Printable metal oxide nanostructures based chemiresistive non-biological analyte sensors

Non-biological analyte sensing refers to the ability to detect and quantify various chemical and physical parameters present in the environment or biological samples that are not directly associated with biological entities such as cells, tissues, or organisms. The field of non-biological analyte sensing has its roots in the early detection of any analytes, and over the years, it has expanded to include a wide range of applications such as environmental monitoring, food safety, and medical diagnostics. This perspective focuses on the current status, challenges and future prospects of metal oxide nanostructures-based non-biological analyte sensors. In this context, the present review aims to delve into the intricate mechanisms, fabrication techniques, and applications of printable chemical sensors for non-biological analytes. Through a comprehensive exploration of the scientific advancements and technological breakthroughs in this domain, this review seeks to provide a comprehensive understanding of the evolving landscape of printable chemical sensors and their pivotal role in modern analytical endeavours.


Introduction
Printable chemical sensors have emerged as innovative tools in the field of analytical chemistry, offering versatile and adaptable means of detecting and quantifying non-biological analytes.Harnessing the principles of materials science and nanotechnology, printable chemical sensors are fabricated using advanced printable materials such as conductive polymers [1], metal oxides (MOs) [2], carbon nanotubes [3], graphene [4], and functionalized nanoparticles.These materials offer the advantage of being easily deposited onto various substrates, including flexible and low-cost surfaces, through techniques like inkjet printing, screen printing, and aerosol jet printing [5].Such approaches facilitate the creation of sensor arrays and customized designs, enabling tailored sensor platforms for specific non-biological analytes.Among all the material MOs based sensors are among the most widely used non-biological analyte sensing platforms.The realm of chemical sensing has been significantly augmented by the emergence of printable sensors based on metal oxides, presenting a sophisticated avenue for the detection and quantification of non-biological analytes.MOs sensors operate by detecting changes in the electrical conductivity of the material when exposed to specific target analytes.These sensors, rooted in the principles of materials science and nanotechnology, leverage the distinct properties of metal oxides to facilitate the transduction of chemical interactions into discernible signals, thereby enabling the sensitive and selective analysis of a diverse array of non-biological substances.These sensors have several advantages, including high sensitivity, fast response time, and low cost [6].The status of MOs-based sensors is constantly evolving, driven by ongoing research and development efforts.Recent advances in materials science, nanotechnology, and device engineering have led to significant improvements in the performance and functionality of metal oxides [7].MOs are a class of materials that have been widely used for printable chemical sensors for non-biological analyte sensing.These sensors have several important properties that make them attractive for this application such as high surface area, chemical stability, selectivity, sensitivity, reversibility, low cost and printability [8].These methods afford the deposition of finely tuned metal oxide nanoparticulate formulations onto diverse substrates, including flexible and cost-effective surfaces.The compositions and structures of oxides used in chemical sensors for non-biological analytes can vary depending on the specific sensing application and target analyte.The deployment of printable metal oxide-based chemical sensors for target analytes is found across an expansive spectrum of applications.Environmental monitoring, industrial process control, and gas-phase detection are domains where these sensors exhibit profound utility.The amalgamation of printable sensors with electronic readout systems facilitates seamless integration into data acquisition networks, thereby enabling real-time and remote monitoring capabilities.
1.1.Effects on the human body upon exposure to non-biological analytes Exposure to various non-biological analytes can have different effects on the human body depending on the specific nature of the analyte, its concentration, duration of exposure, and individual susceptibility.Here are some general categories of non-biological analytes and their potential effects on the human body: (a) Toxic Gases (e.g., carbon monoxide, hydrogen sulfide) [9]: • Carbon Monoxide: Can lead to headaches, dizziness, nausea, confusion, and, at high levels, can result in unconsciousness or death due to its ability to bind to hemoglobin, reducing oxygen transport in the blood.
• Hydrogen Sulfide: Inhalation of high concentrations can cause respiratory irritation, headache, dizziness, nausea, and even unconsciousness or death.It is particularly hazardous due to its rapid onset of effects.
• Formaldehyde: Can cause eye, nose, and throat irritation; long-term exposure may contribute to respiratory issues and certain types of cancer.
(c) Particulate Matter (PM) and Dust [11]: • Inhalation of fine particulate matter can exacerbate respiratory conditions such as asthma and bronchitis.
Long-term exposure is linked to cardiovascular diseases and lung function decline.
• Mercury: Depending on the form (organic or inorganic), exposure can damage the brain, kidneys, and nervous system.It is particularly harmful to developing fetuses.
(e) Industrial Chemicals (e.g., ammonia, chlorine) [13]: • Ammonia: Inhalation of high concentrations can cause irritation of the respiratory tract and eyes, leading to coughing, wheezing, and shortness of breath.
• Chlorine: Exposure to chlorine gas can lead to severe respiratory irritation, chest pain, and even pulmonary edema.
Immediate symptoms might include irritation, headaches, nausea, and dizziness, while long-term exposure can lead to chronic health issues and exacerbation of pre-existing conditions.To mitigate risks, proper ventilation, personal protective equipment, and adherence to safety guidelines are crucial when dealing with non-biological analytes in various environments.

Criteria for evaluating MOs non-biological analytes sensor
A collection of parameters is utilized to depict the sensor's performance as shown in figure 1.The most crucial parameters are detailed below, accompanied by their respective definitions.

General sensing mechanism for MOs non-biological analytes sensor
The fundamental basis of metal-oxide-based sensors is predominantly rooted in chemiresistance.This concept involves scrutinizing the alteration in electrical conductivity or resistivity of the material when exposed to an analyte (heavy metals ions, VOCs and toxic gases), contingent upon its concentration.The interaction between metal oxides and oxidizing or reducing gas molecules or ions brings about a decrease or increase in the resistance value of the metal oxide.Consequently, this mechanism effectively enables gas sensing capabilities.The detection mechanism in MOs gas sensors is associated with the ion sorption of substances onto their surfaces.When the gas sensor comes into contact with oxygen, the absorbed oxygen species are formed, causing the oxygen atoms to extract electrons from within the metal oxide.The subsequent steps (reactions 1-4) depict the kinetics of adsorption [14].
When the MOs are exposed to the target analyte, the analyte interacts with the surface of the oxide, causing a change in the electrical properties of the material.This change in electrical properties is typically caused by the adsorption or chemisorption of the analyte on the surface of the MOs, leading to changes in carrier concentration, mobility, or conductivity.For example, some MOs sensors operate based on the chemical reaction between the analyte and surface, while others operate based on the physical adsorption or desorption of the analyte on the MOs surface.The sensing mechanism can also be influenced by factors such as temperature, humidity, and pressure, which can affect the adsorption and desorption kinetics of the analyte on the MOs surface.In light of these advancements, this review endeavors to comprehensively explore the intricacies of printable chemical sensors rooted in metal oxides for the detection of non-biological analytes.By delving into the nuanced mechanisms, fabrication methodologies, and application prospects, this review seeks to furnish a holistic understanding of the burgeoning landscape within this specialized domain, underpinning the critical role of printable metal oxide-based chemical sensors in modern analytical pursuits.
The conductivity of the sensing element is altered by the interaction between adsorbed oxygen and the reducing or oxidizing target analytes.In the presence of a reducing analyte, the adsorbed oxygen species on the sensing element detach and donate the previously trapped electron to the conduction band [15].Consequently, the heightened concentration of electrons in the conduction band leads to a reduction in the resistance of the sensing element.Thus, when detecting reducing target analytes, n-type materials are favored, while p-type materials are chosen for the detection of oxidizing analytes.The selection of appropriate materials in conjunction with the target analyte requires careful consideration to enhance the sensor's performance characteristics.The changes in resistance of the sensing elements in response to reducing and oxidizing target analytes are summarized in table 1 [14].In figure 2(a), the schematic representations of different categories of sensing mechanisms are depicted.In figure 2(b), the chemiresistive sensor is illustrated, comprising four primary components: an active material (sensing layer), contacts (Au/Cr), a dielectric layer for insulating the contacts (passivation layer), and a substrate.Figure 2(c) portrays the electrical circuit configuration of the chemiresistive sensor.
However, zinc oxide (ZnO) [16], tin oxide (SnO 2 ) [17], titanium dioxide (TiO 2 ) [18], tungsten oxide (WO 3 ) [19], and molybdenum oxide (MoO 3 ) [20] are commonly used for this purpose.The device architectures can be optimized by modifying the MO's material properties and device geometry, such as by changing the particle size, shape, and orientation, to improve the sensitivity, selectivity, and response time of the sensor.Hemoglobin trapped in hollow zirconium dioxide (ZrO 2 ) spheres and sodium alginate sheets electrodeposited on a gold electrode were used to create an H 2 O 2 sensor [21].This electrode produced a Hb Fe 3+/ Fe 2+ quasi-reversible CV response.Methyl parathion, an organophosphate insecticide, was detected by using the high attraction between the zirconia and phosphate groups [22].Maier have developed a flexible sensor for the detection of H 2 O 2 in simulated breath, which has potential applications in health monitoring [23].Liu Yang et al contrasted chemiresistive gas sensors with conventional methods for analyzing Chemical Warfare Agents (CWAs).These sensors offer notable advantages such as compact size, rapid response, and affordability in comparison to traditional instruments.Given the progress in nanofabrication technology and the emergence of novel sensing materials like carbon nanotubes, graphene, and black phosphorus (BP), the exploration of chemiresistive gas sensors remains a significant avenue of research [24].Jae Eun Lee et al introduced a chemiresistive sensor based on a ZnO-CuO p-n heterojunction.This sensor configuration comprises CuO hollow nanocubes attached to ZnO spherical cores as the active materials.Notably, these core-hollow cube nanostructures made of ZnO-CuO display a remarkable response of 11.14 at 1 ppm acetone and 200 °C, surpassing the performance of other metaloxide-based sensors reported in the literature.The response is measurable down to 40 ppb, and the limit of detection is estimated to be 9 ppb.Additionally, the ZnO-CuO core-hollow cube nanostructures exhibit strong selectivity toward acetone over other VOCs and maintain excellent stability for a period of up to 40 days [25].
Printable metal oxide nanostructures based chemiresistive non-biological analyte sensors can be designed using various device architectures, depending on the specific sensing application and target analyte.Here are some common device architectures used in such sensors as a thin film, nanoparticle, interdigitated electrode and flexible.This architecture involves printing or depositing sensing materials on flexible substrates, such as polymer films or fabrics, to create wearable or flexible sensors for non-invasive sensing applications.In  figure 3(a), the schematic representation of printable chemical sensors designed for non-biological analytes on metal oxide is depicted.In figure 3(b), the alteration in resistance is presented when the target analyte is present, while figure 3(c) illustrates the modification in current conduction in the presence and absence of analytes.The utilization of printable chemical sensors spans diverse applications, from environmental monitoring to industrial process control.By interfacing with non-biological analytes such as volatile organic compounds, heavy metals, and gas-phase molecules, these sensors contribute to enhanced detection sensitivity, reduced response times, and improved portability.Furthermore, the compatibility of printable sensors with electronic readout systems ensures seamless integration into data acquisition networks, enabling real-time monitoring and analysis.Zhennan Zhu et al have devised functional ink systems based on metal oxides tailored for precision drop-on-demand printing of thin-film transistors.Their work encompasses an extensive review of distinct ink systems utilized in the context of thin-film transistor printing.Simultaneously, the study provides a comprehensive analysis of challenges encountered in the domain of printed thin-film transistors, accompanied by elucidation of resolutions originating from the ink system perspective [9].Table 2 displays some reports on the sensor performances for non-biological analytes based on MOs nanostructures.

Current and future challenges
Despite the potential advantages of printable MOs nanostructures based chemiresistive non-biological analyte sensors, several current challenges need to be addressed to improve their performance and practicality [37].Figure 4 shows the development of effective sensing characteristics along with the accompanying challenges.Some of these challenges include selectivity, sensitivity, long-term stability, reproducibility, baseline recovery, humidity effect, high temperature, cost and scalability [38].One of the main challenges in the development of sensors is achieving high selectivity towards the target analyte while avoiding interference from other chemical species present in the sample matrix.Metal-oxide sensors have a sensitivity range of parts-per-million (ppm) or even parts-per-billion (ppb), which allows them to detect non-biological analytes at low concentrations.The operating temperature, which must be precisely managed to provide optimal performance, has a significant impact on their sensitivity.In general, higher working temperatures increase sensitivity, but they can also result in decreased selectivity and shortened sensor lifetime.Additionally, metal oxide sensors can exhibit crosssensitivity to other analytes, which can lead to false-positive or false-negative results.With metal-oxide sensors, reproducibility is also a serious problem since changes in fabrication parameters or ambient factors can significantly alter sensor performance.To solve this, rigorous quality assurance processes should be used during the sensor production process.Additionally, calibration techniques should be used to take into account changes in sensor performance over time.This requires optimizing the MOs morphology, porosity, and size, as well as the operating conditions such as temperature and humidity.Additionally, there are challenges in scaling up the production of these sensors to meet industrial demand.Therefore, it is essential to explore cost-effective and scalable fabrication methods, such as inkjet printing or roll-to-roll processing.SnO 2 is a metal oxide that is frequently used in non-biological sensors, although it is susceptible to interference from humidity and other analytes, which can lower its selectivity towards particular target analytes.On the other hand, it has been demonstrated that zinc ZnO [47] sensors have a higher sensitivity to some analytes, although they can be less repeatable due to fluctuations in the growth conditions of the thin films used in their production [48].While the selectivity and sensitivity of other metal oxides, such as tungsten oxide (WO 3 ) [49] and titanium dioxide (TiO 2 ) [50], have shown encouraging results, other variables, including the crystal structure, doping concentration, morphology and operation temperature, can also affect how well these materials work.There is a growing demand for sensors that can be used in portable or handheld devices for onsite and in-field applications.Therefore, future challenges in this area will involve developing sensors that are small, lightweight, and low power-consuming, while maintaining high sensitivity and selectivity.Based on recent research, the synthesized cabbage-like ZnO nanostructure electrode shows potential as an auspicious candidate for use in innovative chemical sensors designed to detect harmful chemicals, such as resorcinol.These findings suggest that the ZnO nanostructures electrode could serve as an efficient and effective tool for detecting resorcinol in various applications, highlighting its potential as a valuable addition to the field of chemical sensing.Although the sensor's detection range (5 μM-320 μM) exceeds the H 2 O 2 concentration in human breath (0.1 μM-1.5 μM in the case of exhaled breath condensate), it serves as proof of concept for a commercial sensor [51].Overall, addressing these current and future challenges will require continued research and development in materials science, chemistry, physics, and engineering, as well as collaboration across different fields and industries.By overcoming these challenges, printable metal oxide nanostructures-based chemiresistive non-biological analyte sensors have the potential to revolutionize sensing and monitoring in a wide range of applications, from healthcare to environmental monitoring to industrial process control.Detecting non-biological analytes presents numerous challenges due to their low concentrations, differential appearance in cells, multi-step sanitization procedures, and the use of expensive reagents.Consequently, there is a growing interest in the research and development of novel sensing materials that surpass the capabilities of traditional sensing materials, alongside investigations into the underlying sensing mechanisms.Metal-oxides non-biological analytes sensors are often characterized by limited selectivity towards different analytes, whereas emerging two-dimensional (2D) transition metal dichalcogenides materials encounter challenges such as reduced sensitivity, difficulties in large-scale production, and poor stability [52].

Morphology dependent selectivity
One of the main challenges in the development of sensors is achieving high selectivity towards the target analyte while avoiding interference from other chemical species present in the sample matrix.Selectivity stands as one of the utmost critical performance metrics in the realm of trace target analyte sensing.Selectivity, or the ability of a gas sensor to specifically respond to a target gas while ignoring interference from other gases, is a critical aspect of gas sensing technology.
Diverse synthesis and fabrication methods have been employed to create gas sensor materials with distinctive morphologies, potentially impacting their selectivity.A compilation of these investigations aims to assess how synthesis and morphology exert an influence on sensor selectivity.While the subsequent section will primarily concentrate on p-type metal oxides due to their frequent application for enhanced chemical sensitization, it's important to note that the discussion of pristine n-type materials will also consider how morphology can impact their response.This focus on p-type oxides arises from their comparatively limited  exploration in the existing literature.In figures 5(a)-(c) we can see that how morphology changes as its gas changes.While the characterization of exposed facets was lacking in numerous investigations, it is evident that the growth mechanisms frequently mentioned might provide insights into how the crystallographic facets of Co 3 O 4 contribute to the alteration of selectivity between ethanol and acetone.

Temperature dependent sensitivity
This requires careful optimization of the material's composition, structure, and surface properties, as well as the design of appropriate sensing.The sensitivity of metal oxide-based sensors is intricately linked to the sensing temperature due to the dynamic changes that occur in the material's surface and electronic properties within varying temperature ranges.Sensing temperature significantly influences the kinetics of chemical reactions occurring at the surface of metal oxide sensors, subsequently affecting their interaction with target analytes.At elevated temperatures, the chemical reactions between the metal oxide surface and the target analyte molecules tend to be more favorable and efficient.This often leads to enhanced adsorption, desorption, and catalytic processes, resulting in increased sensitivity and response towards the analytes.Higher temperatures can also facilitate the release of adsorbed molecules from the surface, making the sensor more responsive to subsequent analyte exposures.
Jun Min Suh et al delved into the impact of Pd and Au decoration on the gas-sensing characteristics of MoS 2 thin films [42].The introduction of Pd and Au decoration proved effective in reducing the operational temperature required to attain the utmost gas response to NO 2 .This was achieved by electronically sensitizing the MoS 2 material.Additionally, the incorporation of Pd decoration led to a substantial enhancement in selectivity towards H 2 , accomplished by the formation of Pd hydride as shown in the figure 6(b).Furthermore, Au decoration yielded an improvement in the overall gas responses across all tested gas species as shown in the figure 6(c).Notably, alterations in conduction types (n-type or p-type) were observed contingent on the specific gas species and operational temperatures.This study elucidates the discernible impacts of noble metal catalyst decoration on the gas sensing properties of MoS 2 , offering valuable insights into the field [42].In the recent advancement in the material science the sensor shows dual sensing behaviour.Sukhwinder Singh et al investigated the gas-sensing performance of the MoS 2 /TiO 2 hybrid that was prepared, measuring it across varying temperatures as shown in the figures 7(a), (b) [43].Interestingly, the device exhibited distinct detection of methanol and ethanol at different temperatures, showcasing a selective response.This selectivity phenomenon is possibly attributed to the alignment between the LUMO (least unoccupied molecular orbital) energy of the adsorbed analyte and the adsorbed oxygen on the sensor's surface.This preference could arise from the matching electronic energy levels between the LUMO energy of these two analytes and the adsorbed oxygen on the device's surface.The LUMO energies for methanol and ethanol are 0.197 and 0.125 eV, respectively.Hence, the interaction of adsorbed gaseous species with the sensor surface, involving electron transfer, is significantly influenced by the LUMO energies of the respective gas molecules.This interaction depends on factors like the operational temperature, as well as the inherent electronic characteristics of the sensing channel.For instance, elemental semiconductors exhibit a decrease in band gap with increasing temperature, which augments their conductivity.In this context, a higher operating temperature can enhance the charge transfer between adsorbed gaseous species and the sensing surface, potentially leading to selectivity due to the varying LUMO energies of different gas molecules.Consequently, by capitalizing on temperature-induced high selectivity and a substantial response value of MoS 2 /TiO 2 against a spectrum of seven distinct gas analytes, practical applications in the medical domain could be unlocked [43].

Reproducibility
Reproducibility of a sensor refers to its ability to consistently provide similar measurements or responses when exposed to the same conditions or analytes across multiple trials.A highly reproducible sensor yields results that exhibit minimal variation between repeated measurements, indicating its reliability and accuracy.Kyungjae Lee et al worked on the precision of reproducibility that was validated by employing multiple devices (depicted in figure 8(a)) and conducting multiple experimental runs (illustrated in figure 8(b)) [44].Notably, the data obtained from these repeated runs also shed light on the regeneration capabilities of the receptor surface.The regenerative process required careful treatment, where purging with N 2 was exclusively used to eliminate layers of DNT (dinitrotoluene) gas molecules while preserving the immobilized peptide (aptamer) intact.For the experimentation involving multiple devices (figure 8(a)), a DNT-bp (DNT-binding peptide) rGO sensor was prepared.The process entailed an initial stabilization of resistance change (5 min), followed by the introduction of 320 ppb DNT vapor (5 min), succeeded by a N 2 gas purge (10 min), iterated five times across multiple rGO sensor chips (N = 5).The calculated average resistance change for these five rGO sensors was approximately 98.8 × 10 −4 , accompanied by a standard deviation of about 0.13%.Furthermore, the coefficient of variation (C.V.) was determined as 13.8% from the data obtained across multiple rGO sensor chips, indicating a commendable level of reproducibility.Similarly, in the case of multiple experimental runs (figure 8(b)), reproducibility was assessed using the same rGO sensor and following the protocol outlined in figure 6(a).The calculated average resistance change resulting from the interaction between DNT-bp and DNT gas was approximately 99.6 × 10 −4 , exhibiting a standard deviation of around 2.3 × 10 −4 .The C.V. was computed at approximately 2.30% from multiple rGO sensor chips, indicating a high level of reproducibility [44].

Humidity effect
At higher humidity conditions, the baseline of a MOs sensor often does not remain stable due to the influence of water vapor on the sensor's surface and electrical properties as we can see in the figure 9 where the sensor baseline not reaching to its initial value.The gradual elevation of the base resistance (Ra) in both pure and Ptanchored CuCrO 2 chemiresistors is a consequence of the retrieval of primary hole carriers (generated thermally).This phenomenon arises through the subsequent reaction [46]: Given that •OH radicals exhibit potent oxidizing characteristics, the reaction kinetics of chemiresistors towards reducing H 2 S experience significant augmentation in the presence of a humid air atmosphere, as evidenced by figure 9.This phenomenon can be attributed to several factors such as the adsorption of water molecules, surface ionization, chemical reactions.However, it's important to recognize that in high-humidity environments, the behavior of MOs sensors can become complex due to the interaction between water vapor and the sensor surface, leading to deviations from a steady baseline.
To overcome these limitations, a promising research direction could be the integration of traditional materials such as 2D Black Phosphorus [53], Ti 3 C 2 MXene-Polymer [54], 2D MOF nanosheets, and carbonbased materials with 2D materials to enable more effective non-biological analytes detection.

Advances in science and technology to meet challenges
The developed non-biological analytes chemical sensors are currently in the prototype stage and have solely been evaluated in laboratory settings.The number of actual samples used for evaluation has been minimal, rendering validation inadequate.The development of marketable products is contingent upon addressing challenges about optimizing the stability, storage, and logistics of the printable chemical sensors for non-biological analytes, as well as ensuring their optimal performance in minimally treated or diluted diverse samples.Recent developments in materials science have led to the development of a new class of materials, termed 'nanocomposite materials', which are composed of organic semiconductors and biomass derivatives.These hybrid materials exhibit superior unique properties such as high electrical conductivity, robust mechanical strength, and excellent thermal stability [55].Chemical sensing devices face significant challenges regarding selectivity and sensitivity.Therefore, there is a pressing need to integrate interdisciplinary fields to create smarter and high-performance chemical sensing devices.The emergence of nanotechnology, which involves the manipulation and control of materials at the atomic and molecular level (typically in the nanometer range), has led to exciting developments in non-biological analytes design.The synthesis of materials at the nanoscale has enabled the exploitation of exceptional physical, chemical, and biological properties, which are central to the success of nanotechnology.
Minggang Zhao et al have constructed distinct hierarchical structures, namely Cu/CuO/rGO/PPy, Cu/ CuO/rGO/CS, and Ni/NiO/ZnO/PPy, successfully incorporating various interfacial energy barriers, such as Schottky or p-n junctions, within electrochemical sensors as shown in the figures 10(d), (e) [62].These structures enable the modification of energy barrier heights through electrostatic interactions between charged analytes and the interfacial junctions.This adjustment corresponds to the specific charged analytes, ultimately amplifying the electrochemical response of the sensors in an exponential manner.Unlike conventional redoxbased electrochemical sensors, the novel sensing mechanism grounded in interfacial energy barriers possesses a distinctive feature-resistance to interference in intricate scenarios, including environments like seawater.Engineered absorbing layers, such as PPy (polypyrrole) for Hg 2+ and CS for Cu 2+ , are specifically designed to accumulate analytes and their charges.These charges' electrostatic induction can impact the interfacial energy barrier without necessitating redox reactions.This newly established sensing mode boasts exceptional antiinterference properties and ultra-high sensitivity as shown in figures 11(c), and (d), thus presenting novel avenues for detecting trace analytes within challenging settings like seawater [62].
In their study, Ana Zubiarrain et al have shown that graphene-like carbon (GLC) can serve as the basis for constructing chemiresistive sensors capable of detecting free chlorine in water across a wide concentration range as shown in the figures 12(a)-(d) [63].They examined the sensor responses within two distinct concentration ranges: 0.01-0.2ppm and 0.2-1.4ppm as shown in the figure 12(e).The thickness of the GLC films had a notable impact on the maximum sensor response and saturation concentrations, with thinner films proving more sensitive but offering a narrower detection range.The thinnest GLC films (12 nm) achieved an impressive limit of detection, as low as 1 ppb.Detecting concentrations exceeding 1.4 ppm seemed feasible with GLC films thicker than 46 nm.These sensors didn't require external resetting and exhibited a reduction in response when exposed to lower free chlorine concentrations, making them suitable for monitoring concentration variations.Although bare devices showed some response to free chlorine, functionalizing them with phenyl-capped aniline tetramer (PCAT) substantially enhanced their performance in quantifying free chlorine in actual tap water samples.In a specific case, the blank devices estimated a water sample containing 0.4 ppm chlorine to be 0.6 ± 0.1 ppm, while the doped sensors estimated it as 0.4 ± 0.1 ppm as shown in figures 12(g), (h).
Furthermore, the sensors exhibited negligible responses to common cations and anions found in drinking water, with signal changes being less than 5% in response to 0.5 ppm of free chlorine.The versatility of GLCbased chemiresistive sensors is exemplified by their ability to adapt sensitivity and dynamic range through varying GLC film thickness.This flexibility opens up various applications, including real-time water quality monitoring, chlorine filter breach detection, and quantifying free chlorine in unknown samples.Their ultra-low power consumption (less than 100 nW during reading) makes them ideal for remote monitoring applications.Moreover, the design of these chemiresistive sensors isn't confined to free chlorine alone; the nanocarbon films can be tailored for other specific chemical targets.Additionally, the study demonstrated the use of two different sensor geometries, expanding deployment possibilities.While the manual, lab-based fabrication method used in this study exhibited some variation in response magnitude between devices due to narrow tolerances, the film format of GLC holds the potential for cost-effective mass production of sensors in the future [63].
Pil Gyu Choi et al achieved the synthesis of Co 3 O 4 nanoparticles through a hydrothermal process, with the growth of these nanoparticles being controlled by crystal formation [64]. Figure 13(a), shows the schematic depiction of an irregularly shaped structure, alongside the alteration in resistance when an analyte is present.Moving on to figure 13(b), presents a schematic representation of a structure under crystal growth control, which displays enhanced sensing capabilities following the introduction of crystal defects, accompanied by a change in resistance when exposed to an analyte.During this process, three distinct raspberry-type structures, a cubic shape, and an irregular form emerged within the Co 3 O 4 nanoparticles.It was observed that the raspberrytype Co 3 O 4 structures exhibited a smaller crystallite size compared to the cubic and irregular variants.This divergence in size could be attributed to the raspberry-type structures possessing a larger lattice volume, a phenomenon resulting from lattice expansion due to the presence of crystal defects.Notably, the presence of these crystal defects yielded an enhancement in gas sensing attributes, particularly towards reducing gases like H 2 , C 5 H 8 , and CH 3 COCH 3 , in contrast to the acidifying gas NO 2 as shown in figure 14(f).Additionally, the hierarchy of gas sensor signal responses shifted as crystal defect areas increased on the Co 3 O 4 surface.The affinity of a gas molecule's higher dipole moment exerted a more potent influence of crystal defects on gas sensing properties.Furthermore, these crystal defects displayed sensitivity to humidity, where water molecules at crystal defect sites and adsorbed H 2 O interacted with target gas molecules, enhancing polar molecule adsorption as shown in figures 14(a)-(e).The Co 3 O 4 nanoparticles, controlled through crystal growth, exhibited comparable durability to the irregularly shaped variants.These findings hold the potential to contribute significantly to comprehending the gas-sensing properties of Co 3 O 4 , particularly those tied to surface states.Furthermore, these insights hold promise for applications spanning human healthcare and environmental protection.These applications include selective sensing of specific target gas molecules such as NO 2 for environmental acidification, CH 3 COCH 3 for diabetic ketoacidosis diagnosis, C 5 H 8 for elevated blood cholesterol, and H 2 for colonic flora [64].
Ruofan Zhang et al have developed a solution combustion synthesis (SCS) approach to create porous Ptanchored CuCrO 2 (CCO) nanoparticles in a single step [45].This process enables atomic-level mixing of reactants and Pt ions, facilitating the even dispersion of Pt atoms within the CCO matrix.Employing an optimized Pt ratio of 1.39 wt %, Pt-anchored CCO nanoparticles significantly enhanced response characteristics towards H 2 S.This enhancement includes a remarkable 1250-fold increase in response from 10 ppm H 2 S at a low operating temperature of 100 °C 35 times higher than pure CCO as we can see in figure 15(a).Additionally, the Pt-anchored CCO exhibited excellent selectivity and reproducibility shown in figure 15(b) [45].
We can also utilize nanoscale synthesis techniques in the development of sensing materials, including methods like electrospinning and hydrothermal synthesis [65].These approaches can yield a variety of innovative one-dimensional (1D) nanostructured sensing materials based on MOs [66].Numerous studies have demonstrated the progression of non-biological analyte detection.These studies have explored various sensor types based on receptors and transducers, along with contemporary techniques utilizing nanomaterials like noble metal nanoparticles (NPs), nanocrystalline [67], MO NPs, nanowires (NWs) [68], nanorods (NRs),  carbon nanotubes (CNTs), quantum dots (QDs), and dendrimers.These advancements underscore the integration of nanotechnology into sensing technology.1D nanomaterial-based sensors, commonly referred to as 1D nanosensors, offer several advantages compared to conventional bulk sensing materials.These advantages encompass enhanced sensitivity to target analytes, swifter response times, and reduced weight and power consumption.These benefits are chiefly attributed to the substantial augmentation of the surface area to volume ratio, heightened surface activity, and augmented charge carrier mobility achievable at the nanoscale.In the case of 1D nanomaterials featuring a limited density of defects, there tends to be a relatively minimal presence of crystalline boundaries that typically scatter charge carriers.Moreover, in the context of 1D materials boasting sufficiently diminutive diameters, charge carriers become constrained radially while retaining freedom of movement axially [69].This dual characteristic synergistically fosters efficient charge carrier transportation.Furthermore, the potential for direct growth of 1D nanomaterials onto established microelectrodes brings forth the opportunity for seamless integration.This integration not only maintains low contact resistance between the sensing material and the electrode but also facilitates the scalability of nanosensors [70].

Advances in Technology
Achieving enhanced functionality and portability of sensors and their associated electronics is a significant challenge that requires miniaturization.Recent progress in nanoelectronics holds promise in addressing this challenge by enabling the development of smaller and more efficient sensors and electronics [71].To meet the difficulties of modern healthcare systems, significant research has been dedicated to developing improved portable devices that utilize microelectromechanical systems (MEMS) and nanoelectromechanical systems (NEMS) for sensing and separating critical non-biological analytes.To achieve this goal, AI (artificial intelligence), advanced ML (machine learning) techniques and IoT (internet of Things) integration can be employed to develop highly selective and sensitive smart sensors.
ML techniques are increasingly being integrated into sensor development to enhance performance, accuracy, and efficiency.ML can help in various aspects of sensor operation such as data analysis and pattern recognition, sensor calibration, feature extraction, dynamic response prediction, fault detection and quality control, sensor array fusion, optimal material design, and adaptive sensing [73].Incorporating ML in sensor development offers the potential to create smarter, more adaptable sensors with improved detection capabilities, reduced false alarms, and optimized performance across various applications.Using supervised learning algorithms like support vector machines (SVM), random forests, or neural networks to study the behaviour of the sensor responses [74].Designing ML models that can provide real-time analysis of sensor data, enabling rapid detection and response to changes in analyte concentrations.In the present day, various researchers working on portable and self-powered systems utilizing piezoelectric and triboelectric nanogenerators exhibit the potential to partially address the aforementioned limitations.This approach can offer effective data processing methods that enable the development of advanced sensing devices.By merging wearable sensors with AI data analytics, it becomes possible to capture signals associated with muscle deformation, joint flexion, temperature fluctuations, heartbeat frequency, and more.These data hold immense significance and find extensive use in domains like healthcare, environmental monitoring, human-machine interactions (HMI), and plant monitoring applications, as illustrated in figure 16 [72].
An array of sensors was affixed onto a printed circuit board and situated inside a sensing chamber with a capacity of 200 mL, illustrated in figure 17(a).Individual sensors were associated with alumina tubes encompassing heaters, which were soldered onto a Bakelite base.This sensor configuration is schematically depicted in figure 17(b).To construct a voltage divider circuit, the sensors were interlinked in series with predefined resistances, as illustrated in figure 17(c).On a personal computer, ML algorithms (figures 18(d)-(f)) were executed.Jin-Young Kim et al examined and resolved the issue of cross-sensitivity in MOs sensors by constructing a sensor array and then employing neural network-assisted pattern recognition analysis ML methods to analyze the data obtained from the array of sensors.To detect hazardous gases such as CO and NO 2 , both individually and as mixtures, a compact array of metal oxide-based sensors was assembled.This array, featuring sensors constructed from Au-ZnO, Au-SnO 2 , Pt-SnO 2 , and unadorned In 2 O 3 , displayed the ability to selectively identify the specified gases even within complex environments, including various humid atmospheres.Utilizing this sensor array, it was possible to distinguish between binary mixtures containing different concentrations of CO and NO 2 .This differentiation was achieved across 25 distinct CO-NO 2 concentration combinations in dry air and 13 such combinations in two atmospheres with differing levels of relative humidity.The array's capacity for effective discrimination was demonstrated using Principal Component Analysis (PCA), which employed response values for various gases and their amalgamations.Furthermore, a novel approach was devised to capture comprehensive response and recovery patterns as inputs.This strategy mitigates information loss arising from these patterns, ultimately enhancing the accuracy of gas detection processes.[75].The precision achieved during the training phase is depicted in figure 18(a).After subjecting the adapted model to 500 iterations, employing the response values as inputs, it demonstrated a discrimination accuracy of 98.6%, as evidenced in figure 18(b).Notably, due to the significant reduction in data dimensions facilitated by PCA, the model selected for training with 500 epochs exhibited robust performance, attaining a flawless discrimination accuracy of 100%, as illustrated in figure 18(c).This robust performance can be attributed to the sensor array's capacity to generate high-quality fingerprints and the streamlined temporal information dimensionality.Consequently, the model achieved a perfect accuracy of 100%, as demonstrated in figure 18(d).Moreover, the researchers concluded that employing PCA on merely 25 and 10 data points and subsequently training these data with MLP-based models can yield accuracy exceeding 99.9%.This assertion is supported by the findings depicted in figures 18(e) and (f) respectively.
Radislav A Potyrailo et al that conventional semiconductor metal oxide materials have the potential to offer high-performance sensors using an impedance measurement method [76].Their strategy involves utilizing dielectric excitation measurements and results in sensors characterized by a linear gas response (with R2 exceeding 0.99), a wide dynamic range for gas detection (encompassing six decades of concentrations), strong baseline stability, and diminished susceptibility to humidity and ambient temperature variations [76].Impedance measurements of this kind can be applied for baseline correction, addressing a significant challenge associated with MOS-based sensors as shown in figures 19(a)-(e) at particular frequency its based line impedance almost remain constant.Considering this methodology is adaptable for implementation with both n-type and p-type MOs materials.Furthermore, this type of demonstration can be used in contexts such as wireless sensor networks, drone-based systems, and wearable devices for monitoring environmental and industrial gas levels.
Arnab Maity et al utilized a blend of wet transfer, impedance and noise assessments, along with ML techniques, to enable the large-scale production of sensor arrays based on graphene field-effect transistors (GFETs) [77].Figures 20(a)-(l) illustrate the schematic experimental configuration for continuous sensing, along with the corresponding outcomes.This strategy effectively identified faulty devices.The sensors they developed demonstrated the capability to concurrently detect heavy-metal ions (lead and mercury) as well as   methods.They established a sensing platform exhibiting a wide dynamic linear range spanning from 0.3 to 80 μM, along with low detection limits (LODs) of 84 nM, 88 nM, and 130 nM for Pb 2+ , Cd 2+ , and Hg 2+ ions, respectively.Correspondingly, the limit of quantification (LOQ) values was determined to be 280 nM, 320 nM, and 450 nM for the respective ions [78].
Min-Ho Seo et al pioneered a self-powered hydrogen (H 2 ) gas sensor employing a photovoltaic cell and a chemo-mechanically deformed Pd-PUA nanograting structure as shown in figure 21(a)-(c) [79].This innovative sensor mechanism capitalizes on alterations in the optical transmittance of the nanograting structure, triggered by the expansion of Pd subsequent to its reaction with H 2 gas.The ensuing change in current output from the photovoltaic cell facilitates detection.Operating solely on ambient light, the sensor negates the need for external electrical power.This sensor exhibited remarkable performance, detecting H 2 gas concentrations as low

Concluding remarks
This review highlights an increasing numeral of publications focused on printed chemiresistive sensors, with a subsequent shift towards investigating fully digital-printed devices.Most non-biological analytes devices used in medical applications suffer large sample volumes, bringing about unreliable outcomes.Although there are a few exceptions such as electrochemical H 2 O 2 sensors and lateral flow pregnancy tests, only a limited number of nonbiological analytes have achieved widespread commercial success.Furthermore, there is a demand for affordable nanostructure-based analytes that can provide quick, accurate, and user-friendly results.To accomplish this, nanomaterials should be integrated into a tiny biochip (lab-on-chip) for efficient sample handling and multiplexed clinical diagnosis.Ongoing academic research in this field is expected to produce commercially viable prototypes soon.Notably, the research presented in wireless environmental and health monitoring has the potential to enrich daily life in the era of IoT, AI and ML.In the field of sensing researchers can more focus on portable and self-power sensors utilizing piezoelectric and triboelectric nanogenerators exhibit the potential to partially address the aforementioned limitations.However, the implementation of these technologies is complex, expensive, power-intensive, and requires a significant hardware setup to be deployed in real-world applications.Hence, it is recommended that novel materials be developed, and efficient data processing techniques utilizing ML algorithms be implemented to address issues related to selectivity and long-term stability in chemical non-biological analytes sensors.In the future, digital printing technology has the potential to enable the complete manufacturing of chemical systems, specifically chemiresistive sensor systems capable of simultaneously detecting various environmental changes.This approach would offer considerable savings in terms of materials, time, and costs when compared to conventional manufacturing and integration methods.Enhanced comprehension and swift advancements in nanotechnology hold the potential to address the impending challenges that modern society will encounter.

Figure 1 .
Figure 1.Criteria for evaluating MOs sensor performance.

Figure 2 .
Figure 2. (a) Schematics of the sensing-mechanism categories (b) chemiresistive sensor with 4 main components: active material (sensing layer), contacts (Au/Cr), dielectric to insulate the contacts (passivation layer) and substrate (c) an electrical circuit of the chemiresistive sensor.

Figure 3 .
Figure 3. (a) Schematic diagram of printable chemical sensors for non-biological analytes on metal oxide.(b) change in resistance in the presence of target analyte (c) change in the current conduction in the presence of analytes and without analytes.

Figure 4 .
Figure 4. Development of effective sensing characteristics along with the accompanying challenges.

Figure 8 .
Figure 8. DNT-bp sensor reproducibility and regeneration test for (a) multiple devices and (b) multiple runs.The resistance change measured during multiple runs is reproducible (shown here for 320 ppb DNT), and the trace in (b) also confirms the complete regeneration of the sensor surface between successive runs at room temperature by using N 2 gas [44] Copyright 2019, Nature.

Figure 10 .
Figure 10.(a) The macroscale photo of the Cu/CuO/rGO/PPy wire.(b) The SEM image of the surface of a Cu/CuO/rGO/PPy wire.(c) The locally magnified SEM image of the peeled surface of the Cu/CuO/rGO wire.(d) The schematic diagram of the Cu/CuO/ rGO/PPy wire.The Cu wire (brown cylinder) in the core acts as the electron transport channel; the CuO (blue shell) and the rGO (semitransparent grey tablets) act as the heterojunction; the PPy (the outmost green shell) as the selective adsorption layer.(e) The schematic diagram of the physical-electrochemical sensing mode for detection of heavy metals (Hg 2+ ).After the specific adsorption of Hg 2+ on the PPy layer, the formed positively charged layer traps electrons from the exposed CuO by electrostatic interaction, which raises the conduction band of CuO, and increases the Φ B [62] Copyright 2020, Wiley.

Figure 11 .
Figure 11.(a) The DPV (differential pulse voltammetry) curves of the fabricated Cu/CuO/rGO/PPy wire in the detection of Hg 2+ in 0.1 M PBS.(b) The calibration curve of Hg 2+ concentrations versus peak current change.(c)-(d) The anti-interference performance of the fabricated Cu/CuO/rGO/PPy wire in the presence of (c) common metal ions and (d) organic pollutants with 100 times concentration (10 nM) of Hg 2+ [62] Copyright 2020, Wiley.

Figure 12 .
Figure 12.(a) Fundamental elements of chemiresistive sensor (b) Dip sensor structural design (c) Flow sensor structural.(e) unprocessed data acquired from a doped sensor with a 12 nm thickness of graphene-like carbon (GLC) and a 100-mV bias (d) Setup for automated surveillance of free chlorine levels in drinking water (f) Fitting calibration curves to Langmuir adsorption isotherms, exponential decay models, and Freundlich adsorption isotherms (g) reaction of a doped sensor to an actual drinking water sample and the concentrations artificially introduced afterwards (h) Mean concentrations foreseen by the sensor with and without contamination, with the actual sample concentration depicted by the purple line, and the standard deviations of sensor readings shown as error bars (all measurements conducted using 24 nm GLC at 10 mV) [63] Copyright 2022, RSC.

Figure 13 .
Figure 13.(a) Schematic representation of irregularly shaped structure and change in resistance in the presence of analyte (b) schematic representation of crystal growth-controlled structure after the crystal defect and change in resistance in the presence of analyte show the superior sensing [64] Copyright 2020, ACS.

Figure 16 .
Figure 16.The development of advanced AI sensors and the evolution of cloud and edge computing [72] Copyright 2023, Wiley.

Figure 17 .
Figure 17.Schematics of a (a) gas sensor array, (b) fabricated gas sensor, (c) voltage divider circuit and heater connection, (d)-(f) MLassisted classification and regression of CO and NO 2 and their mixtures [75] Copyright 2023, Elsevier.

Figure 19 .
Figure 19.Effects of ambient humidity from 0 to 80% RH on exposures to ethanol vapour on the resistance response, linear scale (a), resistance response, (b) logarithmic scale (c) and dielectric response (Z″ at 0.17 MHz) with a dramatic reduction in the RH effects (d), and the frequency dependence of R 2 values of the linear fit for different RH levels [76].Copyright 2020, Nature.

Figure 20 .
Figure 20.(a) Schematic experimental setup for continuous sensing.The sensor array housed in the chamber is sealed with a polydimethylsiloxane (PDMS) mold from the source-drain contact.Water is drawn by the piezoelectric motor into the sensor chamber and then continuously flows back to the external container to mimic the water flow.(b) Photograph of a 3D-printed sealed chamber with the sensors embedded inside.Inset is the cell interior.(c) Photograph of a piezoelectric micromotor and an external water container connected to the sealed sensor chamber.ML modelling for classifying and quantifying various mixed ionic and bacterial species in flowing water.(d) Summary of responses from Sensors 1-3 for Pb 2+ , Hg 2+ , and E. coli.(e-h) Principal component analysis (PCA) plots for the classifications of Pb 2+ , Hg 2+ , E. coli, and their mixtures.(i, j) Schematic of a two-layer artificial neural network (ANN) used for training and mean squared error (MSE) values for various numbers of neurons in the hidden layers.(k, l) Plots of the test and the predicted concentrations for Pb 2+ , Hg 2+ , and E. coli (ppb or cfu/mL) and their mixtures from the trained ANN model.Here, the dash line (Y = T) represents the perfect match between the test and the predicted concentrations for unknown samples while the solid line represents the fitting of the target concentrations predicted from the trained ANN model [77] Copyright 2023, Nature.

Figure 21 .
Figure 21.(a) Schematic illustration of the self-powered H 2 gas sensor structure and (b) H 2 gas detection principle according to volume expansion of Pd on the PUA nanograting.(c) FEM simulation results for Pd-PUA nanograting film by volumetric expansion of Pd in (i) 0.5% and (ii) 2% H 2 gas environments.(iii) The rate of change in the Pd gap of the Pd-PUA nanograting film (|Δw/w 0 |) according to the concentration of H 2 gas.(d) Optical image of fabricated self-powered H 2 gas sensor.(e) SEM images of Pd-PUA nanograting film.Top and cross-sectional images are in upper and lower panels, respectively [79] Copyright 2020, ACS.

Table 1 .
Change in resistance of the semiconductor sensing element made from metal oxide when exposed to reducing and oxidizing analytes.

Table 2 .
Sensor performances for MOs nanostructures based chemiresistive non-biological analyte sensors.