Analytical Gas‐Sensing Responses for Single‐Crystalline Semiconducting Gas Sensors

In this study, an analytical gas response formula is developed based on the classical gaseous molecule adsorption model. To validate the analytical formula, an array of silicon nanowires or microwires is fabricated by patterning the device layer of a silicon‐on‐insulator (SOI) wafer. Gas Hall effect measurements reveal that the surface depletion of the wires narrows down upon exposure to polar molecules that act as dipoles to partially neutralize the surface charges. Gas conductance measurements are performed on different sizes of wires at elevated temperatures. The responses on exposure to different concentrations of water, acetone, and ethanol vapors can be well fitted with the analytical gas response formula. The extracted parameters are consistent with their physical nature. Additionally, the established analytical gas response formula can also well fit gas responses of single‐crystalline nanowires in literature, which further validates the analytical formula.


Introduction
[6][7] To enhance sensitivity, low-dimensional materials such as nanoparticles, [8][9][10][11] polycrystalline thin films, [12][13][14] chemical-vapor-deposited nanowires (NWs) [15][16][17] and 2D atomically thin materials [18][19][20] are often used as gas sensors due to their high surface-to-volume ratio (SVR) compared with bulk material.Although the sensing mechanism of chemiresistive gas sensor has been extensively investigated, [21][22][23][24] an analytical gas-sensing formula that allows for designing gas sensors and predicting sensor responses has yet to be established due to the challenges to control the dimensionality, surface states, and crystal quality of these lowdimensional materials.For example, high-quality 2D materials have an SVR of nearly infinite but with no surface states as binding sites for molecular adsorption. [25]For this reason, gas sensors based on atomically thin 2D materials often require surface DOI: 10.1002/adsr.202300071functionalization by polar molecules. [25]harges near the 2D materials will create a nonuniform conduction channel, which, although creates a high sensitivity, makes it difficult to establish analytical gas-sensing responses.Nanoparticles or quantum dots are also potential materials for high-performance gas sensors due to their extremely high SVR.However, thin films based on these materials have less controllable surfaces and low-quality electrical conduction channels, as a result of which no analytical equations were established for their electrical properties.1D nanowires have a continuous channel and a relatively high SVR, which therefore are a good system for gas-sensing applications as investigated by many research groups from different aspects [26] (diameters, [27] surface states, [28,29] catalysts, [30,31] and orientations [32,33] ).However, chemically synthesized nanowires have a poor quality of surfaces and are less repeatable from device to device due to the incompatibility with standard complementary metal-oxide-semiconductor (CMOS) processes.In contrast, nanodevices fabricated by patterning commercial wafers with the standard CMOS processes have a high precision and reproducibility in terms of device size, surface quality, and material crystallinity, and therefore are the best candidate for exploring analytical gas-sensing responses.36][37][38][39][40] In this work, we fabricated an array of silicon wires with different width by patterning the device layer of a silicon-on-insulator (SOI) wafer.The dependence of conductance on wire width reveals the existence of a depletion region near the wire surfaces.The gas responses of the silicon wires were investigated upon exposure to water, acetone, and ethanol vapors.Gas Hall effect measurements were employed to monitor the width variation of surface depletion regions upon gas exposure.Gas conductance measurements were conducted to investigate the gas-sensing responses on wire width and temperature dependence.An analytical gas-response formula developed on the gaseous molecule adsorption model was validated by these experimental measurements.

Results and Discussion
Silicon wires were fabricated by patterning the device layer (220 nm thick, p-type, N a = 10 18 cm −3 ) of a SOI wafer with electron beam lithography (EBL) and reactive ion etching (RIE).The silicon wires become responsive to gaseous vapors at room temperature only after the oxygen plasma treatment.The wires are 27 m long and the width ranges from 50 to 800 nm.As shown in the false color scanning electron microscopic (SEM) image of a typical wire in Figure 1a, each wire has five silicon pads (gray, A-E), on which 200 nm thick aluminum is deposited to form electrical contacts (yellow).The current of the wires (80-800 nm) is linearly correlated with the bias voltage (Figure 1b) except the narrowest one (50 nm, black).For 50 nm wide nanowire, the current starts to saturate at a bias of 0.5 V with a current decrease by more than three orders of magnitude (from 8 A at 800 nm to ≈2 nA at 50 nm), as shown in the inset of Figure 1b.This phenomenon is attributed to the wire conduction transition from a continuous channel to a pinched-off one when the wire width decreases below the threshold (2W dep ). [40]Surface depletion region width (W dep ) is strongly correlated with the surface charge density Q s (sensitive to surface modification, environmental condition, etc.), which lays the foundation for crystalline semiconductor gas-sensing mechanism.
When the wires are exposed to gaseous analytes, the analyte molecules will be physically adsorbed onto the semiconductor surfaces.The physical adsorption process is best described by the Temkin isotherm model [41] thermodynamically as Equation ( 1) in which q e is the adsorption capacity, K T is the Temkin isotherm constant (g mol −1 ), C e is gaseous analyte concentration in adsorbent at equilibrium (mol g −1 ), R is the universal gas constant (J mol −1 K −1 ), T is the temperature (K), and b is the Temkin constant (J mol −1 ), which is related to the adsorption heat of analytes, as well as the properties of surfaces.
If the analytes are polar molecules, each molecule will act as a dipole.The dipoles of polar molecules will be all aligned to the direction of surface electric field unless the molecules can unipolarly bind to the surface dangling bonds via weak chemical bonds.For this reason, the adsorption of polar molecules is equivalent to the introduction of additional charges on semiconductor surfaces.The variation in surface charge concentration induced by the adsorption of polar molecules will be proportional to the change in adsorption capacity of molecules (assuming that the gas sensor is completely aged), given by Equation (2) (see Section S1 in the Supporting Information for derivations) in which  is a constant related to the dipole strength of the molecules (C g mol −1 cm −2 ), C 0 is the initial equilibrium concentration, referring to the concentration of molecules that are adsorbed and will not be desorbed upon the initial exposure to analytes.This phenomenon is not uncommon in semiconductor gas sensors, which often need to be aged and become adaptive to specific sensing environment.The change in surface charges will induce a variation in surface depletion region ΔW dep ( = W dep − W dep0 ) and thus create a response in electrical conductance.W dep0 is the initial depletion region width of as-etched NWs without analyte gas molecule adsorption, which is estimated from the xintercept by extending the linear correlation of conductance with Si wire width, as depicted in Figure 1c.The total depletion region thicknesses of our p-type Si wires before (black) and after (red) oxygen plasma processing are ≈72 and ≈60 nm, corresponding to W dep equal to ≈36 and ≈30 nm, respectively.The shrinking of depletion region caused by plasma processing leads to an observable upshift of conductance for each p-type Si wires (see Section S2 in the Supporting Information for downshifts of n-type wires).The inset in Figure 1c shows the IV curves of a 400 nm wide Si wire before and after oxygen plasma.
It is interesting to note that only after processed by oxygen plasma do the wires show observable responses to gaseous analytes.Our simulation results show that the depletion region width reaches a maximum of ≈35 nm for a silicon nanowire doped with boron at a concentration of 1 × 10 18 cm −3 (see red dots in Figure 1d), which is in line with the wire surface depletion region width before oxygen plasma treatment in Figure 1c.In this case, the Si wire surface is in strong inversion and a high concentration of inversion electrons is formed near the surfaces.The adsorption of molecule dipoles has a negligible influence on the depletion region width due to the charge-screening effect of the inversion electrons. [42]For this reason, the p-type Si wires before oxygen plasma treatment are insensitive to polar analytes.After oxygen plasma treatment, negative oxygen ions will partly compensate the positive surface charges, reducing the surface depletion region to ≈30 nm.The Si wire surfaces have regressed from the strong inversion into depletion where the free charge carriers are orders of magnitude lower (see open circles in Figure 1d).The adsorbed molecular dipoles can induce a linear change in W dep (solid dots in Figure 1d) due to a large Debye length in depletion region. [42]Therefore, the Si wires become sensitive to gaseous analytes.In the following discussion, all Si wires were processed by oxygen plasma.
In the linear response regime, the slope 1 × 10 −18 cm 3 in our case).Therefore, we have where ΔQ gas is the charge variation induced by absorption of polar analytes.In the following sections, gas Hall measurements and conductance measurements will be introduced separately to verify this theory.

Gas Hall Measurements
For our Si wire gas sensors, there are three (two sidewalls and one top surface) depletion regions, which can respond to the external environment.Only the top depletion region surface has responses in the Hall measurements.In the experiment, a constant current of 100 A was applied through electrodes A and C.
A Hall voltage will be induced across electrodes B and E under an upward magnetic field.As shown in Figure 2a, the Hall voltage was monitored in the ON/OFF flux of gaseous analytes (see Section S3 in the Supporting Information for experimental setup).The analyte concentration is controlled by the flow rate ratio between the pure nitrogen and saturated vapor carried by nitrogen (bubble through the aqueous analyte).The Hall voltage decreases upon exposure of gaseous analytes and recovers when the flow is switched to the pure nitrogen at the same flow rate.To reduce the noise and attain more reliable data, we measured the Hall resistance as a function of magnetic field in pure nitrogen (black squares) and diluted analytes (red dots) separately (Figure 2b).It is known that the Hall resistance R H is linearly correlated with the magnitude field B where H ch is the conduction channel thickness of the wire which is equal to H nw − W dep with H nw being the wire physical thickness (≈220 nm).The slopes of R H ∼ B dependence is equal to 1 qN a H ch which decreases and eventually saturates as the analyte concentration increases (Figure 2c).This decrease comes from the narrowing of the depletion region W dep due to absorption of polarized molecules that are aligned parallel to the electric field near surface space charge region.The narrowing of surface depletion region will widen the conduction channel and lead to a higher conductance.This will be revealed more explicitly in the gas conductance measurements in the following section.Figure 2d depicts the change of depletion region ΔW dep that is calculated from the Hall measurements in Equation ( 4) with ) .Clearly, the correlation of ΔW dep ∼ C e is well fitted with Equation (3) from which RT bqN a and C e0 are extracted as 9.75 nm and 538 ppm, respectively.After plugging those previously known parameters (R, T, and N a ), the intrinsic coefficient  b is calculated to be 6.11 × 10 −3 C mol J −1 cm −2 .

Gas Conductance Measurements
Different from Hall measurements, the conductance of Si wires is modulated by both sidewall and top depletion regions.In practical applications, the conductance is readily attained without the presence of magnetic field, thus is more convenient for gas sensing.As demonstrated before, the narrowing of depletion region will be explicitly reflected in the increase of conductance.In experiments, a dc voltage is applied through pads A and C to provide a controllable current in the channel.The channel conductance  can then be extracted from the slope of current-voltage (between pads D and E) curves.Figure 3a presents transient current response of a 280 nm wide p-type silicon nanowire in a periodic ON/OFF flux of ethanol vapor with an increasing concentration.The results for water and acetone vapors can be found in Section S4 (Supporting Information).The periodicity is fixed at ≈1000 s although the rise time at low concentration may be longer.The responses in current are taken from the difference between the maximum and background current in each cycle.The gas response (R) of the nanowire gas sensors is defined as where Δ is the conductance variation under the ON/OFF exposure of vapor concentration C e , which can be further written as  (see Section S5 in the Supporting Information for derivations) The response of Si wire gas sensors shows reasonably good response-recovery properties, as the baseline changes within less than 5% (indicating a sufficiently aged gas sensor).In addition, the device responds more readily to a high concentration of ethanol vapors (as much as 3 × 10 4 ppm).Interestingly, both p-type and n-type (see Section S6 in the Supporting Information) Si wires have positive responses of current under ethanol vapors exposure.This observation is consistent with the aforementioned  sensing mechanism (dipole-induced narrowing of surface depletion region).Following such a mechanism, our wires would show similar positive responses to all polar analyte molecules.
In Figure 3b, electrical response measurements were performed for three gaseous analytes, including ethanol, water, and acetone vapors.Those three analytes induce positive responses of p-type wires, which means that the three analytes always provide negative dipoles for p-type wires.The negative dipoles on the surfaces lower surface potentials and induce the narrowing of surface depletion, resulting in a higher conductance compared to wires under nitrogen exposure.Note that our model may become invalid when the concentration of analytes is too high so that the depletion region narrows down to zero.In this case, the gas sensor responses often become saturated.Clearly, the responses of our gas sensors are not saturated and our model should be valid.As expected, the responses of the three gaseous analytes fit well with the analytical formula, indicating that the Temin isotherm model is suitable for physical adsorption with gas-solid interaction.From the fitting,  b of water, acetone, and ethanol vapors are extracted as 3.7 × 10 −4 , 1.9 × 10 −3 , and 3.9 × 10 −3 C mol J −1 cm −2 , with aging concentrations C 0 being 1621, 197, and 1203 ppm, respectively.The extracted parameters are summarized in Table 1.

Note that Si NWs have the poorest responses and smallest 𝛾
b for water molecules, although water molecules have a higher polarity (a higher ) than both ethanol and acetone. [43]Therefore, it is not unreasonable to infer that the adsorption heat b for H 2 O is also higher than other two.A larger adsorption heat b will lead to a lower concentration of molecules adsorbed according to our discussions later and Section S7 (Supporting Information).
Due to the noise in current, the gas sensor has a detection limit for each gaseous analyte.Below this concentration limit, gas response signals will be buried in noises.The detection limit thus is calculated as the threshold when the signal-to-noise ratio equals 1 (in Figure 3a).Following this argument, the Si wire gas sensor is sensitive to acetone molecules with a detection limit as low as 282 ppm.For water and ethanol vapors, it can only sense a concentration above ≈2007 or 1273 ppm, respectively.According to Equation ( 5), when the wire width decreases to a value slightly larger than the two times of W dep (the channel is continuous but close to pinch off), the gas sensor response will dramatically increase (consistent with the data in Figure 4a) and significantly lower the detection limit.But a completely pinched-off nanowire (W < 2W dep ) becomes nearly insensitive to gaseous analytes (data not shown).
The sensitivity is defined as the response under the exposure of a unit gas concentration, that is, S = R C e which can be rewritten as Equation ( 6) in which the maximum sensitivity is expressed as . The term in the square brackets of Equation ( 6) will reach 1 when the analyte concentration C e approaches zero, maximizing the sensitivity to S max .It is clear that a smaller conductance (but with a continuous channel) and a lower aging concentration C 0 will lead to a higher gas sensitivity.10][11][12][13][14][15][16][17][18][19][20] Figure 3c depicts the experimental sensitivities.Overall, the sensitivity decreases at higher analyte concentration.The maximum sensitivity S max is calculated in Table 1 from the parameters extracted by fitting Equation (5) to the data in Figure 3b, although it can also be found by fitting Equation ( 6) to the data in Figure 3c.However, the experimental data in Figure 3c are dominated by the high sensitivities at low concentrations of gaseous vapor which have a large uncertainty, leading to less reliable fitting results.For this reason, we use Equation ( 5) to fit the related experimental data for parameter extraction.
Equation (5) indicates that the gas response is a function of nanowire dimensions.To further validate Equation ( 5), we used another set of nanowires with different sizes to detect ethanol vapors.Since the nanowire dimensions are given, the fitting of Equation ( 5) to the experimental data in Figure 4a will render a term  b independent of nanowire size.As expected in Figure 4b (black), the fitting results indicate that as the nanowire width increases from 120 to 280 nm, the term  b remains largely stable around 5.2 × 10 −3 C mol J −1 cm −2 , which is in line with what we found from gas Hall measurements.The aging concentration C 0 (Figure 4b, red) does not vary much among nanowires with different sizes either.The size independence of  b and C 0 is consistent with the nature of these parameters.
The gas response in Equation ( 5) is also a function of temperature.Figure 4c shows the temperature-dependent responses to ethanol gaseous vapors.The results fit well with the derived analytical formula (Equation ( 5)).The extracted  b values are plotted in Figure 4d.It turns out that  b increases as temperature rises.It is likely because the adsorption heat decreases at a high temperature due to a weaker adsorption interaction at an elevated temperature (molecules become less adsorptive to the surfaces).On the other hand, C 0 tends to increase with temperature.Note that a weak adsorption of molecules onto surfaces (smaller adsorption heat) does not imply a lower concentration of molecules attached to surfaces (adsorption capacity).In fact, according to the initial Temkin isotherm model in Equation ( 1), when b is smaller, the adsorption capacity will be higher, which is an implication of two basic assumptions of the Temkin model. [44]The first assumption is that the rate of adsorption/desorption varies exponentially with surface occupation; the second assumption is that the adsorption heat/entropy increases linearly with surface occupation.See detailed illustrations in Section S7 (Supporting Information).Besides, the response behavior of NWs is also influenced by aging concentration C 0 , which also increases with temperature due to a similar reason.Lastly, we would like to point out that our analytical gas response formula in Equation ( 5) can be also applicable to well fit the gas responses of other nanowire sensors in literature (see Section S8 in the supporting Information).It provides additional strong evidence to validate the formula in Equation (5).

Conclusion
In conclusion, we established an analytical gas-response equation for single-crystalline semiconducting gas sensors based on the widely accepted sensing mechanism, which was validated with gas Hall measurements in this work.This analytical equation fits well with experimental data.The extracted parameters are largely consistent between independent gas Hall and gas conduction measurements under various conditions.The analytical gas sensitivity can also be derived, which indicates that a smaller nanowire will render a higher sensitivity, consistent with the documented experimental observations.With this equation, the gas responses of solid-state chemical sensors can be quantitatively designed.Although this equation is derived on Si wire gas sensors, it may be universal for all isotropic and covalently bonded single-crystalline semiconducting gas sensors.

Experimental Section
Gas Path System: The experimental gas atmosphere was supplied by a gas path system.One mass controller (HORIBA, STEC S48 300) regulated the flow of high-purity nitrogen as the background atmosphere.The other line was bubbled through the high-purity liquid ethanol, providing a known saturated vapor flow at room temperature.The gas path system provided a total gas circulation of 500 mL min −1 .By controlling the flow ratio between two mass controllers, the tunable gas concentration at steady pressure was set for response chamber The electrical response was measured in a small designed chamber (1.6 cm × 2 cm × 1 cm), which could be applied as a plug-in for PPMS system.The foundation of measurement chamber was a sample holder with integrated circuit.The sample holder was compatible with physical property measurement system (PPMS, Quantum Design, USA), enabling Hall effect measurements and four-probe measurements at tunable gas concentration.Therefore, the in situ mobility, carrier concentration, electrical current and Hall resistance under certain gas concentration were measured at the same time.
Hall Effect Measurements: The Hall effect was tested with a comprehensive physical property measurement system (PPMS, Quantum Design, USA).The samples were preinstalled on the holder by wire bonding before the test.The maximum current was limited to 100 A, and the magnetic field range was from −1 to +1 T. The gas path system was connected with PPMS Multi-Function Probe (PFB), and continuously fed with ethanol vapor gas for response measurements.The Hall effect measurements were performed under different ethanol partial pressure for monitoring the depletion region width and Hall responses.In addition, the Hall voltage was measured under continuous periodic ethanol partial pressure.The ethanol gas concentration ranged from 4% to 32% of the saturated vapor pressure at ambient condition.
Gas Response Measurements: The fabricated nanowire devices were characterized by high-precision digital source meters (Keithley 2400) on a customized PCB board on which a digital signal source (National Instrument USB-6009) controls a multiplexer (TMUX6208) to scan through several nanowires in each cycle of measurements.
Temperature-Dependent Responses: The PPMS external heater was applied to measure the 280 nm wide Si wire conductance under various temperatures ranging from 290 to 400 K.The measured conductance was used to calibrate the temperatures derived from Joule heating.The 280 nm wide Si wire was electrically biased in a range from 1 to 11 V, corresponding to 301-362 K self-heated temperature.Then the response measurements were performed under voltages ranging from 1 to 11 V.The calibrated temperatures were used to characterize the adsorption process on the Si wire surfaces.
Si Wire Fabrication: The SOI wafers were first cleaned with acetone and deionized (DI) water.Boron ions were then implanted into the device layer of the SOI wafer at an implantation energy of 30 keV and a dose of 2.2 × 10 13 cm −2 .The doping peak was located at 110 nm with a maximum concentration of ≈1 × 10 18 cm −3 .After the implantation, rapid thermal annealing (RTA) was employed to activate the boron atoms at 1000 °C for 20 s.Afterward, the silicon wafers were cut into small pieces (≈1 × 1 cm 2 ).Each piece was cleaned with piranha solution (98% H 2 SO 4 :30% H 2 O 2 , 3:1 (v/v)).A 250 nm thick layer of poly(methyl methacrylate) (PMMA) resist (XR-1541-006, Dow Corning Electronics, USA) was spin-coated on the p-type SOI samples at 4000 rpm for 60 s.Then, the wafers were baked at 180 °C for 90 s.The PMMA resist was exposed by electron beam lithography (Vistec EPBG5200) and subsequently developed in Methyl isobutyl ketone and isopropyl alcohol.PMMA is a positive electron beam resist.After that, aluminum (Al) was evaporated to the exposed region.To form etch mask for the micropads, NR9-1500PY (Futurrex Inc.USA) photoresist was coated on the wafers at 4000 rpm for 40 s.After baked at 140 °C for 60 s, the NR9 resist was exposed to UV light (MDA-400) and developed in the developer after postbaking at 110 °C for 60 s.A 200 nm thick Al film was evaporated (Thermal Evaporator, Angstrom Engineering), followed by a liftoff process.The Al nanowire and micropad patterns defined by electron beam exposure and photolithography were then transferred to the SOI device layer by reactive ion etching (RIE, Sentech ICP Reactive Ion Etching System), forming an array of silicon nanowires with each connecting to five silicon pads for metal contacts.After photo-lithography, 20 nm thick cobalt and 180 nm thick aluminum were evaporated onto the Si pads as electrodes.To form good Ohmic contacts, the wafers were then annealed in argon atmosphere for 15 min at 230 °C.
Plasma Processing: The oxygen plasma was adopted as a postprocessing technique for silicon nanowire surface functionalization.By utilizing oxygen microwave plasma (PVA-Tepla Microwave Stripper/Plasma processing systems) under the atmosphere of oxygen (350 sccm) for a duration of 2 min with a power of 200 W, surface charge density of as-etched silicon nanowires were modified.The positive charge at p-type nanowires surface are partially compensated by oxygen plasma induced ions (O − , O 2 − , etc.).Therefore, the silicon nanowire gas sensors become rapidly responding and recoverable after released from strong inversion situation.
Statistical Analysis: The data obtained in gas Hall measurements were Hall voltage and converted into Hall resistance by dividing the current.The data obtained in gas conductance measurements were channel current and converted into channel conductance by dividing the voltage provided.The gas response was calculated as the conductance change normalized by channel conductance with only nitrogen.The concentrations of testing gases were converted into ppm based on their saturated vapor pressure (water: 23.8 mmHg, acetone: 232 mmHg, and ethanol: 67.0 mmHg) and the standard atmospheric pressure (760 mmHg) by C e = 23.8(water) 760 (atmospheric) × 10 6 × 50% (percentage of water gas flow line).All data analysis and fittings were performed using software ORIGIN with their errors represented by x ± standard deviation (SD.

Figure 1 .
Figure 1.a) False color SEM image of a Si nanowire gas sensors (NW width: 200 nm).b) Four-probe IV characteristics of Si wires with a width from 50 to 800 nm.Inset: IV curves for 50 (black) and 80 nm (red) wide nanowires.c) Conductance of p-type silicon wires and linear fitting curve before (black) and after (red) oxygen plasma processing.Inset: IV curve for 400 nm wide wire before and after oxygen plasma.d) Simulated surface electron concentrations (left) and surface depletion region width (right) for silicon (p-type, N a = 1 × 10 18 cm −3 ) with different surface charge densities Q s /q.Inset: schematic diagram of a silicon wire with surface depletion region induced by surface charges, and energy band bending.

Figure 2 .
Figure 2. In situ gas Hall measurements for a 100 nm wide silicon nanowire.a) Measured Hall voltage response upon the ON/OFF exposure to ethanol vapor with an increasing concentration.Inset: schematic setup of standard Hall measurements under ethanol vapor exposure.b) Hall resistance R H versus magnetic field curves under the exposure of pure nitrogen (black) and 21 158 ppm ethanol vapor (red).c) Extracted slopes of R H −B curves (inset) for SiNWs under the ON/OFF exposure of ethanol vapor with different concentrations.Error bar shows the standard error of linear fit of Hall resistance versus magnetic field curves for different ethanol concentrations.d) Depletion region narrowing under different ethanol concentrations extracted from slopes in panel (c) and the fitting using Equation (3).

Figure 3 .
Figure 3. Gas responses of Si wire sensors.a) Measured real-time current response of a 280 nm wide Si wire gas sensor under increasing ethanol concentrations.Shaded gray regions represent the background fluctuation.Blue arrow indicates the signal current.b) Gas responses under various concentrations of water (black), acetone (red), and ethanol (blue) vapors diluted with nitrogen atmospheres.c) Gas sensitivity calculated from panel (b).

Figure 4 .
Figure 4. Gas responses of SiNWs with different widths at variable temperatures.a) Measured ethanol response curves for SiNWs with different widths under 300 K. b) Extracted  b from panel (a) for NWs with different widths.The gray dashed line is to guide the eye.c) Measured ethanol response curves of the 300 nm wide SiNW gas sensors under various self-heated temperatures.d) Extracted  b from panel (c) at each temperature.

Table 1 .
Summarized response parameters of a 200 nm SiNW gas sensor for water, acetone, and ethanol vapor gas.