The UV Effect on the Chemiresistive Response of ZnO Nanostructures to Isopropanol and Benzene at PPM Concentrations in Mixture with Dry and Wet Air

Towards the development of low-power miniature gas detectors, there is a high interest in the research of light-activated metal oxide gas sensors capable to operate at room temperature (RT). Herein, we study ZnO nanostructures grown by the electrochemical deposition method over Si/SiO2 substrates equipped by multiple Pt electrodes to serve as on-chip gas monitors and thoroughly estimate its chemiresistive performance upon exposing to two model VOCs, isopropanol and benzene, in a wide operating temperature range, from RT to 350 ◦C, and LED-powered UV illumination, 380 nm wavelength; the dry air and humid-enriched, 50 rel. %, air are employed as a background. We show that the UV activation allows one to get a distinctive chemiresistive signal of the ZnO sensor to isopropanol at RT regardless of the interfering presence of H2O vapors. On the contrary, the benzene vapors do not react with UV-illuminated ZnO at RT under dry air while the humidity’s appearance gives an opportunity to detect this gas. Still, both VOCs are well detected by the ZnO sensor under heating at a 200–350 ◦C range independently on additional UV exciting. We employ quantum chemical calculations to explain the differences between these two VOCs’ interactions with ZnO surface by a remarkable distinction of the binding energies characterizing single molecules, which is −0.44 eV in the case of isopropanol and −3.67 eV in the case of benzene. The full covering of a ZnO supercell by H2O molecules taken for the effect’s estimation shifts the binding energies to −0.50 eV and −0.72 eV, respectively. This theory insight supports the experimental observation that benzene could not react with ZnO surface at RT under employed LED UV without humidity’s presence, indifference to isopropanol.


Introduction
The increasing demand of human society for portable detectors able to effectively monitor the environment air forces plenty of research activities in the direction to the development of gas sensors with advanced characteristics which yield a promising lowcost alternative to analytical instrumentation [1]. So far, such sensors primarily rely on a change of optical [2] or electrical [3] properties of various materials interfaced with a gas media [4]. These properties and their gas-induced changes could be properly read with several transduction principles of which the conductance measurement is still retained as the simplest but powerful method [5]. Since the pioneering research by Sejama [6], the commercial chemiresistors are developed employing mostly wide-gap semiconducting metal oxides which however require an external activation to facilitate charge exchange at the surface as the interface between the solid and a gas phase [7]. Such an activation implies conventionally heating up to 250-400 • C range which makes a need in advanced energy consumption of the sensor-based measuring units [8]. The common trend in electronics to miniaturization allows one to reduce the feed power for example with micro-hotplates [9] to go down to low hundreds of mW per sensor [10] but the cost of such systems goes up, too. As a reasonable alternative, rather recently it was discovered that the UV radiation could excite a chemiresistive effect at room temperature (RT) in metal oxides whose gap is close to the radiation energy [11]; even the power of cheap light-emitting diodes (LEDs) might be enough [12]. Nowadays, LED-activated oxide sensors find considerable interest when developing low-power gas-analytical units [13,14]. From this viewpoint, zinc oxide is considered to be an extremely appropriate material because of its bandgap, equal to ca. 3.4 eV [15], which falls into the range where UV, 100-400 nm of wavelength, could excite electron transitions. Therefore, ZnO has been actively investigated, for example, in the form of polycrystalline films [16,17], nanoparticles [18], nanorods [19][20][21], microspheres [22], subµm-thick fibers [23], single-crystalline sheets [24], and various nanostructured layers [25] as promising UV-activated gas sensor versus mainly alcohol vapors. In this case, the structures which have a large surface to interact with the gaseous phase but rather a thin bulk, down to nanodomain, to be effectively modulated by surface processes are advantageous.
Here we have got a look at the ZnO nanostructures grown by electrochemical deposition under constant potential over a substrate [26] equipped with multiple electrodes to further serve as a gas sensor. We have compared the response of grown ZnO to isopropanol and benzene as two model VOC analytes in a wide temperature range from RT to 350 • C with/without UV excitation. From a practical view, the observations further support that sensing of alcohol vapors with ZnO could be effectively supported by UV, 380 nm wavelength, at RT-150 • C temperature range while benzene shows a more complicated chemiresistive performance requiring a presence of H 2 O or higher energies to facilitate the response. From a fundamental view, the results are explained by the density functional tight-binding method (DFTB) to be matured from the differences in the binding energies of these two VOC molecules with the ZnO surface.

Electrochemical Deposition of ZnO Nanostructures
Zinc oxide nanostructures were electrochemically synthesized in an aqueous solution of 0.1 M Zn(NO 3 ) 2 at potential −1 V vs a reference Ag/AgCl sat electrode in the conventional three-electrode configuration as depicted in Figure 1 similar to protocols reported earlier for deposition of other oxides [27,28]. We utilized Si/SiO 2 substrate, 8 × 10 mm 2 , whose frontside was supplied with multiple Pt strip electrodes, ca. 120 µm width, in parallel to be distanced by a gap of ca. 80 µm as a multielectrode chip following the architecture reported earlier [29]. To monitor the operating temperature over the chip, the sensing area was surrounded by two meander Pt thermoresistors, 1 µm height, deposited simultaneously with the measuring electrodes by sputtering under the shadow mask. The rare side of the chip was equipped with four heating meanders made of Pt film, 1 µm height. We consider this platform design as the most suitable one for material characterization in the framework of "lab-on-chip" paradigm. The chip was wired into a ceramic holder which allowed us to selectively connect each Pt electrode or several Pt electrodes under the synthesis process [30]. A graphite rod was employed as a counter electrode. During the synthesis, we have maintained an electrolyte at a temperature of 80 • C using a hotplate defined to be optimal for ZnO growth in preliminary experiments. The electric currents and potential were measured with the help of a potentiostat (P50 Pro, Elins, Russia) employed to run The scheme for deposition of ZnO nanostructured layer over multi-electrode chip employing a three-electrode electrochemical cell: 1-a potentiostat interfaced with PC, 2-a simplified circuit diagram of electric potential application by the potentiostat, 3-electrolyte, 0.1 M Zn(NO3)2, 4graphite rod, 5-Ag/AgClsat electrode, 6-a hotplate, 7-a multi-electrode chip wired in a ceramic holder.
The noted time range has been chosen based on preliminary studies which ensured the absence of percolation for a current in the layers deposited at shorter time intervals while longer time resulted in a continuous thicker layer.

Material's Characterization
The morphology of the oxide layer deposited over the multielectrode chip has been evaluated by scanning electron microscopy (AURIGA CrossBeam, Carl Zeiss AG, Germany) at 5 kV of accelerating voltage. The layer composition has been checked with Raman spectroscopy (DXR™ xi Raman Imaging Microscope, ThermoFisher Scientific, Waltham, MA, USA) using a laser, 532 nm wavelength, at the power of 1 mW with ×50 objective. The crystal structure has been inspected by X-ray diffraction (XRD) employing Bruker D8 advance diffractometer (Bruker, Germany) with CuKα radiation source, λ = 0.15418 nm, under Bragg-Brentano geometry. Patterns were acquired in the 2Θ range from 10° to 90° at a 0.0098° step; accumulation time was 0.2 s. Soller mirrors of 2.5 mm were used. The acquired XRD patterns were processed using @Eva4.3 software. The samples were characterized following the operation of the chip in the course of reported measurements under heating up to 300 °C. The scheme for deposition of ZnO nanostructured layer over multi-electrode chip employing a three-electrode electrochemical cell: 1-a potentiostat interfaced with PC, 2-a simplified circuit diagram of electric potential application by the potentiostat, 3-electrolyte, 0.1 M Zn(NO 3 ) 2 , 4-graphite rod, 5-Ag/AgCl sat electrode, 6-a hotplate, 7-a multi-electrode chip wired in a ceramic holder.
The noted time range has been chosen based on preliminary studies which ensured the absence of percolation for a current in the layers deposited at shorter time intervals while longer time resulted in a continuous thicker layer.

Material's Characterization
The morphology of the oxide layer deposited over the multielectrode chip has been evaluated by scanning electron microscopy (AURIGA CrossBeam, Carl Zeiss AG, Germany) at 5 kV of accelerating voltage. The layer composition has been checked with Raman spectroscopy (DXR™ xi Raman Imaging Microscope, ThermoFisher Scientific, Waltham, MA, USA) using a laser, 532 nm wavelength, at the power of 1 mW with ×50 objective. The crystal structure has been inspected by X-ray diffraction (XRD) employing Bruker D8 advance diffractometer (Bruker, Germany) with CuKα radiation source, λ = 0.15418 nm, under Bragg-Brentano geometry. Patterns were acquired in the 2Θ range from 10 • to 90 • at a 0.0098 • step; accumulation time was 0.2 s. Soller mirrors of 2.5 mm were used. The acquired XRD patterns were processed using @Eva4.3 software. The samples were characterized following the operation of the chip in the course of reported measurements under heating up to 300 • C.

Gas-Sensing Measurements
To measure the chemiresistive effect in the grown ZnO layer, the multielectrode chip has been placed into a setup shown in Figure 2 as primarily reported earlier [31]. The setup consisted of two major parts aimed both to prepare analyte gases and to deliver them to the chip as well as to read out the resistances of the ZnO layer confined between two measuring electrodes with/without UV radiation. The gaseous analytes were supplied under a dynamic flow mode with a flowrate managed by precision mass-flow controllers (MKS Instruments, Inc., Andover, MA, USA) at 500 sccm.
with source mixtures of isopropanol/synthetic air and benzene/air, each of 250 ppm concentration, at the corresponding ratio. The source analyte mixtures were contained in bottles as prepared. The measurements under humidity, 50 rel. %, enriched conditions were carried out via forking the dry air into two lines, where one line was purged through a Drexel flask containing deionized water, with further mixing at a 1:1 ratio. The chosen analyte concentrations correspond to or below the exposure limits of isopropanol and benzene vapors in the air (see, for instance [32]. Both gas pipe lines corresponding to reference air and analyte-enriched air are supplied with a humid generator (a bubbler) to vary independently the humidity content in analyte probes. The humidity content in the gas probe delivered to the chip under study has been independently monitored with a hygrometer (Testo, model 601, Titisee-Neustadt, Germany).  The analyte gas concentration in the probe, 1, 10 and 100 ppm, has been adjusted by mixing the dry filtered air from the generator (SylaTech GmbH, Walzbachtal, Germany) with source mixtures of isopropanol/synthetic air and benzene/air, each of 250 ppm concentration, at the corresponding ratio. The source analyte mixtures were contained in bottles as prepared. The measurements under humidity, 50 rel. %, enriched conditions were carried out via forking the dry air into two lines, where one line was purged through a Drexel flask containing deionized water, with further mixing at a 1:1 ratio. The chosen analyte concentrations correspond to or below the exposure limits of isopropanol and benzene vapors in the air (see, for instance [32]. Both gas pipe lines corresponding to reference air and analyte-enriched air are supplied with a humid generator (a bubbler) to vary independently the humidity content in analyte probes. The humidity content in the gas probe delivered to the chip under study has been independently monitored with a hygrometer (Testo, model 601, Titisee-Neustadt, Germany).
To perform measurements of the chemiresistive effect in the deposited ZnO nanostructured layer, the multielectrode chip was placed in a sealed measuring steel-based chamber equipped with a UV light-emitting diode installed to illuminate the front side of the chip with the layer as depicted in Figure 2b. The chip has been secured by two dielectric rings inside the chamber. The LED emission wavelength was checked with a spectrometer (BTC112 Series Fiber Coupled TE Cooled CCD Spectrometer, B&W Tek, Newark, NJ, USA). The efficient values estimated using a full width at half maximum of the LED spectrum were 375-385 nm ( Figure 2c) that yields photons with energy equal to ca. 3.2-3.3 eV. The LED was powered by a supply unit (BEHA Uniwatt Labornetzteil NG 304T, Germany). To heat the multielectrode chip with temperatures up to 350 • C, four Pt meander heaters were realized at the chip backside. The temperature was independently controlled by calibrated Pt thermoresistors located at the edges of the front side of the chip [30].
We measured the resistance of the ZnO layer between each couple of Pt electrodes as a chemiresistor utilizing a KAMINA electrical measuring unit [33] with a sampling rate of 1 Hz per all the chemiresistor array located at the chip. The electrical measuring unit was controlled by a homemade software interfaced to the PC to allow one reading, storing, and visualization of the received data in real-time. The chemiresistive properties of the ZnO layer were studied in the temperature range from RT, equal to approx. 20 • C, to 350 • C towards isopropanol and benzene, 1-100 ppm of concentration, in the atmosphere of both dried and humid, 50 rel. %, air. The chip operating temperature was driven by the in-built meander heaters, with approx. 1 • C accuracy, and monitored with the thermoresistors. The total temperature in the lab during the measurements was secured by air conditioning at the given value. The temperature of gas flow going to the chamber with a chip was monitored with a thermometer combined with the hygrometer.
The time of analyte exposure to the chip was adjusted to be 90 min while air purging was 120 min accounting for preliminary studies to ensure long-term transients as observed frequently in practice. The chemiresistive response to the analytes was calculated in percent as where R a , R g are resistances of the segment of ZnO layer, confined between each couple of measuring electrodes at the chip, in pure reference air and analyte-enriched air, respectively. The multi-dimensional signal of the array of ZnO layer segments located at the chip was processed by Linear Discriminant Analysis (LDA) technique [34]. Because of the large variation in analyte concentration in the probes, of two orders of magnitude, we have taken a logarithm of raw resistance values recorded under various impacts as a pre-processing technique to feed the LDA algorithm.

DFTB Calculations
The search for the ground state of the considered objects as well as the calculation of their band structures were performed by the self-consistent charge density-functional tight-binding method (SCC DFTB) [35]. In terms of computational speed, the SCC DFTB method is comparable to traditional semi-empirical methods, but, in general, provides the accuracy comparable to first-principal calculations. The method is based on the secondorder decomposition of the total Kohn-Sham energy with respect to the charge density. The matrix elements of the unperturbed Hamiltonian H 0 µv are represented in the minimal basis of atomic orbitals using the two-particle approximation. The total energy of the system is determined by the expression where c i µ and c i v are the coefficients in the decomposition of the molecular orbitals into atomic orbitals, ∆q α and ∆q β are the charge fluctuations on the α and β atoms, respectively, γ αβ is a function, that exponentially decreases with increasing a distance between the α and β atoms and directly depends on the chemical hardness [36], E rep is the term describing the repulsive interaction at small distances, E dis is the term responsible for the van-der-Waals interaction of unbound atoms. The Monkhorst-Pack 12 × 12 × 1 grid was used to sample the Brillouin zone in the reciprocal space. The 3-ob-3-1 basis set was applied to describe the interaction between the Zn, O, C and H atoms [37]. It should be noted that the authors successfully applied the SCC DFTB method earlier to calculate the binding energy between carbon nanostructures and various functional groups including iron oxide [38] and DNA [39]. In addition, this method accurately describes the behavior of H 2 O clusters [40,41] to be within the scope of the current study.
The binding energy of the 2D ZnO surface with VOCs under study, isopropanol or benzene, and with water molecules, equal to number N, was calculated utilizing the following formulas, accordingly: where E(ZnO), E(VOC) and E(H 2 O) are the energies of isolated ZnO supercell, analytes and water structures, E(ZnO + VOC) is the energy of the oxide with bonded analyte at the ground state, and E(ZnO + NH 2 O) is the energy of the oxide with H 2 O molecules at the ground state.

The Physical Characterization of ZnO Nanostructured Layer
Primarily, the ZnO layer has been grown by the electrochemical deposition over Pt electrodes according to known mechanisms [42]. However, the oxide growth has appeared to occur not only over the electrode but also in the between-electrode gaps primarily in the shape of nanodimensional structures. The size and coverage of the gap by the oxide depend on deposition time leading to an electric percolation of the dielectric area between electrodes. The structure of the deposited oxide layer has been verified by Raman spectroscopy; the typical spectrum is shown in Figure 3g. The data of measurements have proved the layer to grow as ZnO polycrystals with wurtzite symmetry. According to literature [43][44][45], we may note various modes characterizing this structure as A 1 + E 1 + 2E 2 + 2B 1 . Raman and IR active modes A 1 and E 1 split into transverse optical (TO) and longitudinal optical (LO) phonons whose corresponding peaks are found at 383.6 cm −1 , 568.7 cm −1 , 586.1 cm −1 , and 1153.0 cm −1 [46]. Raman active E 2 mode includes two nonpolar modes E 2 low and E 2 high of low-and high-frequency phonons, observed at 98.2 cm −1 and 202.3 cm −1 , 437.6 cm −1 , and 331.5 cm −1 for E 2 low , E 2 high and E 2 high -E 2 low , respectively [47]. Phonon LO and TO peaks are identified at 661.3 cm −1 , 1105.5 cm −1 and 1153.0 cm −1 . High-intensity peaks of E 2 confirm the wurtzite structure while the appeared LO phonons indicate the presence of oxygen defects and respective electronic states in the crystals, which have rather a crucial effect for a chemiresistive effect observation. Still, we have not found B 1 mode in the spectrum similar to other works [47,48]. We also observed bands classified as 2TA, 2LA, and TA+LO at 202.3 cm −1 , 483.8 cm −1 and 661.3 cm −1 , respectively [44,48]. The peaks found at ca. 520.5 cm -1 and 971.7 cm −1 seem to come from the SiO 2 /Si substrate [49].
The annealed ZnO layer is characterized by a hexagonal structure, space group P63mc, just according to XRD results (Figure 3h

The Electrical Properties of the ZnO Nanostructured Layer
We have tested the electrical properties of the electrochemically grown ZnO layer under the application of dc electric fields. Because the interface contacts between the oxide and electrode might significantly modify the electric transport, primarily we have looked at the current-voltage (I-V) characteristics in the range of applied potentials at [−5; +5] V in two directions. We have checked the I-Vs almost in all the temperature range of operating temperatures employed in the course of studies, mainly under air conditions, to ensure its linear character. Here, for example, we show I-Vs in Figure 4 under heating to 150 • C recorded in a dry air environment. This temperature is most interesting as a "bordering" one which we show further for the gas-sensing performance of the ZnO layer.
As one can see in Figure 4, the I-V curves are quite linear, which indicates the absence of significant potential barriers at the contact interface in all the ZnO layer areas grown under different deposition times. The slope of I-V curves, which defines a resistance, depends on the deposited volume of the oxide; we have clarified it by plotting the resistance values derived from the I-Vs at the second ordinate axis. The ZnO layer segment resistance goes down with deposition time from ca. 210 kOhm observed at the area grown at 400 s to around 130-140 kOhm in the area grown at 850-1150 s. When comparing them to SEM images of Figure 3, the resistance becomes almost independent on a deposition time at these higher magnitudes of deposition time because of the formation of a rather continuous layer, though still containing mesopores.
As one can see in Figure 4, the I-V curves are quite linear, which indicates the absence of significant potential barriers at the contact interface in all the ZnO layer areas grown under different deposition times. The slope of I-V curves, which defines a resistance, depends on the deposited volume of the oxide; we have clarified it by plotting the resistance values derived from the I-Vs at the second ordinate axis. The ZnO layer segment resistance goes down with deposition time from ca. 210 kOhm observed at the area grown at 400 s to around 130-140 kOhm in the area grown at 850-1150 s. When comparing them to SEM images of Figure 3, the resistance becomes almost independent on a deposition time at these higher magnitudes of deposition time because of the formation of a rather continuous layer, though still containing mesopores.

Gas-Sensing Properties of ZnO Nanostructured Layer under UV Illumination
We have studied the multisensor chip with an electrochemically grown ZnO layer upon exposing to two model analytes, isopropanol and benzene, mixed with dry and wet air in a wide temperature range from RT to 350 °C. The influence of UV irradiation on the chemiresistive effect in the ZnO layer has been checked at these temperatures. These data are summarized in Figure 5, where we depict the temperature dependence of the chemiresistive response to benzene (Figure 5a,b) and isopropanol (Figure 5d,e), given in concentration of 100 ppm, where the chemiresistive response and the observed variations are more remarkable than ones in the case of lower analyte concentration. As a typical behavior, we have taken the ZnO layer segment deposited for 550 s as an example.
The data plots show that the UV radiation has a significant effect to support the appearance of the chemiresistive response in ZnO versus isopropanol at temperatures from room one to 150 °C (highlighted in the figure), no matter the humidity vapors co-exist in the atmosphere. We may see a rather significant response even at RT, about 20-25%, which enhances up to 250% at T = 150 °C. A similar effect is observed, too, when exposing the chip to benzene mixed with wet air. However, the ZnO conductivity stays almost inert

Gas-Sensing Properties of ZnO Nanostructured Layer under UV Illumination
We have studied the multisensor chip with an electrochemically grown ZnO layer upon exposing to two model analytes, isopropanol and benzene, mixed with dry and wet air in a wide temperature range from RT to 350 • C. The influence of UV irradiation on the chemiresistive effect in the ZnO layer has been checked at these temperatures. These data are summarized in Figure 5, where we depict the temperature dependence of the chemiresistive response to benzene (Figure 5a,b) and isopropanol (Figure 5d,e), given in concentration of 100 ppm, where the chemiresistive response and the observed variations are more remarkable than ones in the case of lower analyte concentration. As a typical behavior, we have taken the ZnO layer segment deposited for 550 s as an example.
The data plots show that the UV radiation has a significant effect to support the appearance of the chemiresistive response in ZnO versus isopropanol at temperatures from room one to 150 • C (highlighted in the figure), no matter the humidity vapors co-exist in the atmosphere. We may see a rather significant response even at RT, about 20-25%, which enhances up to 250% at T = 150 • C. A similar effect is observed, too, when exposing the chip to benzene mixed with wet air. However, the ZnO conductivity stays almost inert if benzene vapors are supplied in a mixture with dry air under UV radiation until the temperature goes up to 150 • C. At T = 100 • C the response is ca. 1.4% which is similar to results published in other studies, see, for instance [50]. When heating goes up from 150 • C to 350 • C, the ZnO layer starts to exhibit the chemiresistive properties to be almost independent of UV irradiation. It seems at this temperature range the thermal activation is the major factor to facilitate a charge exchange between both chemisorbed analytes, isopropanol and benzene, and ZnO surface as discussed later. Still, we should note that employing operating temperatures up to 350 • C leads to a substantial enhancing of the chemiresistive response to both analytes by more than one order of magnitude when compared to one observed at T = 150 • C even under UV assistance. The response is saturated at T = 300-350 • C that fully corresponds to conventional operating temperatures known for ZnO chemiresistors [5], the UV effect is almost negligible. Due to the fine morphology of the grown ZnO nanostructures, the response is higher than the observed one in printed ZnO layers doped with Pt [51].
independent of UV irradiation. It seems at this temperature range the thermal activation is the major factor to facilitate a charge exchange between both chemisorbed analytes, isopropanol and benzene, and ZnO surface as discussed later. Still, we should note that employing operating temperatures up to 350 °C leads to a substantial enhancing of the chemiresistive response to both analytes by more than one order of magnitude when compared to one observed at T = 150 °C even under UV assistance. The response is saturated at T = 300-350 °C that fully corresponds to conventional operating temperatures known for ZnO chemiresistors [5], the UV effect is almost negligible. Due to the fine morphology of the grown ZnO nanostructures, the response is higher than the observed one in printed ZnO layers doped with Pt [51]. We have looked at how UV radiation influences the ZnO nanostructures grown in local chip areas at various deposition times. While the UV radiation dominates in the RT-100 °C of the temperature range employed to activate the chemiresistive response in this material, we considered T = 150 °C as a "border" point where to compare the UV contribution in comparison to thermal activation. In order to relatively estimate the UV effect, we have considered the change of chemiresistive response under UV, SUV, in ratio to the response, S, observed without UV, as − 1 and plotted it in dependence on the ZnO deposition time for benzene ( Figure 5c) and isopropanol (Figure 5f). The derived data clearly show that the ZnO nanostructures synthesized for the time in the range 400-700 s, exhibit the most pronounced influence of UV impact, which enhances the chemiresistive response up to 4 times. In the range 850-1150 s of deposition times, the UV effect is less distinct and rather stable, just doubling the response. These clear differences seem to mature from the morphology of the ZnO layers: the layer deposited at lower times has a more We have looked at how UV radiation influences the ZnO nanostructures grown in local chip areas at various deposition times. While the UV radiation dominates in the RT-100 • C of the temperature range employed to activate the chemiresistive response in this material, we considered T = 150 • C as a "border" point where to compare the UV contribution in comparison to thermal activation. In order to relatively estimate the UV effect, we have considered the change of chemiresistive response under UV, S UV , in ratio to the response, S, observed without UV, as S UV S − 1 and plotted it in dependence on the ZnO deposition time for benzene ( Figure 5c) and isopropanol (Figure 5f). The derived data clearly show that the ZnO nanostructures synthesized for the time in the range 400-700 s, exhibit the most pronounced influence of UV impact, which enhances the chemiresistive response up to 4 times. In the range 850-1150 s of deposition times, the UV effect is less distinct and rather stable, just doubling the response. These clear differences seem to mature from the morphology of the ZnO layers: the layer deposited at lower times has a more dispersed meso-nanostructure where the number of free electrons could be effectively facilitated by UV, as we have recently shown in the case of the TiO 2 nanotubes [52]. At thicker layers, the LED power is not enough for the UV radiation to penetrate the entire depth of the oxide layer including internal meso-surfaces, and its effect on the interaction of gaseous analytes to be chemisorbed at the ZnO surface and to participate in the charge exchange is reduced.
To further show the impact of UV radiation on the ZnO resistance under exposure to analytes, we have plotted the typical R(t) transients measured at the operating temperature of 150 • C in Figure 6a,b for the case of isopropanol; the dry air background is given in Figure 6a while wet, 50 rel. %, air background is given in Figure 6b. Firstly, as one can see, the resistance values of the ZnO layer significantly drop following UV exposure: in dry air, the drop is by more than ca. 20 times, while in the wet air it is about 5 times. The drop follows the known UV impact on the wide-band semiconducting oxides related to a generation of free electrons in the oxide conductance band due to inter-gap transitions that, in turn, increases the material's conductivity; see, for instance, [52]. Furthermore, these processes reduce inter-crystal potential barriers that also promote higher conductivity [53]. The observed differences between dry and wet conditions indicate that the excessive free electrons could be partly localized at the surface by additional chemisorbed OHspecies induced under a presence of humidity vapors according to the following dissociation of water molecules at the oxide surface thus reducing the generated number of free electrons. Secondly, the appearance of isopropanol vapors results in a further drop of ZnO sistance which constitutes, in fact, the chemiresistive effect under interest for the sen performance. The reversible character of R(t) behavior with/without the VOC analyte c Secondly, the appearance of isopropanol vapors results in a further drop of ZnO resistance which constitutes, in fact, the chemiresistive effect under interest for the sensor performance. The reversible character of R(t) behavior with/without the VOC analyte confirms the functionality of the synthesized layer as a gas sensor. According to a rather conventional view [5,14], this effect seems to mature from the surface interaction between this adsorbed alcohol and chemisorbed oxygen species, O − or O 2 − , and/or OH − groups according to one of the following chemical routs All these reactions lead to enhancing a free electron concentration, which increases the conductance of ZnO as an n-type semiconductor in accordance with the experimental results.
In the case of benzene, this VOC's interaction with the ZnO surface is more complex accounting for data presented in Figure 5. It seems the employed UV energy has not been enough to facilitate the interaction of benzene directly with chemisorbed oxygen species; we discuss this phenomenon later. However, the humidity-derived species at the ZnO surface appeared due to H 2 O presence in the atmosphere could interact with benzene according to the next chemical routes Furthermore, it should be noted that comparing R(t) curves displayed in Figure 6a,b shows that the characteristic time of the chemiresistive effect is substantially reduced by UV radiation. When LED-UV is Off, 1.5 h of exposure to the isopropanol is still not enough to reach a rather stationary plateau of the resistance value of the ZnO layer. The same concerns the recovery time following other reports [22]. Altogether, the gas response looks partly irreversible within the chosen time intervals of the analyte's exposure because of not enough activation of chemisorption processes on the ZnO surface [5]. At the same time, the ZnO resistance decreases much faster upon/after the exposure to isopropanol vapors under UV assistance. We may distinguish ca. 10-15 min of response/recovery time, which characterizes the resistance transient curves, similar to other research [54] that is already enough for practical applications. Still, the reported values could be further advanced when employing noble catalysts and heterostructures (see, for instance, [55]).
As a major sensor parameter, we have plotted the dependence of chemiresistive response (1) on the analyte's concentration measured at 150 • C in Figure 6c,d towards isopropanol (c) and benzene (d) vapors in dry air and humid-enriched air. The curves could be well fitted by Freundlich isotherm, S ∼ C α , where C is the analyte's concentration and α is the power index which could be derived as a slope of S(C) curve at a log-log scale of the plot [29]. All the values of α are summarized in Table 1. Table 1. The power index α of the S(C) curve for exemplary ZnO layer grown electrochemically for 550 s when exposed to isopropanol and benzene vapors at 150 • C. In general, the S(C) curves characterizing the isopropanol are steeper than the ones for benzene, the α index is rather similar to fluctuate at 0.53-0.62 range which is typical for alcohol's interaction with ZnO [56]. In benzene vapors, the α index drops to a value equal to 0.11 in dry air that indicates a rather weak dependence of sensor signal on the analyte's concentration; still, the sensor signal is low, too. Again, the presence of humidity significantly enhances the S(C) slope that yields α equal to 0.32-0.52. The detection limit which is frequently defined in empirical studies as the analyte concentration corresponding to 10% response [57] goes down to the sub-ppm range for both analytes under the given UV assistance.

UV
Although we may note that the ZnO layer's response to isopropanol is higher when compared to benzene, we cannot distinguish the analyte employing just a single sensor. To advance the selectivity, we have to take a vector signal generated by a multisensor array which we fabricated on the multi-electrode chip's substrate according to the electronic nose's paradigm [58]. With this purpose, again we analyze the most interesting condition of heating the chip up to 150 • C where UV still contributes to the chemiresistive effect. To process the multisensor's vector signal, we employ the LDA technique as detailed in the Materials and Methods section. This technique transfers the vector signal into the optimized coordinate system to separate the vector's classes at the maximum degree [59]. The full dimensionality of the system is (N-1) where N is a number of classes to distinguish. In our cases, we have considered the classes which correspond to analytes, the presence of H 2 O vapors, and UV, altogether equal to 12 in our study. The analyte probes are taken at 10 ppm concentration, as an example. Figure 7 displays the derived LDA coordinate system, a plane of the first two components.  [59]. The full dimensionality of the system is (N-1) where N is a number of classes to d tinguish. In our cases, we have considered the classes which correspond to analytes, t presence of H2O vapors, and UV, altogether equal to 12 in our study. The analyte prob are taken at 10 ppm concentration, as an example. Figure 7 displays the derived LDA c ordinate system, a plane of the first two components. As one can see, the multi-dimensional array signals from the array are clustered in corresponding groups according to their class's origin. Notably, these clusters appear As one can see, the multi-dimensional array signals from the array are clustered into corresponding groups according to their class's origin. Notably, these clusters appear in four sections defined primarily by the presence of humidity and UV (On/Off) as dominant factors for the sensor performance. Within each of the four sections, all the analyte-related clusters and reference air are further distinguished which makes it possible to selectively justify them. The higher selectivity could be attained via introducing other external modifications of the sensor signal via, for instance, a gradual surface impurity by foreign atoms [29] or applying temperature gradients [30].

DFTB Calculation of Analyte Interaction with ZnO Surface
The DFTB calculations have been performed for the two-dimensional ZnO network as a supercell of 96 atoms taken from the ZnO crystal of P63mc space group [60] to be cut along the (1100) plane (Figure 8a). The translation vectors along the X-and Y-axes were 12.819 Å and 16.655 Å, respectively. Such dimensions of the ZnO supercell allowed us both to fit the isopropanol and the benzene ring molecules over the oxide surface and to exclude the interaction between analytes after the translation. At the first stage of the study, the VOC analyte molecules were approached to the pristine surface of the ZnO supercell ( Figure 8b). We have found that the binding energy between isopropanol and ZnO was −0.44 eV, while the same one between ZnO and benzene was −3.67 eV. The latter rather high value is explained by a formation of chemical bonds between two zinc atoms of the supercell and two carbon atoms of the benzene ring. The length of the Zn-C bond was 2.21 Å. supercell (Figure 8b). We have found that the binding energy between isopropanol and ZnO was −0.44 eV, while the same one between ZnO and benzene was −3.67 eV. The latter rather high value is explained by a formation of chemical bonds between two zinc atoms of the supercell and two carbon atoms of the benzene ring. The length of the Zn-C bond was 2.21 Å. Further, we have searched for a maximum number of H2O molecules to possibly set over the ZnO supercell in order to simulate the experimental conditions at humid-enriched air. The humidity molecules were sequentially placed on the ZnO surface. Following each next bonding of the H2O molecule, the total system was re-optimized and the binding energy of the ZnO supercell with the humidity molecule was calculated according to the Formula (2). It was found that the maximum number of H2O molecules on the surface of the ZnO supercell is 19 which allows the system to have minimum total energy under the humidity adsorption. The further process of humidity adsorption is energetically unfavorable.
Finally, the VOC analytes were approached to the ZnO surface covered with 19 H2O molecules (Figure 8c). It turned out that the binding energy of the ZnO supercell with the isopropanol got changed slightly when compared to the pristine case being amounted to −0.50 eV. At the same time, the binding energy of ZnO with benzene was evaluated to be −0.72 eV that yields a dramatic reduction when compared to the situation observed with the pristine surface. It might be explained by the fact that the ZnO coating with water molecules does not allow forming of the chemical bonds between Zn and C atoms as dis- Further, we have searched for a maximum number of H 2 O molecules to possibly set over the ZnO supercell in order to simulate the experimental conditions at humid-enriched air. The humidity molecules were sequentially placed on the ZnO surface. Following each next bonding of the H 2 O molecule, the total system was re-optimized and the binding energy of the ZnO supercell with the humidity molecule was calculated according to the Formula (2).
It was found that the maximum number of H 2 O molecules on the surface of the ZnO supercell is 19 which allows the system to have minimum total energy under the humidity adsorption. The further process of humidity adsorption is energetically unfavorable.
Finally, the VOC analytes were approached to the ZnO surface covered with 19 H 2 O molecules (Figure 8c). It turned out that the binding energy of the ZnO supercell with the isopropanol got changed slightly when compared to the pristine case being amounted to −0.50 eV. At the same time, the binding energy of ZnO with benzene was evaluated to be −0.72 eV that yields a dramatic reduction when compared to the situation observed with the pristine surface. It might be explained by the fact that the ZnO coating with water molecules does not allow forming of the chemical bonds between Zn and C atoms as discussed above. All these observations support the experimental data that UV irradiation of approx. 3.2 eV could stimulate the adsorption of benzene molecules in presence of humidity in the atmosphere.

Conclusions
Following the obtained results, we may conclude that the electrochemical deposition suggests options to grow gas-sensitive ZnO nanostructured layers directly on chip substrates, ready for preparing multisensor arrays, with a varied morphology via adjusting a growth time. Such sensors could be further activated by UV illumination powering by an LED to exhibit the chemiresistive effect at rather low temperatures, below 150 • C. The finer oxide structures grown at lower deposition times are more effectively activated by UV due to a higher surface-to-bulk ratio. We have found clear differences in the observed ZnO response under UV illumination at low operating temperatures towards two model VOCs, isopropanol and benzene. While the isopropanol yields a distinguished chemiresistive signal in dry or wet air backgrounds, the benzene appears to be inert in the absence of interfering H 2 O vapors. We argue that a possible reason for such an effect might be the rather higher energy required to adsorb benzene molecules over the ZnO surface which is supported by DFTB considerations. These results should be accounted for when developing low-power gas (multi)sensors activated by LEDs for environmental monitoring.