A Highly Selective and Self-Powered Gas Sensor Via Organic Surface Functionalization of p-Si/n-ZnO Diodes

these purely inorganic sensors cannot accomplish a meaningful gas selectivity. High selectivities toward a single gas species have been recently reported via modifying the inorganic surface of nanostructured semiconductors with a deﬁ ned organic functionality.


DOI: 10.1002/adma.201403073
interactions, these purely inorganic sensors cannot accomplish a meaningful gas selectivity. [ 4,5 ] High selectivities toward a single gas species have been recently reported via modifying the inorganic surface of nanostructured semiconductors with a defi ned organic functionality. [6][7][8][9] Theoretical simulations based on ab initio density functional theory (DFT) for a system composed of SnO 2 nanowires (NWs) modifi ed with a defi ned selfassembled monolayer (SAM) elucidated the reason for the high selectivity of such gas sensors: the energetic position of the SAM-gas frontier orbitals with respect to the NW Fermi level has been identifi ed to be the crucial factor to ensure an effi cient charge transfer upon gas-SAM binding interactions and thus to sense or discriminate a certain gas species. [ 7 ] The high fl exibility of organic surface modifi cations in terms of functional groups, as well as their steric and electronic structures might possibly enable the targeted design of various specifi c gas sensors. However, all organic surface modifi ed sensor systems so far are based on compact conductometric or fi eld-effect transistor (FET) sensor concepts that still require a remarkable amount of energy to generate a sensor signal (e.g., by applying a source-drain current). To date, none of the semiconductorbased gas sensor systems have been able to accomplish both self-powered/low-powered sensor operation and highly selective gas detection within a single and compact device.
In this work, we present a semiconductor-based gas sensor concept that combines the two substantial requirements of mobile gas sensing in a singular sensor device: self-powered operation combined with high gas selectivity. Beyond the combination of self-powered sensing and high selectivity, a high sensitivity could also be demonstrated for the example target gas, nitrogen dioxide (NO 2 ). The sensor reported here was capable of detecting NO 2 concentrations at the sub-ppm level with a very high selectivity, which is a crucial detection region, since NO 2 is considered to be one of the major threats to human health and is already toxic at very low concentrations (ppb level).
[ 10 ] NO 2 is released in urban and industrial environments by various combustion processes and can produce ozone, cause respiratory ailments, [ 11 ] and is suspected of causing cancer. [ 12 ] For all these, the technology presented here could demonstrate an approach to much-sought-after, widespread sensor networks for monitoring densely populated areas to improve the ambient air quality. Figure 1 a schematically shows the production process of our self-powered and selective sensor device based on nanostructured p-Si/n-ZnO diodes. After patterning of a p-Si layer on SiO 2 via reactive ion etching (RIE), photolithographic methods Low power consumption and reliable selectivity are the two main requirements for gas sensors to be applicable in mobile devices. [ 1 ] These technological platforms, e.g., smart phones or wireless sensor platforms, will facilitate personalized detection of environmental and health conditions, hence becoming the basis of the future core technology of ubiquitous sensing. Even today, health control as well as environmental monitoring relies on immobile and complex detection systems with very limited availability in space and time. Recent work has shown promising concepts to realize self-powered gas sensors that are capable of detecting gases without the need of external power sources to activate the sensor-gas interaction or to actively generate a read-out signal. [ 2,3 ] These sensors drastically reduce power consumption compared with conventional semiconductor gas sensors and additionally reduce the required space for integration. All the attempts so far have been based purely on nanostructured inorganic metal oxide sensor materials that provide a good sensitivity toward different gases due to their high surface-to-volume ratio. However, due to their nonselective sensing mechanism based on oxygen-vacancy-gas COMMUNICATION were used to deposit layers of 20 nm of ZnO selectively on the p-Si sidewalls that served as a seed layer for site selective growth of n-ZnO NWs, to form p-Si/n-ZnO heterojunctions (Figure 1 b,c). X-ray diffraction spectra revealed a good crystallinity, and the disproportionate peak attributed to the ZnO (002) facet indicated an oriented growth direction along the c -axis ( Figure S1, Supporting Information). Under illumination by solar light (AM1. 5), an open-circuit voltage ( V oc ) is induced by the built-in potential at the p-n heterojunction, [ 2,13,14 ] which is used as a self-generated signal for the sensor. To ensure a strong and measurable V oc signal, even under low illumination conditions, every device contained several diodes connected in series by way of evaporated gold contacts (9, 16, or 26 diodes; see Figure 1 c). This confi guration resulted in a linear increase of the measured V oc signal, proportional to the number of diodes in series under light, with an average contribution of V oc,exp = 72 ± 6 mV/diode (typical value for p-Si/n-ZnO diodes) [ 2 ] and leads to measured V oc values of 0.65 V (9 diodes), 1.35 V (16 diodes), and 1.84 V (26 diodes) respectively (see Figure 1 c and Figure S2 in the Supporting Information). Figure 1 d shows the band diagram of the Au/p-Si/n-ZnO/Au system in equilibrium (see Supporting Information for further details). While the bandgap of ZnO ( E g,ZnO = 3.37 eV) makes it unsuitable for photovoltaic conversion in the visible range, only the Si serves as a photon absorber. By design, the depleted region of the heterojunction (depletion width X = 0.646 µm) mainly extends within the p-Si ( X Si = 0.637 µm), according to the impurity concentrations of the materials used here ( N a = 1.355 × 10 15 cm −3 for p-Si) and assuming a typical value of N d = 1 × 10 17 cm −3 for n-ZnO produced by this method. [ 15,16 ] Under these conditions, the theoretical V oc in an ideal heterojunction has been estimated to be V oc,theor,ideal = 479 mV/diode, which relates to the built-in potential formed between the ZnO conduction band and silicon valence band under non-biased conditions (see Supporting Information). The discrepancy with our experimentally observed values ( V oc,exp = 72 ± 6 mV/diode) can be explained by a signifi cant current leakage due to defectinduced carrier recombination at the p-n interface. This effect is typically found in Si-ZnO junctions owing to the large lattice mismatch [ 17,18 ] and sputter effects during the dry etching process of silicon. In fact, simulations (see Figure S4a in the Supporting Information) predict V oc,theor,leakage = 68 mV/diode considering a shunt resistance of around 30 Ω cm 2 to account for these effects (to be compared with values around 1 kΩ cm 2 for a commercial solar cell). [ 19 ] Furthermore, the band diagram shows that the back contact is not Ohmic, as the ZnO bands bend 0.59 eV at the ZnO/Au interface. This limits the forward current of the p-Si/n-ZnO structure but will not substantially affect the V oc , as no current is fl owing under open-circuit conditions.
For the organic modifi cation of the n-ZnO NWs, two different methoxysilanes were used to form SAMs on the surface of the n-ZnO NWs, owning amine ([3-(2-aminoethylamino)propyl]-trimethoxylsilane ( 1 )) and thiol ((3-mercapto-propyl)trimethoxysilane ( 2 )) functionalities respectively. The successful immobilization of the SAMs terminated with amine 1 and thiol 2 on ZnO NWs was investigated via high-resolution Auger spectroscopy (spatial resolution < 20 nm). Characteristic carbon and oxygen signals were detected for both devices. Nitrogen and sulfur signals for the amine-and thiol-functionalized devices respectively documented the as expected organic surface character of the sensor devices (see Figure S5, Supporting Information).
To measure the sensing characteristics of the devices functionalized with amine-and thiol-SAMs, the respective substrates were placed in a sensing chamber and illuminated through a quartz window to generate the monitored V oc signal. All the measurements were performed with samples composed of 16 diodes in series, as they offered the best compromise in terms of signal value and noise level. The test gases were diluted in dry synthetic air (blend of 21% O 2 and 79% N 2 ) and applied via a mass fl ow controller system into the sensing chamber.
where V oc,gas is the measured sensor signal and V oc,air the baseline value measured in pure synthetic air atmosphere. Even the lowest value of 250 ppb could be detected with a signal response of Δ V oc = 7.5% and the V oc increased by 23.5% with the introduction of 750 ppb NO 2 (see Figure 2 a and c). For all concentrations, a full signal recovery was observed after switching back to pure synthetic air (typical recovery times were about 5 min). The results prove that the device is capable of detecting NO 2 in selfpowered operation at low ppb values, which is the above-mentioned critical range for environmental monitoring.
[ 10 ] From the observed linear sensing characteristics of the amine-functionalized sensor, the detection limit for NO 2 can be approximated as [NO 2 ] min = 170 ppb with a signal-to-noise ratio of 2. In sharp contrast to the amine-modifi ed device, the response of the thiolmodifi ed device was observed to be negative. An even higher absolute response of Δ V oc = −9.8% toward 250 ppb of NO 2 was achieved for this device. In any case, the increase of the sensor response for the thiol-modifi ed samples did not proceed linearly with the NO 2 concentration, and reached a saturation value of Δ V oc = −12.8% for 500 and 750 ppb. This characteristic is suitable for highly sensitive NO 2 detection, whereas a quantitative sensing could not be realized at concentrations >250 ppb by using the thiol SAM. Remarkably, the inverted sensor signals for the two different organic functionalities prove that the critical gas-surface interactions of the sensor device, and consequent electrical modulations are guided by the nature of the SAM molecules. The fact that no response toward NO 2 was observed for devices without SAM modifi cation further indicates that the sensor-gas interaction is solely subsiding by SAM-gas interactions in the case of the modifi ed samples (see  Figure S6, Supporting Information), as the incident light is not suffi ciently activating direct interactions between the inorganic ZnO and the surrounding gas atmosphere. [ 2 ] Since illumination with light in the visible range (e.g., natural sunlight) is sufficient to facilitate gas-desorption processes for fast response and recovery processes of SAM modifi ed gas sensors, [ 7 ] the organic groups furthermore ensure a sensing operation without the need of external thermal or UV-light energy that is usually required for the activation of the metal oxide surface of traditional gas sensor systems and consume a major fraction of the total required energy in conventional sensor systems.

COMMUNICATION
In order to assess the selective sensing capabilities of our devices, we compared the response toward NO 2 with that of common interfering and combustion gases (SO 2 , NH 3 , and CO) of signifi cantly higher concentrations (2.5-25 ppm). Figure 3 a,b show that the response of the amine-functionalized sensor toward all the referenced gases is signifi cantly lower than in case of the presence of low NO 2 concentrations. For 2.5 ppm of SO 2 , a measurable response of Δ V oc = 5.0% was observed. The response per ppm of 2.0%/ppm was remarkably lower than was observed for NO 2 (31.3%/ppm). The responses of the amine-terminated sensor toward 25 ppm of NH 3 and CO were just above the noise level with Δ V oc = 1% (0.04%/ppm). The same measurements for the thiol-modifi ed sensor revealed a relatively low negative response of Δ V oc = −3.4% toward 2.5 ppm of SO 2 (−1.35%/ppm) and an increase of the measured V oc by 4.5%, when 25 ppm of NH 3 were introduced (0.18%/ppm) (see Figure 3c,d, and Figure S7 in the Supporting Information). No signal change was observed after the introduction of 25 ppm of CO. The high response toward low concentrations of NO 2 (−39%/ppm for 250 ppb; −17.1%/ppm for 750 ppb) and a lower signal noise qualifi es the thiol-modifi ed sensor as being suitable for the selective detection of tiny NO 2 concentrations. Nevertheless, the quantitative sensing ability and a higher response at concentrations above 500 ppb, combined with a very high discrimination of other interfering gas species qualifi es the aminefunctionalized sensor as being more adaptive in real applications. Besides these promising fi ndings, the stability of the functionalized sensors is still challenging. After exposing the sensors to sequential pulses of NO 2 for about 8 h under constant illumination, the sensing response was found to be lowered to about 80% of the initial values. This effect could be attributed to photocatalytic cleavage caused by UV-induced charge transfer from the covalently bound semiconductor substrate or directly UV-induced degradation of the organic SAM. [ 20 ] Therefore, our ongoing investigations are focusing on the understanding of such degradation processes as well as their suppression via less harmful operation conditions, e.g., by restricting the incident light spectrum to the visible spectrum. Such illumination energy would be suffi cient to facilitate the adsorption/desorption processes of the sensor and still could be absorbed by the p-Si/n-ZnO junction to generate the sensor signal. Thus, higher illumination intensities could be used to improve the moderate dynamic performance of the sensors, just following the strategy observed in a previous report. [ 7 ] In order to better understand the SAM-gas binding interactions and to evaluate variations of the sensing characteristics between amine-and thiol-terminated species, theoretical DFT calculations were performed. In order to determine the moststable binding geometries of NO 2 for both the amine and thiol groups, 115 different SAM-NO 2 confi gurations with random initial positions and orientations of the NO 2 molecule were relaxed ( Figure S8, Supporting Information). The thus-obtained most-stable binding geometries of NO 2 with both the amine and the thiol SAMs are illustrated in Figure 4 a. For the amine SAM, NO 2 preferably adsorbs near the secondary amine group, where the highest occupied molecular orbital (HOMO) of the SAM is located. For the most-stable amine-NO 2 geometry, the binding energy E B amounts to −0.44 eV and a considerably high covalent character of the NO 2 -SAM binding was observed.
Here, more-negative values of E B mean stronger binding.
In the case of the thiol SAMs the thiol-NO 2 binding energy is found to be considerably weaker ( E B = −0.22 eV). For the most-stable geometry, NO 2 adsorbs near the sulfur atom, where the HOMO of the thiol SAM is located, and additionally forms a weak hydrogen bond with one of the OH groups (Figure 4 a). For the thiol-NO 2 system, some degree of hybridization of the SAM and gas frontier orbitals was also observed, which indicated a considerable covalent binding character ( Figure S8b, Supporting Information). However, dispersion forces here play a more-dominant role as compared with the amine case; hence, the portion of the covalent bonding interaction is lower compared with that of the amine-NO 2 system (see the Supporting Information for further explanation).
Based on these SAM-NO 2 binding geometries, we investigated the positions of the energy levels of the SAM and the most-stable SAM-NO 2 systems to get an idea about the change Adv. Mater. 2014, 26, 8017-8022 www.advmat.de www.MaterialsViews.com of the electronic structure of the functional groups due to gas adsorption and the differences between the thiol and amine systems with respect to the affi nity to transfer charges between the SAM-gas system and the ZnO NWs. Since the bonding of the amine and the thiol SAM to the ZnO NWs in each case most likely occurs via the silicon atom, which, in both cases, has the same chemical environment locally, we chose the average electrostatic potential in the core region of the Si atom of each SAM as a common energy reference. Interestingly, we found that NO 2 adsorption leads to a downward shift of the HOMO by 0.52 eV in the thiol case and, in sharp contrast, an upward shift of 0.13 eV in the amine case (Figure 4 b). For both systems the LUMO level of the SAM-gas system is much lower than in the NO 2free case. This low LUMO level is mainly derived from the NO 2 LUMO. The DFT calculations thus show that the electronic structure of the SAM-NO 2 system indeed depends on the specifi c choice of the SAM and is strongly altered by NO 2 adsorption (shift of the HOMO and LUMO energies). Moreover, the relatively low position of the thiol-NO 2 LUMO (high electron affi nity (EA) of the system) and the relatively high position of the amine-NO 2 HOMO (low ionization potential (IP) of the system) indicate a predominantly electron-acceptor character for the thiol-NO 2 system and a predominantly electron-donor character for the amine-NO 2 system. [ 21,22 ] This is consistent with the experimentally observed reversed signals of V oc of the thiol-and amine-functionalized NWs. This relative change of the electron donor and acceptor character of the organic functionalities upon NO 2 binding (change of HOMO and LUMO energies) could directly infl uence the electronic structure of the SAM/n-ZnO system and thus, macroscopically result in a change of the sensor signal (Δ V oc ). However, further studies are needed to better understand the mechanism of the consequent interactions at the SAM-NW interface and the resulting change within the electronic n-ZnO structure.
In conclusion, we have demonstrated a selective and selfpowered gas sensor, capable of detecting low NO 2 concentrations in the ppb range without the need of an external power source. The sensor signal ( V oc ) was generated by microfabricated p-Si/n-ZnO diodes upon visible-light illumination. The selective sensing qualities were introduced by the functionalization of the n-ZnO surface with amine-as well as thiol-terminated organic SAMs. Furthermore, the use of an organic SAM facilitated the gas-surface interaction without the need of heat or UV activation, as is required for bare inorganic gas sensors. Detailed DFT simulations of the SAM-NO 2 binding interactions and subsequent changes of the organic surface group frontier molecular orbitals indicated that the nature of the chemical SAM structure directly determines the gas response of the hybrid material. The contrary relative changes of the ionization potential (IP) and electron affi nity (EA) upon NO 2 binding for amine-and thiol-terminated SAMs correlate well with the experimentally observed sensing results. Therefore, this work gives an insight into the complex sensing mechanism of inorganic-organic hybrid gas sensors and shows the feasibility of transferring chemical signals from specifi c organic-gas interactions into active electronic signals solely driven by visible light.

Experimental Section
Preparation of Series p-Si/n-ZnO Diodes : The p-Si strips on SiO 2 were prepared by photolithographic patterning of a silicon-on-insulator (SOI) wafer (boron-doped, R = 1-20 Ω cm) and subsequent reactive ion etching (RIE) using a mixture of SF 6 /O 2 (129/9 sccm) as the process gas ( T = −95 °C, p = 1.5 Pa). For the site selective growth of the ZnO NWs, the etched substrate was again photolithographically patterned, and a 20 nm Zn fi lm was deposited via DC sputtering (99.9999% Zn target, Ar), followed by an annealing at 300 °C in air to form a ZnO seed layer. ZnO NWs were grown by immersing the seeded substrate into a 1:1 solution of zinc nitrate (Zn(NO 3 ) 2 ·6H 2 O; c = 0.025 mol/L) and hexamethylenetetramine ( c = 0.025 mol/L) and stirring the solution for 3 h at 90 °C. After refreshing the solution, the substrate was stirred for another 3 h at 90 °C. Subsequently, the substrate was removed from the solution, sonicated for 5 s and rinsed for 30 s in deionized water.
Adv. Mater. 2014, 26, 8017-8022 www.advmat.de www.MaterialsViews.com Figure 4. a) Most-stable geometries of NO 2 molecules adsorbed at the thiol and amine functionalities. The charge densities of the HOMO wavefunctions are shown as green isosurfaces. The isosurfaces are drawn at a value of 0.075 e/Å 3 . In both cases, the HOMOs are hybrids of the HOMO of the organic functionality and the NO 2 LUMO. In the amine case, strong mixing between the amine HOMO and the NO 2 LUMO occurs, whereas in the thiol case, the HOMO of the combined system is dominated by the thiol HOMO. b) Energy levels of the HOMO and LUMO of the functionalities without adsorbed NO 2 and with adsorbed NO 2 . The averaged electrostatic potential from the core region of the Si atom of each functionality was used as reference potential. The HOMO level of the isolated thiol functionality was set to 0 for convenience.