array where a sensor pixel is accessed by additional switching FETs, is presented. Last, typical examples of pH sensor arrays in biomedical applications, such as health monitoring and DNA sequencing and elongation are presented.


Introduction
pH, defined as the H + activity in the aqueous solution, has a profound impact on biomedical applications, including cell DOI: 10.1002/adsr.202200098 signaling, tumor growth, protein activity, DNA deprotection, and viral infection. [1][2][3][4][5][6][7] The regulation of pH in the human body is crucial for maintaining a healthy balance in the biological environments, which can reflect health, physiological, and biological status of a person. The normal pH of the skin surface of most human body sites is acidic with pH value between 4.1 to 5.8, while an abnormal pH environment demonstrates skin lesions, such as acne vulgaris, irritant contact dermatitis, and atopic dermatitis. [8][9][10] It is reported that inflammatory skin diseases, itchy, dry, and aged skin show increased skin pH. [8] In addition, dysregulated pH is an adaptive feature of many cancers. In the normal differentiated adult cell, intracellular pH is ≈7.2, lower than extracellular pH. However, most cancer cells have higher intracellular pH of > 7.4 and a lower intracellular pH of ≈6.7-7.1. [5,[11][12][13] This "reversed" pH variation leads to a storm for cancer progression.
It has also been reported that pH in normal and tumor cells are different. The hypoxic phenomenon is often obtained during the development of tumors, resulting in intracellular and extracellular acidosis, which has been a trigger in the early stage of apoptosis. [14,15] It is found that many cellular processes and enzymatic reactions rely on pH. When the pH of body fluid decreases from 6.9 to 6.5, living cells are destroyed, potentially promoting tumor formation. [16] Acidic environment has been proven to inflame blood cells and lower oxygen levels, which can impair the function of DNA and hinder the metabolic state of living cells, causing liver, sweat glands, and liver exhaustion. [17] pH is also central in other fields, such as environmental science, chemistry, agriculture, and food processing. Therefore, it is essential to exploit suitable technologies for accurate detection of pH.
Methods for monitoring pH with highly complex biological systems; however, rely on situational restrictions and needs. Considering the importance and tight regulation of pH in biological systems, pH sensors nowadays have attracted immense research attention. Several approaches involving electronic sensors have been employed to record pH values, consisting of capacitive, chemiresistive, amperometric, and potentiometric technologies. [18] Besides, other transduction techniques, including optical, mass-sensitive, and cantilever-based pH sensors have also been applied to determine pH. [19][20][21] In this review article, www.advancedsciencenews.com www.advsensorres.com we focus on the potentiometric pH sensor strategy, whose development has undergone considerable effort over the past few decades.
Potentiometric sensors provide information about the analyte in the electrolyte by recording the potential difference between the two electrodes, that is, the potential difference between a reference electrode and a working/sensing electrode. Typically, Ag/AgCl electrode is one of the most popular references to provide stable potential relative to the working electrode. Ideally, no current flows between the two electrodes during the electrical measurements, establishing a local equilibrium with the sensor interface. [22] Potentiometric sensors for pH recording are widely spread on account of their advantages, including simplicity, low-cost, excellent electrical performance, as well as real-time monitoring. In addition, they are easy to integrate with readout circuits. In this regard, monitoring pH of human body fluids has attracted increasing attention for promoting biomedical applications. [23] Field-effect transistor (FET)-based biosensors (bio-FETs), a kind of potentiometric sensors, have been rapidly developed for pH detection due to a milieu of advantages, including lower temperature dependence, high portability and scalability, less bias at extreme pH, and significant durability, rendering it promising for biomedical applications. [24][25][26][27] FET-based biosensor is a kind of FET where the conventional dielectric structure and gate are replaced by sensitive membrane, reference electrode, and electrolyte. [25,28] Different from traditional electrical sensors involving redox reactions with certain electroactive species, bio-FETs depend on the effective charge of the binding biological analyte at the surface, which leads to the variation in effective carrier concentrations in the semiconductor channel through capacitive coupling. [29,30] By functionalizing bio-FET surface with a sensitive membrane, site binding event to targeted molecules such as proton in electrolyte can be achieved. Thus, FET-based biosensor could provide an excellent pH sensing platform.
Recently, with breakthroughs in artificial intelligence, an intelligent system is rapidly progressing, which needs pH sensor arrays with the function of real-time collecting of data and transmitting it to the cloud serves for artificial intelligence and big data analysis. [32] The rapid enhancement of intelligent systems has promoted the requirement for pH sensors from individual devices to sensor arrays with high spatiotemporal resolution, low power consumption, and improved performance. Sensor array adds new dimension to measurement, making it possible for spatiotemporal mapping of signals and real-time information feedback.
FET-based pH sensor array can achieve real-time pH mapping with high sensitivity and resolution. So far, two sensor array technologies, namely, static and active-matrix arrays, have been explored for pH sensor systems. The static array exhibits simple design and fabrication process but suffers from redundant driving components and leading wires, as well as low device density. [31] As for active-matrix array, the individual pixel could be active exclusively for the signal record, while other pixels are in the off state, which is an effective approach to achieve accurate readout, less electrode wires, and improved spatiotemporal resolution. The active-matrix array tenders a potential platform for integrating multi-sensing elements possessing high pixel density, equipping with intelligence system with enhanced functionality. Schematic illustration of pH sensor array including static and active-matrix array as well as corresponding applications, such as peripheral arterial disease (PAD), ischemia, e-skin, brain healthy management, high-throughput DNA sequencing, and elongation. Healthcare management: Reproduced with permission. [29] Copyright 2015, American Chemical Society. Reproduced with permission. [164] Copyright 2014, Wiley-VCH. DNA sequencing and elongation: Reproduced with permission. [1] Copyright 2022, American Association for the Advancement of Science. Reproduced with permission. [178] Copyright 2015, IEEE. Smart e-skin: Reproduced with permission. [39] Copyright 2022, American Association for the Advancement of Science. Reproduced with permission. [154] Copyright 2020, Wiley-VCH.
FET technologies lie at the heart of the sensor array, and there are high demands on the switch FET components' performance when integrated into the active-matrix array. For instance, the signal crosstalk issue between adjacent pixels and power consumption can be promoted by increasing current on/off ratio of FET. To achieve a fast readout of the detection pH signal, the FETs should exhibit high electrical performance, for example, high mobility. For a stable operation of the active-matrix array, the variation of the electrical characteristics needs to be suppressed, in which threshold voltage is the core point. Herein, we present a review of the recent progress on FET-based biosensor array for pH mapping; discuss the basic information of FET-based pH sensors including principle, device structure, and characteristics; emphasize on the static and active-matrix array for pH mapping; and introduce their applications in healthcare management, highthroughput DNA sequencing, and elongation (Figure 1).

Device Principle
In 1970, P. Bergveld first proposed the ion sensitive FET (ISFET) based on metal oxide semiconductor FET (MOSFET) with the liquid gate replacing the metal gate electrode. [32] The typical device structure of ISFET is depicted in Figure 2a, where the ISFET can  [191] Copyright 2020, Elsevier. c) Site-binding model. Reproduced with permission. [35] Copyright 2021, Wiley-VCH.
be converted into the bio-FET by modification of the sensing surface over the surface with specific receptor molecules. The potential distribution and electrostatic interaction can be modulated by the binding events referring to charged targets.
For the conventional MOSFET, when the device is operated in the linear region, the equation of drain-to-source (I ds ) is given by: where C OX is the insulator capacitance per unit area, μ is the carrier mobility in the channel, V G and V D represent the applied gate voltage and drain voltage, and L and W are the length and width of the FET, respectively. V TH is the threshold voltage, which can be expressed by: where the (Φ M− Φ S ) represents the difference between the metal work function (Φ M ) and semiconductor work function (Φ S ); Q OX , Q SS , and Q B represent charge in the oxide layer, charge at the interface between oxide and semiconductor layers, and depletion charge in semiconductor, respectively; and Φ f is the fermi potential. In the case of the bio-FET, as the gate is exposed to the solu-tion and gate voltage is applied through the reference electrode, the threshold voltage for the bio-FET (V TH * ) is expressed by: Compared to Equation (2), the metal work function (Φ M ) is replaced by the voltage applied to the reference electrode (E REF ), sol is the dipole potential of the solution, which is often regarded as a constant, and Ψ 0 is the interfacial potential between the insulator and electrolyte, which varies with pH. In this case, if we apply fixed V D and a suitable voltage in the reference (V REF ) simultaneously, I DS is pH dependent. It is also essential to ensure that E REF remains constant during the measurement by choosing an applicable reference electrode.
Considering that Ψ 0 changes with proton concentration, the concepts of the electrical double layer are needed to be discussed. Figure 2b shows ion distribution at the dielectric/solution interface. There are surface hydroxyl groups in the gate dielectric layer, which could adsorb anions and cations and create the net charge on the surface. According to the Gouy-Chapman-Stern (GCS) model, most anions tend to be close to gate dielectric because cations are willing to keep their hydrated molecules. [33] The charged ions will present near the surface in an aqueous solution environment. Thus, the route from the center of adsorbed charged ions to the surface is known as Inner Helmholtz Plane (IHP). The locus of centers from hydrate ions to IHP is called Outer Helmholtz Plan (OHP). The interfacial potential drop is owing to the GCS double layer at the insulator and electrolyte.
On the other hand, the theory of pH sensing is based on the site-binding model, which can be described by potential distribution at the electrolyte-insulator (E-I) interface. [34] Figure 2c illustrates the model of pH-induced surface charge states on the (3aminopropyl)triethoxysilane (APTES)-modified Al 2 O 3 /In 2 O 3 interface. The potential difference at the E-I interface is generated by the interaction of the insulating layer with ions in the electrolyte. The distance required for intermolecular chemical reactions can be carried out only at one molecular distance. However, salt ions will form hydrates with H 2 O molecules in the electrolyte, which cannot approach the E-I interface, and many sites appear at the interface. These sites can only bind with hydrogen or hydroxyl groups. Thus, the existence of terminating group (hydroxyl group or amino group) on the gate insulator surface allows for deprotonation or protonation in various pH electrolyte environment, resulting in negative (high pH), neutral, or positive surface charge (low pH). The total amount of surface charge (negative or positive) will influence the dependence of the surface potential of the bio-FET on the pH value in the electrolyte, depending on the material properties, interface potential, and hydrogen ions near the surface. [35] The equilibrium involves the surface hydroxyl group and the corresponding reaction: and H B + are hydrogen ions near the interface. In this case, a charged layer 0 formed on the gate insulator surface. At the same time, an equal amount of reverse space charges dl are induced in the solution, which could be explained by the GCS model. The relationship between the pH at the surface interface (H S + ) and bulk (H B + ) could be described by the following equation: Therefore, the sensitivity (S V ) is defined as: and where is dimensionless sensitivity parameter, which changes from 0 to 1; B pH S is the variation of net groups with the surface pH; and C diff is differential double-layer capacitance. Assuming that B pH S is infinite, when the pH B changes, pH S remains original state; at this time, = 1. The sensitivity reaches the Nernst limit about 59.2 mVpH −1 at room temperature.

Mode of the Signal Acquisition of FET-Based Biosensor
Over the last a few decades, biosensors have gained much attention owing to their well-characterized behaviors, which generally combine two basic components: biological recognition and transducer. The biosensor system can convert biological recognition events, such as antigen-antibody binding, nucleic acid hybridizations, enzyme-substrate reactions, and ions response, directly into detectable signals using bio-transducers. Recently, much attention has been paid to FET-based biosensor, a kind of biotransducer, due to its excellent sensitivity, miniaturization, and portability. [36] The biological recognition information acquired from the sample corresponding to either analyte concentration or activity can be transduced to electrical signals via bio-FETs. Subsequently, the electrical signals can be illustrated, amplified, post-processed, and displayed or sent to the data center for backend applications (Figure 3a). [37] To demonstrate the mode of signal acquisition, it is important to understand the operation mechanism of the FET-based biosensor. As depicted in Figure 3b, a typical FET-based biosensor is constructed, where the liquid gate electrode is away from the dielectric with the intervening electrolyte. [38] The conductivity of the FET-based biosensor can be modulated by the electric potential or charge density at the channel layer. The biological recognition events upon the channel region facilitate the electric field modulation of devices, which is analogous to change the gate voltage through the reference electrode. N-type device shows an improvement in conductance when binding biomolecules with positive charges at the device surface owing to the accumulated electrons in the channel region. However, p-type FET binding with molecules with positive charges will deplete the holes, resulting in a corresponding decrease in conductance.

Electrical Data Analysis
After obtaining a large amount of experimental data, it is essential to process the electrical data with appropriate analysis methods. In this section, two approaches to analyze the relationship between the concentration of the analyte and device response will be introduced. Figure 4a depicts the transfer curves (I ds −V Ag/AgCl ) of the bio-FET at various pH ranging from 3 to 10 with drain voltage (V ds ) fixed at 0.05 V. In order to understand the relationship between the response signal and pH, Ren et al. extracted threshold voltage (V th ) from the transfer curves of every pH value, which is demonstrated in Figure 4b. As a result, the device illustrated a linear dependence between pH values and V th with the linearity of 99.7% and the sensitivity of 61.9 mV pH −1 (Figure 4c). [35] On the other hand, Wang et al. reported another analysis approach to further calibrate FET responses, which is used to minimize the device-to-device variations. [24,39] Figure 4d shows the transfer curves in sweat samples at different concentrations of cortisol. During the measurements, the drain voltage (V ds ) was fixed at 10 mV when sweeping the gate voltage (V gs ). The calibrated response (ΔV) was determined by absolute sensor response (ΔI ds ) and the variation in source-drain current with gate voltage sweep (dI ds /dV gs ). [40] The equation is shown below: Reproduced with permission. [53] Copyright 2022, American Chemical Society (f). Reproduced with permission. [54] Copyright 2020, American Chemical Society (g). 1D material (nanowire) and corresponding bio-FET. Reproduced with permission. [56] Copyright 2021, American Association for the Advancement of Science (h,i). 2D material (graphene) and corresponding device (j,k). Reproduced with permission. [64] Copyright 2009, American Association for the Advancement of Science (j). Reproduced with permission. [62] Copyright 2013, American Chemical Society (k). Bulk material (thin film) and corresponding sensor (m). Reproduced with permission. [35] Copyright 2021, Wiley-VCH (l,m). Hybrid material and corresponding bio-FET. Reproduced with permission. [66] Copyright 2021, Springer Nature, Ltd. (n,o).
In this setup, the calibrated response was calculated when gate voltage (V gs ) was fixed at 200 mV ( Figure 4e). Therefore, the relationship between the calibrated response and cortisol concentrations is obtained in Figure 4f, illustrating reduced device-todevice variations.

FET-Based Biosensor Architectures
FET-based biosensors could be constructed to distinguish device architecture, which depends on the sensing mechanism. As shown in Figure 5a, the gate is positioned over the channel, where the devices with sensing films placed on the top of the channel layer are classical designs that can be used to monitor the properties of cells, such as pH variation. Compared with other structures, it is deemed a convenient method by bio-functionalization of semiconductor channels. [41][42][43][44] There is another design with a top gate structure, where the sensing films are embedded on the gate electrodes, which is applicable for peptides, enzymes, antibodies, and genomic probes (Figure 5b). [45] A third FET-based biosensor architecture is the side-gate geometry, in which the gate is placed in the same platform as the channel layer (Figure 5c). [46,47] It is commonly utilized in the neuromorphic application, where the biocompatible channel lies in direct contact with human tissue to record electrophysiological signals. [48] Another device structure is the extended gate, in which the sensing electrolyte compartment extends off the FET chip ( Figure 5e). One of the main advantages of the extended gate configuration is the separation of dry and liquid environments, where the device isolates the electronics part (FET) from the electrolyte sensing part, preventing possible ion corrosion and making the device more stable during long-term measurement. [49] The extended gate geometry is commonly used to detect pH, extracellular ions concentration, and enzyme-linked immunosorbent assay. [37,50] A special architecture of dual-gate geometry with additional gate dielectrics is shown in Figure 5d. The capacitive coupling of the two gate dielectrics significantly enhances sensitivity, even surpassing the Nernstian Limit (59 mV pH −1 ).

Materials for FET-Based Biosensor
An overview of typical semiconductor materials used for FETbased biosensors is presented in this section, focusing on 0D, 1D, and 2D, bulk, and hybrid materials. A comprehensive overview of these materials and correlated devices is shown in Figure 5.
Considering the advantages of quantum dots (QDs) with low cost, high biocompatibility, and ease of fabrication, QDs have been widely used in biosensing. [51,52] The TEM image of QDs shows the aspherical nanoparticle morphology with uniformly distributed particles (Figure 5f). [53] Fan et al. introduced the QDmodified gating electrodes in a FET-based biosensor, demonstrating high sensitivity and selectivity (Figure 5g). [54] The sensing mechanism of the device is that the coordination of QDs and analytes induces the change in capacitance of the double electri-cal layer close to the gate electrode, causing the shift in channel current.
It is commonly known that 1D nanowire has enhanced electrical performance, where the conductance of nanowire-based FETs could be restrained by changing channel density and electric potential, making bio-FETs to be potential candidates for biosensing. [55] The transduction behavior in the nanowirebased biosensor is similar to the planar sensor, except that the nanowire-based FETs have ballistic properties of small size with a bulk transport of carriers. [56] Thus, nanowire-based FETs, allowing direct conversion of biological features such as binding, enzymatic reactions, and charge transfer, ultimately lead to the detection of biomolecular based on an effortless approach. [57] Figure 5h shows the SEM image of silicon nanowire, reported by Hu et al. The prepared nanowire-based sensor has been proven to detect ions with single charge resolution. The events of emitting and capturing a single hydrogen ion at the liquid/solid interface are directly recorded by sub-10 nm electrical double layer-gated nanowire-based bio-FETs ( Figure 5i). [56] As the first case for exploration of graphene in 2004, currently, researchers have developed a broad application of 2D materials. [58] The 2D semiconductor materials with single-atom or few-atom thickness are commonly less than 5 nm, and the lateral size ranges from sub-micrometers to centimeters. [59,60] Graphene is one of the most heavily researched ones, whose lattice is composed of six-membered rings and sp 2 -hybridized carbons with a honeycomb structure ( Figure 5j) and the basic structure of 2D bio-FET is shown in Figure 5k. [61][62][63][64] The working principle of 2D bio-FET is the amplification effect of effective signal transduction close to the 2D material surface.
Bulk materials, such as metal oxide thin films typically serve as sensing materials for pH sensors owing to their excellent chemical resistance and mechanical strength. Previous reports for different metal oxides, such as In 2 O 3 , SnO 2 , and PtO 2 , have proven optimized pH sensitivity. [65] In these materials, hydroxy groups are formed on the electrolyte's sensing surface, followed by dissociating, causing electrochemical equilibrium. Based on this equilibrium, the potential at the interface of solid and liquid is stable and varies with the hydrogen ion concentration. Figure 5l shows the HRTEM image of the In 2 O 3 semiconductor layer and the corresponding device for pH sensing is depicted in Figure 5m. The device offers a high sensitivity of 62.4 mV pH −1 with good linearity of 99.3%. [35] The hybrid structure, such as WSe 2 /MoS 2 heterostructure, taking advantage of the interface, exhibits excellent electrical performance. The hybrid structure-based device can modulate gate transconductance by introducing a screening layer of the interface and channel charges. This enables super-Nernst pH sensitivity by decreasing effective back gate capacitance without improving the sensor dimension. [66] Figure 5n shows the HRTEM image of the WSe 2 /MoS 2 heterostructure, demonstrating excellent interface quality. Bio-FET device with the WSe 2 (top)/MoS 2 (bottom) vertical heterostructure is depicted in Figure 5o.

Surface Modification for FET-Based Biosensor
Besides sensing principle, device architectures, channel materials, and surface modification are also significant parameters in FET-based biosensor configuration. The operation of the FETbased biosensor is determined by the bioreactions between the recognition element modified on the device surface and analyte in the solution. The interaction will lead to either an increase or decrease in the transconductance of the device, which modulates the carriers transport of semiconductor layer. [67] Therefore, proper surface modification is imperative for enhancing both sensitivity and specificity of target detection, preventing interfering substance reactions and minimizing the noise. An essential step in fabricating a FET-based biosensor device is to anchor a specific recognition group with highly specific binding affinities to the analyte on the FET surface. Before that, the bifunctional linker molecules providing binding sites are often utilized to help anchor the bioreceptor elements to the surface. The modification layer at the gate or channel surface may also electrically passivate the device by changing local electrical field and refuse the other interactions of ions or non-specific molecules with the device surface.
For silicon and metal-oxide surfaces, covalent modification of the specific bioreceptors through chemical agents is commonly used. Prior to surface modification, silicon and metaloxide surfaces are often treated with oxygen plasma or ultravioletozone (UV-O 3 ), generating ─OH groups on the device surface, which facilitates the reaction with chemical agents. In this regard, organosilanes such as (3-aminopropyl)trimethoxysilane (APTMS), APTES, and 3-(trimethoxysilyl)propyl aldehyde are widely used. As shown in Figure 6a, a 1:9 (v/v) solution of APTMS and propyltrimethoxysilane (PTMS) were incubated on the surface by vapor deposition for 1 h at 40°C. [68] 90% PTMS deposited on the SiNW surface not only adjusted to the density of APTMS to decrease the steric hindrance of following bioreceptors but also to prevent non-specific reactions by forming a chemically insert surface. Functional linkers such as 3maleimidobenzoic acid N-hydroxysuccinimide ester (MBS) could be subsequently reacted with APTES with the amine group, and maleimide group introduced to incubate DNA-aptamer with sulfhydryl group. On the other hand, phosphonic acids with ─COOH groups are also available for metal-oxide or SiNWs covalent binding. [69] Phosphonic acid-based modification is less sensitive to moisture and less self-condensing than organosilane to facilitate storage and following reaction. [70] Figure 6b depicts that In 2 O 3 , NWs have been modified by 3-phosphonopropionic acid with ─COOH groups, which were subsequently activated using carbodiimide chemistry to covalently functionalize antibodies with amino groups. [71] In the case of the Au surface, one of the most optimum linker molecules is the sulfhydryl compound, such as cysteamine. As shown in Figure 6c, a self-assembled monolayer of cysteamine with ─SH group was deposited on the Au surface. Through a chemical linker (4-formylphenyl boronic acid), biomolecule-like dopamine can successfully bind to device surface. [72] In addition, the noncovalent modification approach has been widely used to functionalize bioreceptors. For example, 1pyrenebutyric acid N-hydroxysuccinimide ester (PASE) molecule as the linker can noncovalently couple to graphene surface through -interactions. Subsequently, the spike S1 proteins are deposited on the graphene surface through the reaction between the amine groups on spike S1 protein and hydroxyl-free succinimide ester on PASE (Figure 6d). Different types of substances, Reproduced with permission. [68] Copyright 2013, American Chemical Society. b) In 2 O 3 NW-based biosensor's surface modification route involves 3phosphonopropionic acid deposition, carbodiimide chemistry activation, and antibodies incubation. Reproduced with permission. [71] Copyright 2005, American Chemical Society. c) Reaction sequence for the functionalization of Au surface by a self-assembled monolayer of cysteamine with ─SH groups. Reproduced with permission. [72] Copyright 2013, Elsevier. d) Schematic diagram of graphene-based biosensor. PASE molecules as the linker can noncovalently couple to graphene surface through -interactions. Reproduced with permission. [192] Copyright 2021, American Chemical Society. e,f) Hydrogel film with functional groups and ion selective membrane can be deposited on the device surface by non-covalent bonding. Reproduced with permission. [67] Copyright 2023, Elsevier (e). Reproduced according to the terms of the CC BY license. [73] Copyright 2022, The authors, published by MDPI (f).
such as hydrogel and ion-selective membrane, can also be noncovalently modified on the device surface (Figure 6e,f). [67,73]

Characteristics of FET pH Sensor
In this section, we introduce the vital parameters such as stability, response time, sensitivity, reproducibility, hysteresis, and drift, which influence the sensing performance of bio-FETs.

Response Time
Response time is one of the most critical factors for potentiometric sensors, regarded as the time required for its open circuit potential (OCP) equilibrium. [74] Some factors influence the signal response time of the potentiometric sensor, including thickness, device structure, morphology, pore size, and pH range. It is reported that improved surface area-to-volume ratios would reduce response time, owing to the biomolecule binding kinetics. [75] On the other hand, nanoscale geometric structure facilitates great potential for device miniaturization, providing fast response time. [76] Channel materials with high capacitance, distinguished electrical conductivity, and low contact resistance will increase the electron transfer rate, resulting in dropping response time. The response time of pH sensors also depends on the pH range during the measurement, where most sensors demonstrate a shorter time at low and high pH values but slower in between. [74,[77][78][79] Typically, the pH sensor sometimes depicts a slower response time in the alkaline buffer compared to the acidic region. Libu et al. reported a potentiometric RuO 2 -Ta 2 O 5 pH sensor measured in various pH solutions. They observed that in the acid solution, the sensor shows a shorter response time (less than 8 s) and slightly longer in the alkaline buffer (less than 15 s), which is related to H + ions diffusion and dominant in the acidic buffer. [80] The response time performances of FET-based pH sensors based on different influencing factors are compared in Table 1.

Stability
The stability of various FET-based sensors is one of the most crucial features requiring continuous measurement or long incubation. The devices often suffer from instability when operating in the liquid environment such as buffer solution, pure water, or directly real biofluids, which makes it practically difficult to acquire stable signals. The stability could be influenced by many parameters including channel materials, liquid gate electrodes, device architecture, bias conditions, and pH level. As a result, in the process of creating biosensors, the transducers need to be adjusted to guarantee their stability and multiple efforts have been taken to boost their stability.
One of the current approaches to address the stability of FETbased biosensor focuses on choosing proper channel materials. Metal-oxide based biosensor such as IrO x shows promising stability when immersed in various pH electrolytes for a long time. It illustrates a fast and stable response in corrosive, nonaqueous,  [200] or aqueous media. [18] However, most metal-oxide materials, such as In 2 O 3 and ZnO always suffer from corrosion in liquid environment with extremely high or low pH values, owing to their chemical instability, which results in a shift of the electric signal.
Besides metal-oxide pH sensors, the conductive polymers with the ability of ion-exchanging serve well for developing pH sensors. [81] A popular polymer material is polyaniline (PANI). To evaluate the stability characteristics of PANI-based biosensors, Rahimi et al. immersed the devices into the various buffers at pH 6, 7, and 8 at the sampling frequency of 1 Hz over 24 h. [82] There was no significant output voltage change during the first 5 h, and the total voltage change was less than 12 mV over 24 h monitoring, showing good stability at long time measurement. However, the conducting and semiconducting polymer as active channel materials would react with the molecular oxygen under the device operation, resulting in electrochemical side reactions. Such side redox reactions can produce hydrogen peroxide (H 2 O 2 ) that may influence the device' stability. [83] In comparison, carbon-based materials such as carbon nanotubes gained much attention for pH sensing owing to their high surface-to-volume ratio and environmental stability. However, the stability performance of carbon nanotube-based bio-FETs always suffers from interference by the inhomogeneous interface between carbon nanotubes and metallic source and drain electrodes. [84] In addition, most ISFET configurations and glass-based electrodes exhibit poor long-term stability, while EGFET has demonstrated better performance. [85] One often neglected issue in FETbased biosensors is the requirement for a proper reference electrode. The electrode should maintain a consistent potential throughout the operation as any fluctuation could overshadow the real biological or chemical signals.

Signal Drift
Threshold voltage (V TH ) instability of nanomaterial-based potentiometric pH sensors, commonly known as drift, is another challenge and has become one of the main factors for FET-based biosensors to be difficult to apply in the marketplace. [86] The drift is defined as the potential difference over time where the bio-FETs are immersed in the buffer of permanent pH value. Different factors affect the device drift, including electrochemical non-equilibrium situation at the insulator/electrolyte interface, ion diffusion at the gate insulator layer, slow surface effect, and injection of defects from the solution to the insulator layer. [87] The summarized solutions to ameliorate signal drift are: choosing a suitable passivation layer to prevent ions in solution from the channel layer, designing special compensative readout circuits, and using the proper structure of the device with metal oxide as the gate contact. [35,88,89] Signal drift behavior of insulator films based bio-FET has been reported; According to Ren et al., In 2 O 3 -based bio-FET after Al 2 O 3 deposition using solutionprocessed approach shows a low current drift (0.35 mV h −1 ), compared to bared In 2 O 3 device (12.5 mV h −1 ), because Al 2 O 3 dielectric layer effectively passivates In 2 O 3 channel layer. [35] For 2D bio-FETs with a high-k insulator layer, the trapping/de-trapping model of carriers from the channel at the high-k insulator oxide layer is the main reason for signal drift. Thus, the excellent quality sensing layer with low defects and ions diffusion is essential to achieve a low signal drift on 2D material-based bio-FET. Wei et al. depicted that extended gate (EG) ISFET with top gate molybdenum disulfide (MoS 2 ) FET and EG Al 2 O 3 /h-BN sensing stack with 5 nm Al 2 O 3 by ALD method provides a drift value of ≈4 mV h −1 . [90] Another type of setup is FinFET in liquid gate configuration. Rigante et al. reported a FinFET with 20 nm critical features, using HfO 2 as gate dielectric and silicon nanowire as channel. The device showed excellent stability after 105 h test in the aqueous environment at pH 6, where the single wire FinFET exhibited a drift of 0.13 mV h −1 , a three-wire FinFET was 0.1 mV h −1 , and a five-wire FinFET illustrated 0.12 mV h −1 . [91] The experimental temperature also could change the drift, where signal drift was obvious at the first few days and later became stable quickly if the fixed temperature was relatively high. Table 2 illustrates the influence of several paraments on stability, such as type of setup, quality and composition of the sensing film, method of fabrication, and reference materials.

Hysteresis
Hysteresis is a parameter to evaluate the devices' reproducibility, which also limits the accuracy of the bio-FET during the measurement. While it is generally regarded as a memory effect, the device will suffer from serious errors if not considered. The hysteresis behavior refers to the fluctuation of the corresponding current performed by titration in both acidic and basic loops, which are correlated with surface defects of interaction between the gate insulator and solution ions at the liquid/solid interface. [92,93] Three possible reasons may influence the hysteresis: one is the design  of a new structure device. There are abundant corners and sidewalls in the channel layer, where ions diffusing to the sensing film constitute a high local electric field and forbid the ions to close the surface. [94][95][96] In this case, the extraneous ions of the sensing film are dropping, resulting in improved hysteresis. Another is the "passivation" parameter, in which the passivation layer could safeguard the semiconductor channel from extrinsic ions to build up religious hysteresis. Compared to ISFET with 30 nm SiO 2 (−54 mV), hysteresis of device with APTES/SiO 2 stack-sensing membrane (21 mV) significantly improved, owing to the well-ordered amino groups (−NH 2 ). [97,98] The third one is "pH loop" factor. It was found that the hysteresis width in the FET-based sensor is dependent on measurement pH loop time, where the hysteresis increases with the improvement of loop time. [99] On the other hand, post-processing of the measured signal by a given correction algorithm could also address the error from the hysteresis effect. [100,101] For comparison, Table 3 summarizes the measured potential values for different pH loop values to demonstrate hysteresis width in different FET-based pH sensors. Hence, the hysteresis width strongly depends on the device configuration, pH measurement loop, and passivation.

Sensitivity
Sensitivity is one of the most vital factors of FET-based sensors, which is determined by surface potential as well as the capaci-tance of gate dielectrics. A sensor with high sensitivity is always desirable, and the possible solutions are presented for sensitivity enhancement. The sensitivity can be improved by using a proper dielectric layer and modification of the semiconductors' properties. Nowadays, SiO 2 , Ta 2 O 5 , Al 2 O 3 , Si 3 N 4 , HfO 2 , and other high-k dielectric as sensing layer have been used in ISFET. Although intensive studies have been applied to various materials, their own drawbacks with signal drift, memory effect, and compatibility with CMOS technology limit further application. [102,103] Considering the disadvantages of SiO 2 films with low interface states that are easy to be damaged by the ions in the solution, multilayered sensing materials have been implanted in the device. Meanwhile, high-k dielectric films such as HfO 2 have gorgeous signal-to-noise ratios but suffer from non-ideal effects. Conversely, Al 2 O 3 layers have strong power against non-ideal performance. [104][105][106] Thus, Al 2 O 3 /HfO 2 stacking sensing layer achieves good sensitivity, enhanced chemical stability, and high output current owing to its rigid band structure. [107] However, for traditional ISFET, the maximum achievable sensitivity is 59 mV pH −1 due to the Nernstian limit, whereas DG-ISFET has overcome the situation of higher sensitivity beyond the Nernstian limit tuned by transforming the capacitance of dielectrics. The sensitivity of DGISFT relies on the surface electrolyte capacity of the top gate (TG) and the dielectric layers' coupling constant. [108] Bae et al. reported a DGISFET for pH sensing, where Si was utilized as a semiconductor layer, SiO 2 as the back  gate (BG), and Al 2 O 3 was regarded as TG as well as the sensing film. Compared to the single gate (SG) operation model (48 mV pH −1 ), the sensitivity of the double gate (DG) was up to 407 mV pH −1 . [109] On the other hand, by integrating signal amplification, a complementary inverter based on OECTs illustrated an ultrahigh sensitivity of 2300 mV dec −1 , which overcame the fundamental limit. [110]

FET-Based Biosensor Array Configurations
FET-based biosensor array is an essential implement used to acquire biological electrochemical signals. The two main categories of biosensor arrays are static and active-matrix geometry, making it possible for high-throughput data processing and image mapping. The schematic circuit diagram of the static array is shown in Figure 7a, which is an approach by connecting the outputs of all individual devices to each pixel for signal transmission. Each device is individually wired for multiple sensor arrays, so the loading wired design requires N 2 wires for an N × N array. [111] On the other hand, the active-matrix array consists of switch FET and sensor FET. By putting an active element in each cell of the device, the active-matrix array shows enhanced functionality, in which the active component in the array is referred to switch FET. [112] The circuit schematic of the active-matrix array is shown in Figure 7b, demonstrating diminutive cross-talk issue; simple readout circuit as well as less power consumption. [113][114][115] In addition, each individual cell of the active-matrix array can be controlled, leading to high spatiotemporal resolution. In terms of the wired connection, the number of wires for N × N pixel active-matrix array only needs 2N wires. As a result, compared to the static array, active-matrix array has the advantage of minimizing the whole size of the device due to the reduction of the number of loading wires, which allows for higher device density. Therefore, the static and active-matrix arrays are inspiring approaches, which have been developed for various biomedical fields such as bioelectronics, sensors, actuators, memory, and electronic skins (eskin) far beyond the display industry. [116,117]

Typical Fabrication Process for Array
During the past decades, the motivation for fabricating largearea and high-definition sensor arrays has driven the development of microfabrication technologies. Although many fabrica-tion toolkits in modern semiconductor techniques have been proposed, the low-cost and large-area manufacturing method is required for applications. [118] Figure 8 illustrates the state-of-the-art sensor system fabrication approaches, including inkjet-printed technologies, direct light pattern strategy, and photolithography method.
In recent years, inkjet-printed technology has been extensively explored to fabricate sensor systems. [119][120][121] Different from traditional photolithography, inkjet-printed technology directly deposits patterned layers onto the substrates without complex chemical processes and expensive instruments for constructing the microelectronic device. Due to these advantages, inkjetprinted technology has been employed to fabricate large-area sensor systems, and the fabrication process is shown in Figure 8a. [119] The direct light pattern is a resistance-free patterning method where the semiconductor channel can be patterned through masks and deep ultraviolet (DUV) processing (Figure 8b). The source and drain electrode also could be evaporated and deposited onto the substrate via the shadow masks, providing opportunities to obtain low-temperature and low-cost roll-to-roll printing of non-silicon-based devices. Low electrical performance and high-temperature processing of devices, especially for solution-based amorphous metal oxide semiconductors, are among the most significant trade-off issues because high temperature is not suitable for roll-up, flexible, and conformal electronics. [122] Here, direct light pattern technology provides the potential power to balance it. Although the direct light pattern and inkjet-printed strategies decrease multiple steps compared to the photolithographic process, they always suffer from nonideal layer registration and low pattern resolution. [25,30] Photolithographic process is one of the leading fabrication strategies for high-performance sensor systems. Figure 8c shows the typical steps for photolithographic technology, in which the materials are deposited on the whole surface and then etched to form patterns with the assistance of photoresists. [123] Photolithographic technologies can achieve high-resolution patterns down to the nanometer scale, but the trade-off is needed to be considered, including complicated instruments and multiple steps, such as spin-coating photoresists, light exposure, development, and stripping photoresists. [124,125] After repeating the photolithographic process, the integrated device with excellent pattern registration can be formed.

Static Array for pH Sensing
Static arrays have advantages in terms of reducing measurement time and testing errors, as well as facilitating the analysis of analytes compared to the single device. The regulation of pH is vital to keep a healthy balance in the biological environment for sustaining a lifestyle, which is the reason for dysfunction or severe disease within a physical system. Thus, the fundamental parameter of pH is of significant interest in the biomedical domain. This section will focus on recent progress in pH sensing using the static array.
Aroonyadet et al. reported a top-down fabricated In 2 O 3 nanoribbon FET sensor array using radio frequency (RF) sputtering. [126] Figure 9a shows the wafer-scale photo of In 2 O 3 nanoribbon FET array with 100% yield, and the inset image ex-hibits a magnified view of one chip containing four subgroups of six FET devices. The fabrication process is illustrated in Figure 9b, in which 500 nm Si 3 N 4 is first deposited on silicon substrate served as dielectric, followed by deposition patterned source and drain electrodes (5/45 nm Ti/Au) using photolithography and e-beam evaporation. Then, the nanoribbon-sharped mask is defined by the photolithography step, and In 2 O 3 is deposited by sputtering. Thus, In 2 O 3 nanoribbons are formed after the lift-off process. After fabrication of multiple single devices on the same substrate by the above method, the static array is created. Owing to the bind-site model, hydroxy groups on the In 2 O 3 nanoribbon film will deprotonate or protonate, which could successfully detect pH value. In 2 O 3 nanoribbon FET array shows excellent sensitivity in both physiological pH ranging from 6.7 to 8.2 and a wide range between 4 to 9 with good long-term stability and uniform electrical performance. However, with the rapid development of precision medicine, designing flexible and wearable platform for continuous monitoring is demanded. [26] Flexible biosensor array is an intelligent electronic device worn on human bodies as an accessory. [127][128][129][130][131] Recent improvements in the microelectronics and biosensor domain offer the possibility of employing wearable sensor arrays to continuously monitor personal physical status without restricting user's movement. [132][133][134] However, the wearable bio-FET array with the function of detecting physiological signals remains a challenge. To meet the requirements above, new flexible systems based on the FET biosensor array are introduced. Recently, a highly sensitive and conformal In 2 O 3 nanoribbon FET array was proposed by Liu et al. Figure 9c Figure 9b. [126] The device reflects an improved conductivity when the pH value of the buffer decreases (pH range from 10 to 5) because hydroxyl groups on the In 2 O 3 nanoribbon surface are protonated owing to more H + ions in the solution, which results in the positive gating effect on semiconductor channel.
Rim et al. also demonstrated a flexible static array on ultrathin polyimide (PI) films, and the schematic view of the array is shown in Figure 9e. [29] First, the ultrathin PI films (2 μm) are spin-coated on glass substrates, followed by spin-coating In 2 O 3 films patterned by direct light pattern technology. The interdigital electrodes are deposited on the patterned In 2 O 3 films by ebeam evaporation using shadow masks. Next, the prepared devices are delaminated underwater; then, transferred to artificial polydimethylsiloxane (PDMS) substrate, where an ultrathin flexible PI-based array conformably contacts the PDMS. The crosssection view of the individual device is shown in Figure 9f, where the device structure is bottom-gate and top-contact; meanwhile, In 2 O 3 serves as a semiconductor layer. This flexible array exhibits good sensitivity with a linear range from 5.5 to 9.0, and the rate of linear pH response is ≈8.6 ± 0.4 μA pH −1 .
Wearable sensing technology that detects biological signals such as pH in human skin or biofluids provides specific personalized physiological and psychological information. [135,136] How- Figure 8. Flow diagram of typical fabrication processing to construct sensor system. a) Illustration of inkjet-printing technology. Reproduced with permission. [119] Copyright 2021, Springer Nature, Ltd. b) Illustration of direct light pattern strategy. Reproduced with permission. [122] Copyright 2014, American Chemical Society. c) Illustration of the photolithographic process. Reproduced with permission. [123] Copyright 2022, American Chemical Society.  [126] Copyright 2015, American Chemical Society (a,b). c) Photograph of flexible In 2 O 3 pH sensor array. d) SEM image of an individual In 2 O 3 nanoribbon sensor (L/W = 500/25 μm). Scale bar is 1 cm. Reproduced with permission. [193] Copyright 2018, American Chemical Society (c,d). e) The schematic image of the flexible In 2 O 3 array. f) Cross-section view of individual device. Reproduced with permission. [29] Copyright 2015, American Chemical Society (e,f). g) Expanded image of the wearable sensing system in which the sensor array, microfluidic system, liquid crystal display (LCD), and flexible printed circuit board (FPCB) components are integrated to form pH sensing smartwatch. h) Photography of the FET-based pH sensor array with In 2 O 3 semiconductor layer based on a flexible polyimide substrate. i) Unknow pH solution compared by FET sensor array and the pH meter. Reproduced with permission. [39] Copyright 2022, American Association for the Advancement of Science (g-i). ever, the existing wearable system lacks the wireless and dynamic capabilities to capture physiological changes accurately and seamlessly. Owing to the weakness mentioned above, FETbased pH sensor is currently unavailable to advance personalized precision medicine. Wang et al. designed a flexible In 2 O 3 FET array-based smartwatch equipped with a multichannel, custom, autonomous, and self-referencing measurement unit, achieving seamless and real-time data acquisition, where FET arrays are modular and generalizable. [39] They could be straightforwardly applied in mobile and wearable formats for pH detection in sweat or body fluids. As shown in Figure 9h, thinfilm In 2 O 3 is deposited on a flexible substrate (polyimide) via the solution-processed method and then patterned by the photolithography approach. Interdigitated Ti/Au electrodes are de-posited to form source and drain electrodes. By functionalizing with (3-aminopropyl) triethoxysilane (APTES) diluted with trimethoxy(propyl)silane (PTMS) (1:9 v/v ratio), a pH-sensitive film is self-assembled on an In 2 O 3 channel, and pH can be successfully detected through protonation/deprotonation of amine group in APTES, which changes the conductivity of semiconductor channel. Subsequently, the FET array is embedded in a thin-film microfluidic system and a liquid crystal display powered by a 110-mAh lithium battery to produce a skin-adherable "smartwatch" (Figure 9g). The device performs pH measurements ranging from 4.6 to 7.6 via implementing calibrated response method, [137] which shows a high linearity of 99%. As shown in Figure 9i, the pH value of the FET-based sensor is considerably close to the pH meter with r = 0.999, P < 0.001.  [194] Copyright 2017, IEEE (a,b). c) A 256 × 256 implantable measurement setup for pH sensing. d) Operating principle of a single device with two transistors by converting proton change in electrolyte to the electrical potential variation. Reproduced with permission. [195] Copyright 2019, IEEE (c,d). e) The cross-section view of individual ISFET structure for H + , Na + , K + , and Ca 2+ detection. f) Ion mapping image including H + , Na + , K + , and Ca 2+ . Reproduced with permission. [145] Copyright 2020, American Chemical Society (e,f). g) Image of pH sensor array based on packaged CMOS integrated circuit consisting of 64 × 64 = 4096 Al pads and 16 × 16 = 256 electrochemical cells. Inset image is the magnified view of the individual cell consisting of an outer cathodic Pt ring electrode combined with four underlying Al pads, an inner anodic Pt ring combined with another four Al pads, and a center OCP sensor with the circular Pt electrode combined with one underlying Al pad. h) Principle of the device to produce acids and bases by oxidation-reduction reaction on the surface of chip. i) pH localization in an individual concentric pixel, showing nonexplosive diffusion as time goes by. j) Array-wide pH localization for selected pixel. Reproduced with permission. [1] Copyright 2022, American Association for the Advancement of Science (g-j). k) Schematic of 5 × 5 flexible active-matrix array consisting of IGZO switch FET and sensor FET. l) Mapping image of pH distribution after dropping pH 2.5 solution to pH 6.5 mother liquor bath. Reproduced with permission. [154] Copyright 2020, Wiley-VCH (k,l).

Active-Matrix Array for pH Sensing
Recently, the active-matrix array has attracted much attention in pH sensing, which allows individual and random access to each pixel with fast addressing speed and keeps a high device density owing to shared electrode wires and active components. [138] The circuit diagram of the active-matrix array is shown in Figure 7b, where the drain electrodes of each wire are connected to the data line, gate electrodes are coupled with a scan line, and source electrodes are grounded. As the malfunction or breakdown of the individual device will lead to an entire row or column failing, the reliability requirements of the cross-point insulation, gate dielectric, and electrodes in the active-matrix array are needed to be considerably improved than the static array Ion-sensitive field-effect transistors (ISFETs) have gained extensive research interests since their invention by Bergveld in 1972. Cong et al. designed a 3600 × 3600-pixel ISFET array and corresponding readout circuitry using a conventional 0.18 μm CMOS process, which proposes a large-scale and dense pH sensor array with a simple architecture. Figure 10a shows the layout of the ISFET array chip, consisting of ISFET array and peripheral circuits. The extended gate ISFET is depicted in Figure 10b, in which the polysilicon gate is combined with the passivation layer through the intermediate metal layer. Typically, Si 3 N 4 is a pH-sensitive film that prevents the metal top gate from the measured buffer and is highly sensitive to the analyte's proton. Thus, the drain current and the threshold voltage of the device change with various pH values. This active matrix-array achieves high pH sensitivity (73.56 mv pH −1 ), large-area (3600 × 3600), and high-throughput (26 frame per s) properties, which have great potential in the biological application.
While active-matrix sensor arrays for detecting in vitro electrophysiology, such as pH, have been developed, signal recording technology in vivo remains limited. The electrophysiological signal recording methods based on neural interfacing devices, such as microelectrode array and multichannel probe array, have been proposed for various biomedical applications. [139,140] In addition, an implantable neural recording probe has been employed as formidable technology to monitor brain activities with high spatiotemporal resolution. [141,142] However, the neuroimaging tools with a low spatial resolution for neurochemical detection, such as protons, have limited the accurate measurement. An implantable 256 × 256 active-matrix array with an integrated readout circuitry for pH recording and imaging is reported by Lee et al. Figure 10c demonstrates the block diagram of the active-matrix array system, where the 256 × 256-pixel arrays are integrated with the readout circuit and the entire chips are 512 × 512 μm and 10.42 × 3.55 mm, respectively. The proposed sensor array structure is fabricated based on CMOS technology, including the deposition of Ta 2 O 5 as the ion-sensitive film. The critical point of this array with the high resolution is achieved by employing two transistors (2-T) in each pixel circuit. This 2-T in-pixel structure makes it possible to reduce pixel pitch to 2 μm, which is smaller than previous work using seven transistors. [143] Figure 10d shows the cross-section view of a single pixel. The operation mechanism is that the threshold voltage (V TH ) shift of the device changes with different concentrations of hydrogen ions when the constant gate voltage is applied to the reference electrode. V TH shifts mean that the potential depth beneath the sensing area changes, and these vibrations are read out by the peripheral circuit. In order to realize the implantable function, the pixel array is willing to be on the side near the edge. This sensor array could detect pH distribution change during the inserting trials in brain-like objects and then realize 2D image mapping in real-time measurement. Therefore, we confirm that the implemental active-matrix array could visualize spatial pH distribution successfully.
Although large-scale hydrogen ions mapping has been achieved, high-resolution, real-time, and simultaneous monitoring of multiple ions and image mapping using sensing learning are still challenges with existing spectrophotometric, chromatographic, and potentiometric techniques. [144] Moser et al. designed an active-matrix array of 1024 ISFETs employing sensor-learning technology to perform multi-functional mapping for concurrent detection of hydrogen, calcium, sodium, and potassium ions. [145] The ISFET array as the solid-state sensor is based on unmodified CMOS technology, eliminating the requirement for unique fabrication processes while providing the advantages of scalable and ease of integration. [146] The mechanism of the array is dependent on the modulation of the threshold voltage by generating electric field when ions are bonded to the ion-sensitive membrane. Analyte-specific ions sensitivity membrane is incubated on the surface of an ISFET array chip, which yields pixels with quasi-Nernstian sensitivity to ions to train array and discriminate species. Taking advantage of the pH sensitivity of Si 3 N 4 passivation film in CMOS technology, ISFET operation toward proton sensing is demonstrated. [147] Then, the sensor learning of the IS-FET array depending on back-end offline algorithms is applied in the device to process data and generate a sensitivity matrix for accurate monitoring functions via ion mapping. Figure 10e shows the basic structure of the individual pixel emphasizing that both MOSFET and ISFET sensors can be integrated to the same substrate. The device is based on standard CMOS technology using the extended-gate method, consisting of extending the ISFET gate to the top metal through the vias. The sensing area is considered the top metal region that is combined with the solution via the passivation layer. As shown in Figure 10f, this 32 × 32 ISFET array achieves real-time ions mapping, including H + , K + , Na + , and Ca 2+ . The notion of sensing learning based on offline training algorithms is also introduced to multi-functional ISFET array to accommodate the versatility through training the platform for acquiring more accurate mapping results.
The pH sensor array, a potential platform for pH mapping, has been extensively studied. However, most previously reported arrays are noncapable of in situ regulation of pH by simulation but only address the pH values. In addition, owing to the fast diffusion of protons in water, achieving pH localization by array remains challenging. [148] In a traditional pH sensor array, due to the diffusion effect, the pH values between pixels will affect each other; so, only the pH value of the whole solution can be measured, and the solution in each pixel cannot be accurately localized. [149,150] Aside from these issues, few previous studies have realized an on-chip pH sensor that can localize the pH across the array in real time, except a pH indicator based on a fluorescence approach with off-chip optics. [151] Here, Jung et al. proposed a fully electrochemical and high-density array for pH localization constructed on the CMOS microelectrode array. [1] As shown in Figure 10g, the CMOS integrated array consists of an array of 64 × 64 = 4096 aluminum (Al) pads electrodes and an electrochemical array of 16 × 16 = 256 cells with a single pad dimension of 10.5 × 10.5 μm and a pad-to-pad pitch of 20 μm. [152] Each pixel consists of an inner andic platinum (Pt) ring connected to four underlying Al pads, as well as an outer cathodic Pt ring connected to another four underlying Al pads, arranged in the concentric geometry, and also an open circuit potential (OCP) sensor with a circular Pt electrode localized in the center that connected to one underlying Al pad. The OCP sensor in the device is regarded as a variant of ISFET with pH-sensing capabilities. Each pixel can not only electrochemically produce protons with a positive stimulus current via the inner anodic ring but also electrochemically generate the basic wall with the negative stimulus current via the outer cathodic ring, as well as a central OCP sensor to monitor pH. The mechanism is depicted in Figure 10h, where the anodic Pt ring is applied to a positive current to oxidize 2,5-dimethyl-1,4-hydroquinone (H 2 Q) into 2,5dimethyl-1,4-benzoquinone (Q) for generating protons, and the cathodic Pt ring is injected to the negative current to generate quinone dianions (Q 2− ) and base molecules that could neutralize protons. Thus, the base molecules produced by the outer ring act as an electrochemical wall that limits the number of protons generated from the inner ring, resulting in the pH being confined to the pixel and not diffusing. As shown in Figure 10i, an acidic pH of 5.26 is observed and remains within the pixel of current stimulation, while the pH measured outside the activated pixel is more than 7 during the stimulation, demonstrating that the pH inside the pixel maintains localized rather than bursting. Figure 10j shows pH can be localized in any position and any number of chosen pixels activated with the corresponding current.
Although there have been enormous efforts of active-matrix arrays aimed at enhancing pixel density, device size, cross-talk issues, and electrical performance, the flexible sensor still needs to be considered due to the requirement of the Internet of Things (IoT) system. [153] Honda et al. designed a macroscale detachable pH sensor array that consists of a reusable, flexible ISFET array and a disposable flexible pH-sensitive membrane array through the flexible electrical connector, where ISFET is proposed to be extended gate geometry. [154] Figure 10k illustrates the optical image of a flexible 5 × 5 pH sensor array, in which polyaniline (PANI) is employed for a flexible pH-sensitive membrane because it has the capacity of low environment toxicity and can also detect protons by protonation/deprotonation effect on its film. The reaction between PANI and protons serves as potential pro-duction, leading to a variation of the transistor potential because PANI electrodes are combined with gate electrode through the flexible connectors. Thus, by observing the change of drain current, the pH value of the buffer can be detected successfully. Figure 10l shows the pH mapping results when adding a pH 2.5 solution into a pH 6.5 electrolyte bath after 6 s, demonstrating that this flexible array offers a potential platform for flexible electronic domain.

Applications of pH Sensor Array
This section describes applications of relevant FET-based pH sensor arrays in healthcare management, high-throughput DNA elongation, and sequencing. The application proposed highlights specific features of the pH sensor array.

Healthcare
The pH regulation in the human body is critical to maintaining a healthy balance in biological environments to sustain life. Variations and disturbances in pH may predict dysfunction or disease, including peripheral artery disease, ischemia, atherosclerotic plaque development, tumor growth, cancerization, brain dysfunction, and inflammation. [79,[155][156][157][158] Nowadays, FET-based biosensor arrays are reported to predict the physiological alteration mentioned above. Hence, long-term and continuous employment of a non-invasive FET-based pH sensor array could greatly help assess personal health and early disease prediction, allowing for timely and effective treatment. This part focuses on pH regulation in healthcare management, specifically ischemia, peripheral artery disease, and brain health management.
Peripheral arterial disease (PAD) is the third most common atherosclerotic vascular disease, second only to stroke and coronary artery disease. It triples the risk of cardiovascular disease, leading to amputation in severe cases. [159] However, most patients with chronic or mild PAD do not experience uncomfortable clinical symptoms that make them unaware of the onset or progression of the disease, where the PAD will induce disrupted blood flow, causing ischemic symptoms such as intermittent claudication. [160] Although ankle-brachial index (ABI) with high sensitivity and specificity provides effective monitoring, it is unable to assess tissue damage caused by blood flow disturbances and hypoxic injury. [161] As hypoxia enhances anaerobic glycolysis causing lactic acidosis, observation of pH values can fully reflect tissue damage resulting from ischemic injury. [162] Thus, using pH sensor array can effectively prevent PAD occurrence. However, there are disadvantages to the lack of accuracy of pH monitoring in human skin owing to the presence of pollutants such as cosmetics, sweat, and oil contamination. In addition, skin with inflammation or infection will change the skin pH level, hampering accurate measurements. [8] The microneedle system is a great candidate for satisfactory penetration, which can access dermal sites, cover large skin areas without surgery, and record physiological signals such as pH with high accuracy. Therefore, Lee et al.   Figure 11a, where the Au electrode is deposited on the device via a shadow mask and perylene serves as the passivation layer. To evaluate the pH values, Ag/AgCl and polyaniline (PANI) are electrochemically deposited on the microneedle. The average sensitivity of 5 × 5 pH sensor array is as high as 82 mV pH −1 after covering with poly(3-methoxypropyl acrylate) (PMC3A) that is a kind of blood-compatible material, demonstrating large-area sensitivity and high yield. Then, conformable microneedle pH sensor array is used to monitor the dermal pH in a rat model with PAD, to validate the biomedical application. To mimic PAD in the rat, the femoral artery is chosen to be ligated and the pH value of the lower extremity is recorded one week after surgery ( Figure 11b). As shown in Figure 11c, the microneedle array exhibits gradual pH variation of the lower extremity from the thigh to the perineal region in a PAD rat model, while the control group shows a constant pH value over the whole recording area. The distal acidity is higher than the proximal region, demonstrating that the damage of ischemic tissue is more abominable with increment distal distance from obstructive lesions. Thus, this flexible microneedle pH sensor array is a potential platform to provide reliable data and timely treatments for patients.
Another disease requiring intervention at an early stage is ischemia. Ischemia is defined as the restriction in blood supply to tissues, muscle groups or organs, especially the heart muscles, leading to the shortage of oxygen of cells that only generates ATP via glycolysis and creates H + as the outgrowth, which impairs cellular metabolism. The accumulation of lactic acid explains this myocardial acidosis of cells during anaerobic glycolysis. These electrophysiological changes are the electrical substrate for developing ventricular arrhythmias, causing the high mortality associated with cardinal ischemia. [163] Instability in pH level is known to indicate the abnormality of metabolism, and it does allow to assist clinicians by providing available cardiac information through the advanced implantable platform or specialized surgical tools. [164] However, it is hard to develop technology for recording and addressing pH in cardiac due to its curvilinear shape, heterogeneous surface, and continuous chattering. Previous research reports printed circuit board electrodes showing good sensitivity and repeatability in buffer solution; however, the electrical performance tends to fluctuate under in vivo animal models. [165] Besides, traditional pH sensor, such as glass electrodes are bulky, rigid and easy to break, thereby, difficult to be miniaturized for application. Currently, Chung et al. have developed a conformal matrix-addressed pH sensor array based on IrO x for recording the signals of cardiac ischemia-reperfusion. [164] The sensing setup divides into miniaturized sensor array and a reference electrode. The fabrication of devices exploits various techniques including conventional microfabrication, printing, and electrochemical approaches. Figure 11d shows the basic structure of a typical IrO x matrix where the magnified views in the lower state depict Au electrodes before and after electroplating IrO x . The sensor arrays consist of Au traces encapsulated by polyimide (PI), and the interconnections are designed by thin and filamentary serpentine to reduce material strains during mechanical deformation. Then, after employing transfer printing technology, the array can integrate onto the elastomer sheet or inflatable balloon Figure 11. Application of pH sensor arrays for healthcare management. a) Structural image of flexible pH array with microneedles. b) Optical image of the operation performed with peripheral artery disease (PAD) model. Scale bar is 2.5 mm. c) Image demonstrates location and pH distribution with pH sensor array in a PAD model rat. Reproduced with permission. [196] Copyright 2021, American Association for the Advancement of Science (a-c). d) Photography of IrO 2 pH sensor array fabricated on flexible substrate. e) Human right ventricular wedge in contact with pH array. f) Time against pH curve with no-flow ischemia, reperfusion, and baseline. g) pH maps after 18 min of ischemia. Reproduced with permission. [164] Copyright 2014, Wiley-VCH (d-g). h) Optical image of proton image sensor array consisting of sensor chip packaged with printed circuit board (PCB). i) Visual stimuli of eight different directions, positioned 13 cm away from the eye of mice. j) Visibility of changes in pH by visual stimuli. Reproduced with permission. [167] Copyright 2020, Springer Nature, Ltd (h-j).
catheter, following which IrO x layer is successfully deposited on the Au electrodes by electrochemical method. The minimized sensing area of 1 mm, sub-millimeter-scale arrays, and precision pH sensor distribution make it possible for in vivo testing. As IrO x is sensitive to pH, the device matrix array shows an average sensitivity of 69.9 mV pH −1 with a linearity of 99.7% and a response time as low as 0.5 s. The linear response of the device to temperature is ≈1.6 mV C −1 , and minimal influence of the extracellular ions is <3.5 mV. Figure 11e shows the IrO x -based pH sensor arrays are tested on human right ventricle (RV) in open circuit potential (OCP) configuration. Time against pH curve is plotted in Figure 11f, indicating a decrease in pH from 7.4 to 6.55 during the cardiac ischemic episodes and the pH mapping result of array is shown in Figure 11g.
The regulation of pH level is also essential for the human brain to maintain normal function. When the brain is in healthy status, extracellular pH is maintained between 7.2 to 7.4, while intracellular pH ranges from 6.8 to 7.0. [166] Besides, protons can directly participate in neurotransmission, illustrating that the correlation of pH changes with brain function is increased under physiological and pathological conditions. [167] A 128 × 32 pixel proton active-matrix sensor array applied for analyzing the brain in vivo is reported by Horiuchi et al. Figure 11h shows the photograph of complete proton active-matrix pH sensor array, consisting of a sensor chip integrated with printed circuit board (PCB), in which the sensor chip is used to record pH variation and is combined with PCB by wire bonds passivated by black epoxy. The dimension of the sensing area in the sensor chip possesses a length of 11.47 mm, a width of 1.76 mm, and a thickness of 0.1 mm. The reduced scale of the sensor array minimizes the damage to surrounding brain tissues for living animals. The temporal and spatial resolution of the proton sensor array are 50 frames s −1 and 23.55 × 23.55 um per pixel. The mechanism of the sensor array to detect proton concentration is based on chemical equilibrium in terms of Si 3 N 4 membrane that alters the surface potential at each cell. [168] The pH sensor array with high sensitivity (51.6 mV pH −1 ) and excellent spatial-temporal resolution is used to record localized pH change in the brain of live mouse, specif- Figure 12. Application of pH sensor arrays for high-throughput DNA sequencing and elongation. a,b) ISFET array for high-throughput DNA sequencing. c) Packaged array with molded fluidic lid to allow sequencing reagents. d) The mechanism of DNA sequencing. Reproduced with permission. [178] Copyright 2015, IEEE (a). Reproduced with permission. [177] Copyright 2020, Springer Nature, Ltd (b-d). e,f) pH sensor array and mechanism for highthroughput DNA elongation. g) The mapping results after Cy5-labeled ddATP incorporated into the deprotected substrate single-stranded DNA. Reproduced with permission. [1] Copyright 2022, American Association for the Advancement of Science (e-g).
ically in the primary visual cortex (VI) region. As shown in Figure 11i, the researchers evoke brain activity in the V1 of mouse by visual stimulation, causing changes in pH and the visual stimuli consists of drifting gratings in different orientations. Figure 11j shows the changes in pH caused by visual stimuli, in which diverse spatial patterns of pH variations in V1 region of living mice are observed in response to visual stimuli of eight drift gratings in different directions. These results illustrate that the sub-second order temporal variation in brain pH can be observed by the proton pH sensor array, and further confirm that pH change of the array can reflect actual variation in brain pH. Thus, the pH sensor array has potential to seek the relationship between different pathologies and cellular pH dysfunction.

High-Throughput DNA Sequencing and Elongation
Massively parallel DNA sequencing has been one of the most successful and widespread applications for pH sensor arrays, which has a profound impact on life science, biotechnology, and medicine for a lower-cost and more scalable solution. [169][170][171] The decrease in the time and cost of DNA sequencing has opened up new domains, including human genetics, infectious diseases, personal genomes, as well as ecological and cancer studies. [172][173][174][175][176] Nowadays, researchers design a DNA sequencing technique based on a low-cost and scalable semiconductor fabrication technique in which the integrated circuit is engineered to directly perform non-optical DNA sequencing of the genome. [177,178] The most widely used technology of the CMOS process is employed to fabricate high-density and large-scale production arrays. The device consists of ISFETs in perfect register with plenty of wells, providing parallel and simultaneous independent sequencing. As shown in Figure 12a, each individual site pixel of the sensor array contains a primary FET-based sensor aligned with the upper well so that the sensing areas of transistors take up the majority of the bottom of the wells, and the adjacent pixel select transistor is utilized to capture and for the readout sensor signal. [178,179] Figure 12b shows the array chips manufactured on wafers, and the device passivated with a flow cell is depicted in Figure 12c, in which fluid above the sensor is isolated with the supporting electronics to protect electronics and offer fluidic sample loading during DNA sequencing. [177] The crosssection image of the individual pixel of the array is shown in Figure 12d, where each sensor element with an individual floating gate is connected to the underlying ISFET. The device with tantalum oxide layer provides the sensitivity of 58 mV pH −1 . The sequencing process starts when the DNA is cut up into millions of fragments, and each fragment then attaches to its own bead. These beads flow across the chip, depositing into the well. DNA polymerase and sequencing primers are then bound to the template and pipetted into the loading port. Protons are released, a nucleotide is incorporated into the growing DNA strands, changing the pH of the solution in the wells. This induces a variation in the surface potential of the ISFET beneath the well, converting pH into voltage. Thus, this voltage change is recorded, indicating that the nucleotide was incorporated. For example, cytosine is a polymerase of C nucleotide in DNA strand, if a complementary G is present. If the nucleotide is not complementary to the next base, there are no ions released, and no voltage changes recorded. If there are two identical bases next to each other, two nucleotides are incorporated: the voltage doubles and the chip will record two identical bases . This process happens simultaneously in millions; that's why it's often described as a massively parallel sequencing. The pH sensor platform is touched by creating a direct connection between chemical and digital information, providing a fast, simple, scalable sequencing solution that is less expensive and more reliable.
The second example employing pH sensor array is high throughout DNA elongation by the pH-regulated enzyme at any selected cells. The utility of synthetic biology has yielded numerous chemical substitutes for natural DNA and RNA, supporting the development of diagnostic and therapeutic applications. [180] Synthetic DNA is an attractive long-term date storage medium owing to sustainability, longevity, density, and ease of replication. [181] The array provides the DNA synthesis parallelism in which each sequence is elongated at any site simultaneously for high throughout DNA elongation. Jung et al. used pH sensor array to execute DNA elongation, combining dideoxyadenosine triphosphate (ddATP) to the substrate of single-stranded DNA molecules by employing pH-regulated enzymes of chain-terminating nucleotides at any locations in parallel. [1] The optical image of packaged sensor array featuring 64 × 64 = 4096 Al pads and 16 × 16 = 256 electrochemical cells, is shown in Figure 12f, and the detailed geometry and electrical performance were described previously (Figure 10g-j). As shown in Figure 12e, the substrate DNA molecules are incubated on the glass ceiling, facing the sensor array with a distance of 14 μm. Only when the substrate DNA strands are deprotected in an acidic environment can they selectively incorporate nucleotide molecules in solution. A local acidic microenvironment is created by applying an appropriate current to stimulate the sensor array to form a local pH of ≈5.5. Subsequently, Cy5-labeled ddATP is added to the electrolyte, which could enzymatically incorporate with deprotected DNA strands, where Cy5 is a kind of fluorescence pattern. As depicted in Figure 12g, after enzymatic binding of Cy5-labeled ddATP to the selective deprotection site, epifluorescence imaging demonstrating that fluorescence pattern (Figure 12g right) is identical to the random pixel activation (Figure 12g left), confirming that the nucleotides are incorporated with the spatial selectively deprotected sites by enzyme and pH localized at the ceiling successfully. The highly parallelized and densified pH-regulated enzymatic DNA elongation may enable high-throughput synthesis of DNA enzymatically in a mild solution medium, applied to the storage and archiving of DNA data, as well as providing access to numerous molecular biology tools. [181][182][183][184]

Conclusion and Outlook
Recent advances in FET-based biosensor arrays for pH sensing, including static and active-matrix arrays, have been summarized in this review. pH sensor array has received considerable attention owing to its unique features and various applications. Nowadays, the developments of novel materials, promising device geometry, and semiconductor processing technology have led to virtual improvements in pH sensor arrays. In this review, recent advances in pH sensor arrays are introduced, including geometry design, working principle, advanced electrical performance, and progressive applications. The structure configurations of sensor arrays primarily consisting of static arrays and active-matrix arrays are concluded above. The recent achievements of pH sensor arrays with various mechanisms, structures, and fabrication processes are sorted out in this work. Differences between FET-based sensor arrays including static and activematrix array are demonstrated, where static arrays with simple fabrication approaches and lower cost always suffer from redundant driving components and leading wires, as well as low device density, while active-matrix array containing switch FET and sensor FET has accurate readout, low power consumption, less electrode wires, high device density, and improved spatiotemporal resolution, but more complex craft and expensive technology. In addition, FET-based pH sensor arrays can be widely employed in extensive domains due to the achievement of real-time pH mapping, including human healthcare management, highthroughput DNA sequencing, and elongation. Thus, based on the unique intrinsic characteristics and structure geometries, FETbased sensor arrays provide promising platforms to satisfy the requirements of human society.
Despite the significant progresses that have been made in FETbased pH sensor array in real-time mapping, excellent sensitivity, large-area, and high-density integration to CMOS technology, several factors in sensing characteristics are still required to be further considered for future advanced pH-sensing applications: 1) Devices suffer from potential drift along with time, making it challenging to acquire consistent potential, leading to instability, which will be influenced by the passivation layer, testing setup, reference electrode, sensing electrode material, and fabrication method. Such issues could be addressed by using EGFET configurations, which can separate the sensing layer from the transistor and protect the device from electrolyte corrosion. Deposition of a high-k dielectric layer such as Al 2 O 3 to the semiconductor layer can also improve long-term stability and signal drift. [35] In addition, results of alternative reference electrode materials based on Ti/Au/Ag/AgCl show further improvements. [185] 2) In addition to the challenge mentioned above, efficient sampling and interfaces between fluids and solid surfaces are vital in continuous health monitoring, ensuring high repeatability and preventing contamination. [186,187] Moreover, considering that analyte concentration in the real sample is not as reliable as that of buffer, a simple purification approach is commonly adopted, requiring the device to resist contamination. 3) Fabrication of transparent, flexible, high-density, less crosstalk, and the stable active-matrix array is challenging. Although ISFET array with high-integration is demonstrated by CMOS technology, the current CMOS fabricated process cannot be applied to transparent and flexible substrates due to the limitations of crafts. Static arrays with a more straightforward process and low cost have been illustrated to acquire the flexible device, but they suffer from low resolution. Considering that emerging IoT technologies require massive wearable sensor arrays, flexible and large-scale integrated device platforms with low-cost fabrication methods should be investigated. A new approach, such as a solution-processed method using metal oxide semiconductors, has been proposed, which makes it possible to achieve a flexible and transparent activematrix array for pH sensing. At the same time, crosstalk problems are minimized. Owing to the complexity of techniques and nonuniformity by spin coating, there are few available devices; thus, efforts need to be concentrated on addressing the abovementioned issues. 4) The resolution of the array should be improved further, including less response time, and higher temporal and spatial resolution. By combining a switch sensor and FET sensor for pH sensing and multiplexing in each cell, the active-matrix array can be fabricated to reduce the number of wires, leading to high spatial resolution. In addition, minimizing line width and semiconductor size to create pixels as compact as possible can also improve spatial resolution. However, reducing the sensing area may decrease device sensitivity and signal-noise ratio (SNR); so, there is a trade-off between them. In addition, researchers have started pushing the array's temporal resolution by creating a device with high frame rates, which is achieved by using high signal frequency. [188] 5) The ultimate goal of the sensor array is to realize a closeloop system for human healthcare management that can be utilized in the real-life world. By considering the integration of a miniaturized sensor array, power supply, communications, data management, and machine learning, a sensor system with multifunctional, fully portable, wireless data collection, intelligent skin plasters with continuous and longterm sensing capabilities will be achieved; thus, real and virtual space can interact in the future seamlessly. [189,190] Future research on a close-loop system of pH sensor array is expected to coordinate with advances in computation and signal processing.
With regards to FET-based biosensor arrays, despite achieving tremendous success, further improvements are needed to realize practical applications. More endeavors can be made in device stability and repeatability, pixel design, flexible active-matrix array, and a close-loop system, including power supply, data collection, machine learning, and intelligent skin plasters, for future advanced human healthcare management. This would revolutionize personalized medicine, biomedical research, as well as automation in manufacturing technologies. Given the challenges in the pH sensor array, we believe that more interdisciplinary collaborations are required to advance this emerging domain.