Introduction

Catalytic surface functionalization of nanostructured metal oxides presents a promising and versatile approach across various chemical research fields, including heterogeneous catalysts1,2,3 and catalysis-based sensors4,5. Leveraging this surface functionalization, researchers have extensively explored catalysis-based electrical sensing on nanostructured metal oxide surfaces for various volatile organic compounds (VOCs)4,5,6. Catalysis-based molecular sensors have attracted considerable attention because of their ability to tailor sensitivity and selectivity via catalytic events on sensor surfaces, which is challenging to achieve with bare metal oxide sensor surfaces5. Among the numerous surface functionalization methods for nanostructured metal oxides, organic molecular surface modification is among the most facile approaches, leveraging intermolecular interactions and affinity between target analytes and surface-decorated molecules7,8,9. The effects of such organic molecular surface modification on molecular selective sensing data have typically been interpreted in terms of attractive or repulsive intermolecular interactions between analytes and modified organics10,11,12. For example, (3-aminopropyl)trimethoxysilane (APTMS) self-assembled monolayer (SAM)-modified multi-ZnO nanowires sensors have demonstrated enhanced sensitivity for acetone by modulating both the depletion layer thickness on semiconducting ZnO surfaces and the electric interactions between the -NH2 groups of APTMS SAMs and C = O of acetone13.

However, considering the aim of catalysis-based metal oxide sensor surfaces, which is to maintain and/or increase the surface density of catalytic sites of metal oxides4,5, organic molecular surface modification presents a clear deviation from this general design principle based on catalysis-based mechanisms. This is because organic surface modifications inherently reduce catalytic site density on metal oxide surfaces. On the other hand, recent studies on inhomogeneous catalysts have revealed that organic molecular surface modification can direct the pathway of catalytic reactions on metal oxide surfaces14,15. For example, electronic modulations of oxide surfaces using SAMs have been shown to influence the dehydration activity of alcohols by regulating transition-state structures and activation energies16,17,18. In catalytic processes on metal oxide surfaces and in catalysis-based electrical sensing of VOCs, the “catalyst deactivation effect” is an inherent and inevitable issue. This effect occurs when catalytic sites are deactivated by residual analytes and/or catalyzed products present on the surface during operations, thereby significantly deteriorating catalytic performance19,20,21. While proposals to mitigate these surface deactivation effects using UV radiation or high-temperature annealing have been made22,23,24, such harsh and destructive surface treatments are clearly unsuitable for catalyst surfaces modified with organic materials25,26. Among various surface chemical species, carboxylates are particularly important for two primary reasons. Firstly, carboxylates are highly common species in chemical sensors and heterogeneous catalysis because organic compounds composed of carbon, hydrogen, and oxygen are ultimately oxidized into carboxylic acids in oxidative ambient environments. Secondly, carboxylates form stronger bonding to typical metal oxide catalysts, resulting in significant catalyst deactivation by blocking active catalytic sites on the metal oxide surface.

Herein, we demonstrate that weak van der Waals interactions between hydrophobic aliphatic alkyl chains and hydrophilic ZnO surfaces, which have been previously underestimated during catalytic molecular sensing, prevent catalyst deactivation in the electrical sensing of aliphatic aldehydes on ZnO sensor surfaces. Surface modifications with aliphatic phosphonic acids (octadecylphosphonic acid-ODPA) on ZnO unexpectedly enhanced catalytic aldehyde (nonanal) electrical sensing by suppressing catalyst deactivation. Electrical sensing measurements revealed that ODPA surface modification significantly reduced the recovery time of the sensor response to nonanal molecules without compromising sensitivity. Spectroscopic measurements revealed that nonanal molecules directly coordinate with surface Zn ions as oxidized carboxylates by penetrating the trans-zigzag ODPA SAMs. Temperature-dependent spectroscopic analysis revealed that the desorption temperature of carboxylates as reaction products on ODPA-modified ZnO surfaces was substantially reduced to less than 150 °C, whereas carboxylates on bare ZnO nanowires remained above 300 °C, highlighting the significantly reduced catalyst deactivation effect. IR p-polarized multiple-angle incidence resolution spectroscopy (pMAIRS) was employed on deuterated SAMs to successfully observe the changes in the conformation and orientation of the alkyl chains of ODPA caused by nonanal penetration and adsorption.

Results and Discussion

Figure 1a and b shows schematic of the bare and ODPA-modified single-crystalline ZnO nanowire sensors fabricated in this study. Figure 1c and d show the effect of surface modification on the resistance changes of both ZnO nanowire sensors under cyclic exposure to volatile aldehyde (nonanal as a biomarker in human breath27,28) vapor and purified air. The concentration of nonanal vapor was 32.7 ppm, as determined by gas chromatography-mass spectrometry (GC-MS) measurements (Supplementary Fig. 1), and the operating temperature for sensing was maintained at 160 °C. When nonanal vapor was introduced, the resistance consistently decreased for both the bare and ODPA modified nanowire sensors, indicating the capability of the ZnO nanowire sensors to electrically detect nonanal molecules. This decrease in resistance can be attributed to the reduction of ZnO sensor surface via the oxidation of aldehydes on the sensor surfaces2,15,29. Upon the introduction of air, a significant difference became evident in the recovery process between the bare and ODPA-modified nanowire sensors. The resistance of the bare ZnO nanowire sensor did not fully recover during the recovery process (400 s) under airflow, resulting in an accumulated permanent change in resistance. In contrast, the resistance of the ODPA-modified nanowire sensors fully recovered to the initial value during the recovery process, even at the same operating temperature. Figure 1e–g shows a significant enhancement of the recovery process of the ODPA-modified nanowire sensors for nonanal concentrations ranging from 270 ppb to 60.2 ppm. As shown in Fig. 1g, the ODPA-modified sensors exhibited a shorter recovery time, even at high concentrations of 60 ppm. In contrast, the sensor recovery time for the bare ZnO nanowire sensor increased with increasing nonanal concentration. Note that ODPA modifications did not significantly alter the electric transport properties of ZnO nanowire devices (Supplementary Fig. 2), suggesting that ODPA modification does not cause significant changes in transport parameters including space charge width, carrier density, and carrier mobility in the nanowires at the sensor operating temperatures. Therefore, the difference in the nanowire resistance under air flow was caused by inevitable structure variations of hydrothermally grown ZnO nanowires. The improvement of sensor recovery process by ODPA modification was observed between ZnO nanowire sensors with approximately the same resistances (Supplementary Fig. 3), highlighting the enhancement of sensor performance due to ODPA modification is effective regardless of the nanowire structural variations. The correlation between the reduction of recovery time and ODPA modification was further examined by controlling ODPA density on ZnO sensor surface, as shown in Fig. 1h, which shows the relationship between recovery time and ODPA modification time. As shown in Supplementary Figs. 4 and 5b, the ODPA density on the ZnO surface increases as ODPA modification time increases and reaches saturation after 60-min modification. This trend is in good agreement with the saturation behavior of sensor recovery time shown in Fig. 1h, indicating that the saturation characteristics of the sensor recovery time can be interpreted in terms of the saturation of the surface ODPA density on ZnO nanowire surfaces. On the other hand, prolonged modification times potentially lead to the dissociation of ZnO surfaces, namely the formation of zinc phosphates, as reported in our previous study30. In this study, the ODPA alkyl peak area, which had been saturated around 60 min modification, starts to increase again at 120 min modification (Supplementary Fig. 5b), suggesting possible partial dissociation of the ZnO surface. Therefore, approximately 60–90 min of modification time is optimal for forming an ODPA SAM that enhances sensor performance without disadvantages. Thus, these correlations between sensor recovery time and ODPA modification indicate that ODPA modification led to a shorter sensor recovery process.

Fig. 1: Electrical sensing of nonanal using bare and octadecylphosphonic acid (ODPA) modified ZnO nanowire sensor devices.
figure 1

Schematics of the device structure and sensing surface of (a) bare and (b) ODPA modified ZnO nanowire sensors. In bare ZnO nanowire sensors, ZnO (1 0 − 1 0) surface is directory exposed to ambient and nonanal vapor. On the other hand, in ODPA modified ZnO nanowire sensors, the ZnO surface is covered by an ODPA layer. Time-dependent resistances of (c, e) bare and (d, f) ODPA modified ZnO nanowire sensors exposed to nonanal vapor (32.7 ppm in c and d) or purified air (N2:O2 = 4:1). Relationship between recovery time of sensor response and (g) nonanal vapor concentration or (h) ODPA modification time. Recovery time is defined as the time required for 90% recovery in resistance values. All sensing measurements were performed at 160 °C under ambient pressure. The analytical fit lines in (g) are based on the function \(y=a+b\exp ({cx})\), optimized using the minimum chi-square method. The line in (h) represents the interpolation between data points using a loose Bezier function. The error bars in (h) represent the standard deviation of three independent measurements. Each data point was obtained by averaging the results of three separate experiments, and the error bars indicate the variability between these measurements. Source data are provided as a Source Data file.

Full size image

Next, we will discuss the origin of the reduced recovery times of the sensor responses. In general, permanent sensor resistance changes in metal oxide sensors are closely correlated with sensor surface catalyst deactivation or poisoning (e.g., residual analytes or catalyzed products on the surfaces)6,31. Notably, ODPA modification does not prevent aldehyde catalytic reactions on ZnO surfaces because sensitivity, which corresponds to the change in resistance under analyte exposure, does not decrease for ODPA-modified ZnO sensors. Thus, ODPA surface modification significantly suppressed the surface catalytic deactivation effects of ZnO nanowire sensors without lowering their electrical sensitivity, which is an inherently detrimental issue for chemical sensors19,20,21,23,24.

As shown Fig. 1, direct measurements of the molecular adsorption states on the sensor surfaces are essential for understanding the mechanism of the suppression of catalyst deactivation effects in ODPA-modified ZnO nanowire sensors. FT-IR measurements of nonanal on single-crystalline ZnO nanowire array samples were conducted to achieve consistency between the spectroscopic data and the single-nanowire electrical sensing data26,29. Figure 2a shows the IR spectra of the carbonyl group vibration band (1500–1800 cm−1) for liquid nonanal, vapor nonanal adsorbed on the bare ZnO surfaces, and vapor nonanal adsorbed on the ODPA-modified ZnO surfaces. The spectra exhibit three major peaks at 1728 cm−1, 1689 cm−1, and in the range of 1550–1600 cm−1. As reported previously29, the peak at 1728 cm−1 in liquid nonanal data can be assigned to the C = O vibration band (ν(C = O)), and the peak at 1689 cm−1 in ZnO surface data can be attributed to the red shift of ν(C = O), as illustrated in Fig. 2b. A comparison with density functional theory (DFT) calculations reveals that the red shift indicates the coordination of carbonyl groups with Zn ions on the ZnO (1 0 − 1 0) plane29. The relatively broad peak at 1550–1600 cm−1 can be attributed to the asymmetric COO− vibration band νas(COO−) of catalytically oxidized products-carboxylates29. The νas(COO−) band can be decomposed into two peak components of 1575 and 1550 cm−1. As shown in Fig. 2b, these two components can be attributed to the two different adsorption states of the carboxylates on the surfaces32,33. The lower peak at 1550 cm−1 corresponds to carboxylates that are more strongly adsorbed on ZnO surfaces than those at 1575 cm−132,33. Upon increasing the ODPA surface modification time, the ratio of these two different adsorption states gradually changed. The strongly adsorbed peak component at 1550 cm−1 decreased, whereas the weakly adsorbed peak component at 1575 cm−1 increased, as shown in Fig. 2c. These experimental trends in the IR spectra revealed the following implications for the adsorption states on the ZnO sensor surfaces. First, surface ODPA modifications did not suppress the catalytic oxidation of aldehydes on the ZnO surfaces because catalytically oxidized product carboxylates were consistently observed for all samples (Fig. 2a), including ODPA with a trans-zigzag alkyl conformation (Supplementary Fig. 5). This also suggests that nonanal molecules pass through the packed ODPA layer and reach surface unoccupied Zn ions, since ODPA does not occupy all the surface Zn sites on the ZnO surfaces, even in the packed trans-zigzag SAMs conditions (Supplementary Fig. 4). Second, the surface ODPA modification significantly weakened the adsorption states of the carboxylates on the sensor surfaces, as shown in Fig. 2d and e. Since catalytically oxidized carboxylates on ZnO nanowire surfaces have been reported to persist even at temperatures up to 350 °C29, the presence of carboxylates strongly adsorbed on sensor surfaces may be closely related to observed incomplete sensor recovery process owing to catalyst deactivation effects on sensor surfaces shown in Fig. 1. Thus, these IR measurements reveal that ODPA modification suppresses the catalyst deactivation effects shown in Fig. 1 by weakening the adsorption states of carboxylates on the ZnO surfaces.

Fig. 2: IR characterization of carbonyl group of nonanal on ZnO nanowires.
figure 2

a IR spectra of liquid nonanal, nonanal adsorbed on bare or ODPA modified ZnO nanowire array. ODPA modification times of 30, 60, and 120 min are shown for the ODPA modified samples. IR spectra were measured under ambient conditions. The orange and pink shaded regions represent decomposed peaks at 1575 and 1550 cm−1, respectively, using Gaussian functions. This color notation also applies to the color-coded parts in (b–e). b Vibrations of C = O (ν(C = O)) and asymmetric COO− (νas(COO−)). c Compositions of peaks at 1550 and 1575 cm−1 for νas(COO−). Schematic of the carbonyl group of nonanal adsorbed on (d) bare and (e) ODPA modified ZnO nanowires. Source data are provided as a Source Data file.

Full size image

Next, we will focus on the desorption behavior of aldehydes and carboxylates from the ZnO sensor surfaces to understand the sensor recovery characteristics shown in Fig. 1. The temperature dependent adsorption states of the aldehydes and carboxylates on the ZnO surfaces were analyzed using IR spectroscopy with thermal treatments. IR spectra of nonanal-adsorbed ZnO nanowire samples were measured after accumulative thermal treatments at temperatures ranging from 50 to 200 °C in ambient air for 10 min. Detailed surface desorption and reaction characteristics were analyzed using IR peak analysis and temperature-programmed desorption (TPD)-MS measurements. As shown in Fig. 3a and b, catalytically oxidized carboxylates from nonanal on bare ZnO surfaces persisted even at high temperatures up to 200 °C. Conducting thermal treatments at a higher temperature range (300–400°C) is necessary to completely remove carboxylates from bare ZnO surfaces, as also shown in Fig. 3b, Supplementary Fig. 6, and a previous study29. This strong residual trend of carboxylates on bare ZnO surfaces at such temperatures is closely related to the unrecovered sensing behavior during molecular desorption process at 160 °C in Fig. 1. In contrast, as shown in Fig. 3c and d for ODPA-modified ZnO surfaces, catalytically oxidized carboxylates tend to easily desorb even at a significantly lower temperature range between 50 and 75 °C. This significantly enhanced desorption behavior of carboxylates on ODPA-modified ZnO surfaces is consistent with the fully recovered sensing data during the molecular desorption process at 160 °C in Fig. 1. Furthermore, as shown in Supplementary Fig. 7, TPD-MS measurements revealed that nearly all nonanal molecules were oxidized as nonanoic acid and desorbed below 150 °C. Thus, the results in Fig. 3a–d reveal that ODPA surface modification on ZnO sensor surfaces substantially enhances the desorption of the catalytically oxidized product-carboxylates at a lower temperature range by weakening the adsorption states between the carboxylates and ZnO surfaces. The next issue is to understand how ODPA surface modification changes the adsorption states between the carboxylates and ZnO surfaces. Since phosphonic acid moieties are known to most strongly interact with ZnO surfaces via acid–base interactions, one plausible mechanism is based on the electrostatic modulation of ZnO surfaces through electron-drawing phosphonic acids16,17,34. Previous studies have also reported the modulation of catalytic activity or work function of metal oxides through surface electrostatic changes via phosphonic acid modification16,17,34. To examine the electronic contributions of the head phosphonic acid group of ODPA on ZnO surfaces in Fig. 3c, we conducted thermal treatment experiments for ZnO surfaces modified by methylphosphonic acid (MPA) without an aliphatic chain. Figure 3e and f shows the IR spectra of the MPA-modified ZnO nanowire surfaces after thermal treatments at various temperatures. The catalytically oxidized product-carboxylates on the MPA-modified ZnO surfaces clearly persisted up to 200 °C of thermal treatment. This residual trend of carboxylates at high temperatures contrasts sharply with the IR spectra of ODPA shown in Fig. 3c. The TPD-MS measurements also confirmed the presence of residual carboxylate components on the ZnO surfaces (Supplementary Fig. 8). Although the temperature-dependent IR spectra of MPA-modified ZnO surfaces resemble those of bare ZnO surfaces in Fig. 3a, the primary vibration band νas(COO−) of carboxylates on the MPA-modified ZnO surfaces was observed at 1575 cm−1, which can be attributed to the relatively weak adsorption states of carboxylates, as shown in Fig. 2b. Even with 30 min modification of MPA, the weak adsorption state is as dominant as it is with 60-min modification of ODPA (Supplementary Fig. 9), indicating that the effect of the phosphonic acid head group is comparable between the 30 min modification of MPA and the 60 min modification of ODPA. Therefore, the MPA surface coverage is similar to or even higher than that of ODPA. This observed inconsistency between the adsorption states of carboxylates and the desorption temperatures for MPA-modified ZnO surfaces suggests potential unknown contributions of other molecular moieties of ODPA to the enhanced desorption behaviors. Because the structural difference between MPA (C1 phosphonic acid) and ODPA (C18 phosphonic acid) is solely the length of the alkyl chain, interactions involving the C18 alkyl chains must cause the suppression of catalyst deactivation by ODPA modifications. Therefore, these temperature-dependent IR spectra data demonstrate that the desorption temperatures of catalytically oxidized product-carboxylates from ZnO sensor surfaces are significantly influenced by surface modifications.

Fig. 3: Desorption temperature analysis of nonanal on ZnO nanowires.
figure 3

IR spectra of the carbonyl region of nonanal adsorbed on (a) bare, (c) ODPA-modified, and (e) methylphosphonic acid (MPA)-modified ZnO nanowires after thermal treatments in ambient air for 10 min at temperatures ranging from 50 to 200 °C. Temperature-dependent surface reaction and desorption states of nonanal on (b) bare, (d) ODPA-modified, and f MPA-modified ZnO nanowires. The color scheme of the carbonyl and carboxyl moieties in (b), (d), and (f) corresponds to the wavenumbers shown in (a), (c), and (e). Source data are provided as a Source Data file.

Full size image

Next, we will computationally investigate the role of ODPA surface modifications in suppressing catalyst deactivation effects on ZnO surfaces. This investigation aims to address the discrepancy in catalytic sensing data between MPA and ODPA surface modifications. The electronic effects of phosphonic acids on catalytic ZnO surfaces alone cannot fully explain the observed suppression of catalyst deactivation effects on the catalytic sensing characteristics of the ODPA-modified ZnO surfaces. The adsorption structures and energy diagrams for nonanal oxidation on bare and ODPA-modified ZnO surfaces were investigated using DFT calculations combined with molecular dynamics (MD) and molecular mechanics (MM) methods. Details of these calculations are provided in the Methods section. Figure 4a shows snapshot images of the calculated stable structures of oxidized products (carboxylates) of nonanal adsorbed on the ZnO (1 0 − 1 0) surface. The calculations indicate that the carboxylate group was coordinated with the unsaturated surface Zn sites, and the alkyl chain moiety was unexpectedly aligned perpendicular to the ZnO c-axis [0 0 0 1] via van der Waals interactions. This is because the accumulated van der Waals interactions between the alkyl chain and the ZnO surface are no longer negligible as the carbon number of the aliphatic compounds increases, despite the relatively small van der Waals interactions between each CH2 and the ZnO surface35,36. Figure 4b shows the calculated stable structures of the ODPA-modified ZnO surfaces. In the ODPA-modified ZnO surface calculations, the surface ODPA density was set to be consistent with the experimentally estimated ODPA density with a 60 min modification (Supplementary Fig. 4). As expected for typical SAM structures, the calculated alkyl chains mostly formed a trans-zigzag-dominated conformation. This trend is consistent with the IR data obtained in this study (Supplementary Fig. 5c). However, some alkyl chains of ODPA tended to lie down on the ZnO surface because of van der Waals interactions with unsaturated Zn sites, despite the relatively strong alkyl–alkyl interactions of the C18 structure. Again, this is because the accumulated van der Waals interactions between the alkyl chain and the ZnO surface tend to be comparable or even stronger than the alkyl–alkyl interaction of C18 structure. Although the fraction of alkyl chains lying down and interacting with the ZnO surface is estimated at around 20%, the actual fraction should be less than the estimated value because the estimation is based on the most stable structure with the ideal alignment of ODPA on the ZnO (1 0 − 1 0) surface. This estimation is consistent with the FT-IR experimental observation that the alkyl chains of ODPA SAMs on ZnO surface form packed trans-zigzag conformation at the longer modification time. Figure 4c shows the calculated nonanal adsorption structure of the ODPA-modified ZnO surface, with the adsorbed carboxylate highlighted in green. Notably, the alkyl chain of the carboxylate (an oxidized product of nonanal) lies on the ZnO surface rather than in a trans-zigzag-dominated conformation because of interactions with the surrounding alkyl chains of ODPA. This results in a conformational change in the alkyl chain of ODPA, as shown in Fig. 4c. This computationally predicted alkyl conformational change is experimentally validated in the next section using spectroscopy with deuterated phosphonic acids. Next, we will compare the different adsorption states of nonanal and the oxidation product-carboxylate on bare ZnO and ODPA-modified ZnO surfaces in the calculated adsorption energy diagram. Figure 4d shows a comparison between the bare ZnO surface and the ODPA-modified ZnO surfaces on the calculated energy diagram of the four states during the adsorption process of nonanal and the oxidation product-carboxylate, including (1) the total energy of the surface and nonanal before adsorption, (2) the total energy of the physisorption state, (3) the total energy of the oxidization product-carboxylate after adsorption, and (4) the total energy of the surface and carboxylic acid after desorption. The ODPA molecules clearly destabilized the adsorbed nonanal, showing an energy difference of 0.719 eV. Thus, the experimentally observed suppression of catalyst deactivation effects (reduced desorption temperatures of the oxidized product of nonanal on ODPA-modified ZnO in Fig. 3) can be reasonably explained by the adsorption–desorption energy diagram. This reduction in desorption energy, as well as the energy required for the dissociation of carboxylates from the ZnO surface, increases the turnover frequency (TOF) of the catalytic reactions, promoting the oxidation of aldehydes and enhancing sensor response. On the other hand, the IR analysis (Fig. 3) indicates that the amount of nonanal reaching onto the ZnO nanowire surface decreases via ODPA modification, resulting in competing effects on sensor response between oxidation enhancement and suppression of adsorption. As a result, ODPA modifications significantly improve recovery speed without sacrificing sensor response. To further understand why the adsorbed carboxylates became destabilized, the calculated adsorption energy difference (0.719 eV) was further decomposed into (A) “ODPA alkyl chain deformation energy including the loss of interaction between the alkyl chain and ZnO,” (B) “difference in interaction energy of nonanoate (oxidized product of nonanal) to the surface,” and (C) “nonanoate alkyl chain deformation energy,” as shown in Fig. 4d. The ODPA alkyl chain deformation energy term dominated and contributed 81.3% of the total energy difference (0.719 eV), indicating that the conformational change in the alkyl chains of ODPA was the origin of the destabilization of the adsorbed carboxylates on the ODPA-modified ZnO surfaces. As shown in Fig. 4e, the large energy difference caused by the conformational change in the ODPA alkyl chain can be attributed to the strong van der Waals interaction between the alkyl chain and the ZnO surface. In other words, the competing effects of ODPA and nonanal in terms of alkyl–ZnO interactions reduced the desorption temperature of oxidized nonanal and suppressed catalyst poisoning. Although both physical filter and chemical interactive effects should exist in the ODPA modified ZnO nanowire sensors, the chemical effects play a major role in the measured sensor characteristics. The FT-IR analysis (Fig. 3) shows that the adsorption amount of nonanal onto the ZnO surface decreases by ODPA modification, highlighting the presence of a physical molecular filter effect via the ODPA SAM. However, the sensor response does not significantly change by the ODPA modification on ZnO nanowire sensors, indicating that a sufficient amount of nonanal supplied from vapor can penetrate through the ODPA layer and coordinates to catalytic sites on ZnO surfaces. On the other hand, the chemical interactive effects of ODPA modification include (1) the electron-drawing effect of the phosphonic acid head group and (2) the destabilization of adsorbed carboxylates through van der Waals interactions between the alkyl chains and the ZnO surface. The desorption experiments of MPA-modified ZnO nanowires with the DFT calculations highlight that the van der Waals interactions between the alkyl chains and the ZnO surface play a critical role in the suppression of catalyst deactivation. The van der Waals interactions between alkyl chains and metal oxide surfaces are significantly influenced by the metal elements with a large number of electrons. Therefore, if molecules composed of light elements such as water, interpose between the alkyl chains and the metal oxide surface, the van der Waals interactions are generally expected to decrease. However, the densely packed, hydrophobic alkyl chains of the ODPA SAM reduce the number of water molecules reaching the ZnO surface. As a result, the effect of relative humidity on the suppression of catalyst deactivation via the van der Waals interactions might be limited. In Fig. 1, although residual water molecules in ambient air were not removed in the sensor measurements, the effects of the van der Waals interactions remain evident. The desorption temperature analysis based on FT-IR spectroscopy (Fig. 3) was also performed in ambient air, supporting the humidity-resistant enhancement of sensor performance. This suggests that the ODPA SAM effectively mitigates the impact of humidity variations on the sensor performance.

Fig. 4: Density functional theory (DFT) calculations.
figure 4

Snapshots and schematic images showing the most stable structures of: (a) nonanal adsorbed on the ZnO (1 0 − 1 0) surface as catalytically oxidized carboxylate, (b) ODPA-modified ZnO surface, and (c) nonanal adsorbed on ODPA-modified ZnO surface. The nonanal and ODPA molecules near the adsorbed nonanal (ODPA 1) are highlighted in the side and top views. d Energy diagram of the surface catalytic nonanal oxidation on ZnO and ODPA-modified ZnO surfaces, with the composition ratio of the total energy difference between bare and ODPA-modified ZnO surfaces with adsorbed oxidized carboxylate. e Schematic showing the destabilization of adsorbed nonanal on ODPA-modified ZnO surfaces via the deformation of alkyl chains of ODPA.

Full size image

These results for the bare, ODPA-modified, and MPA-modified ZnO sensors consistently indicate that the alkyl chain of ODPA significantly reduces the desorption temperature and sensor recovery time of the ODPA-modified ZnO nanowires. Although spectroscopic measurements of the conformational changes in ODPA and nonanal alkyl chains are required to directly validate the model in Fig. 4, discriminating the conformational changes between the ODPA alkyl and nonanal alkyl chains is challenging. To overcome this issue, deuterated phosphonic acid modification was performed. Hexadecyl phosphonic acid-d33 (D-HDPA) was synthesized and immobilized on the ZnO nanowire surfaces (Fig. 5a). Detailed procedures for the D-HDPA synthesis can be found in the Methods section. Figure 5b shows the IR spectra of the alkyl and deuterated alkyl (D-alkyl) regions of nonanal, ODPA, and D-HDPA on the ZnO nanowire surfaces. D-alkyl exhibits IR absorption within the range of 2000–2300 cm−1, which is lower than that of the alkyl and carbonyl of nonanal (2800–3000 cm−1)37,38. This difference in IR spectra enables the individual evaluation of conformational changes in nonanal alkyl and phosphonic acid D-alkyl on ZnO surfaces. Notably, a reduction in the desorption temperature was also observed even for the D-HDPA-modified ZnO nanowires, confirming that deuteration does not change the surface nonanal reaction and desorption (Supplementary Figs. 10 and 11). The reduction of desorption temperature due to D-HDPA modifications also indicates that phosphonic acids with a chain length of 16, which is shorter than that of ODPA (C18), are still effective for the reduction of sensor recovery time. Therefore, phosphonic acid SAMs with a chain length of 16 are expected to suppress the sensor recovery time similar to ODPA cases. In the spectroscopic analysis, IR pMAIRS was conducted to assess the degree of alkyl chain orientation39,40,41. Figure 5c shows the IR pMAIRS data of the alkyl and D-alkyl regions, which comprises in-plane (IP) and out-of-plane (OP) spectra, for nonanal on D-HDPA-modified ZnO nanowires. The IR absorption peaks for the nonanal alkyl and D-HDPA D-alkyl groups can clearly be observed. The symmetric CH2 and CD2 vibration bands (νs(CH2) and νs(CD2)) were analyzed to measure the conformation of alkyl chains. Figure 5d shows the wavenumbers of νs(CD2) of the samples with varying D-HDPA modification times. Figure 5e shows the ratio of the OP/IP spectra, which represents the degree of orientation of the alkyl chain. A decrease in νs(CD2) wavenumber suggests a transition of the alkyl conformation from gauche-rich to a trans-zigzag-dominated structure42,43. Therefore, the trend of Fig. 5d indicates that the D-alkyl of D-HDPA changes from gauche-rich to trans-zigzag-dominated conformation as D-HDPA surface density increases, which is consistent with the general trend of alkyl SAM formations, as shown in Supplementary Fig. 530. When the modification time is less than 60 min, where the surface coverage of D-HDPA is relatively low, nonanal adsorption significantly alters D-alkyl conformation to a trans-zigzag-dominated structure. Conversely, the orientation of the D-alkyl of D-HDPA changed slightly to an ordered structure upon nonanal adsorption (Fig. 5e). When the peak ratio equals 1, the alkyl conformation corresponds to a completely disordered structure, while a higher peak ratio indicates ordered alkyl structures39,40,41. Figure 5f highlights that conformation and orientation changes in the alkyl chains of surface phosphonic acids, resulting from nonanal adsorption, play a critical role in poisoning-less catalytic sensing. Furthermore, the suppression of catalyst deactivation via ODPA modification was investigated for various analytes including aldehydes of linear, branched, and cyclic structures with smaller carbon numbers of 6 or 7 compared to nonanal (Supplementary Fig. 12). The reduction in sensor recovery time due to ODPA modification was smaller for these molecules compared to nonanal, supporting the proposed mechanism that the strong van der Waals interactions caused by the long alkyl chain of nonanal plays a key role in the catalytic deactivation. These results promise that the ODPA modification enables molecular skeleton discrimination based on the obtained trends on the recovery time. In addition, this molecular discrimination mechanism could be effective for other functional groups such as alcohols because the molecular selectivity relies on carbon skeleton structures rather than functional groups.

Fig. 5: IR p-polarized multiple-angle incidence resolution spectroscopy (pMAIRS) characterizations of alkyl chains of nonanal on organic-modified ZnO nanowires using deuterated phosphonic acids.
figure 5

a Schematic of hexadecyl phosphonic acid-d33 (D-HDPA) modification on ZnO nanowires. b IR spectra of alkyl chains of nonanal, alkyl chains of ODPA, and D-alkyl chains of D-HDPA on ZnO nanowires. c IR pMAIRS spectra of alkyl and D-alkyl regions for nonanal adsorbed on D-HDPA-modified ZnO nanowires. d νs(CD2) OP wavenumbers for D-HDPA modified ZnO nanowires with varying modification time. Wavenumbers before and after nonanal adsorption are shown. e IR peak ratio between OP and IP spectra (AOP/AIP) for νs(CD2) of D-HDPA-modified ZnO nanowires with varying modification time. IR peak ratios before and after nonanal adsorption are shown. All IR measurements were conducted at ambient temperature. f Proposed mechanism for the suppression of catalyst poisoning via ODPA surface modification during aldehyde sensing on ZnO. Source data are provided as a Source Data file.

Full size image

In summary, we demonstrated that weak van der Waals interactions between hydrophobic aliphatic alkyl chains and hydrophilic ZnO surfaces, which have been underestimated as interactions during catalytic molecular sensing, prevent catalyst deactivation effects during the electrical sensing of aliphatic aldehydes on ZnO sensor surfaces. ODPA modification significantly reduces the recovery time of the sensor response to nonanal without compromising sensitivity in electrical measurements of single-crystalline ZnO nanowire sensor devices. Spectroscopic analysis revealed that nonanal penetrates through packed trans-zigzag ODPA SAMs and directly coordinates with surface Zn ions as an oxidized carboxylate. The desorption temperature of the carboxylates on ODPA-modified ZnO (<150 °C) was substantially lower than that on bare ZnO nanowires (>300 °C), indicating a significantly reduced catalyst poisoning effect. IR pMAIRS with deuterated phosphonic acids showed that nonanal penetration and adsorption forced the alkyl chains of ODPA to form a trans-zigzag-dominated conformation and an ordered orientation. DFT calculations revealed that the deformation of ODPA significantly destabilized the adsorbed oxidized carboxylates because of the unexpectedly strong interactions between the alkyl chains and the ZnO surface. Thus, through a comprehensive examination of potential interactions between the ODPA SAM and the ZnO nanowire surface, we demonstrated that van der Waals interactions of ODPA critically enhance ZnO nanowire sensor performance. Engineering strong interactions between alkyl chains and metal oxide nanostructures using organic and/or inorganic modifications offers a promising strategy for developing chemical sensors and heterogeneous catalysts that combine durability and molecular or reaction selectivity.

Methods

ZnO nanowire growth

Single-crystalline ZnO nanowires were hydrothermally grown on a ZnO seed layer/SiO2/p-Si substrate. A 5-nm Ti adhesion layer and a 100 nm ZnO seed layer were sequentially deposited onto a 100 nm SiO2/p-type Si substrate via radio frequency (RF) sputtering. The hydrothermal reaction solutions consisted of 5 mM zinc nitrate hexahydrate, Zn(NO3)2·6H2O (Wako, 99.0%) and 5 mM hexamethylenetetramine (HMTA), (CH2)6N4 (Wako, 99.0%). The ZnO-deposited substrate was immersed in the growth solution and maintained at 80 °C for 48 h. The solution was replaced with freshly prepared solution every 24 h until completion of the hydrothermal synthesis. Following the reaction, a ZnO nanowire array was grown on the substrate. Subsequently, the samples were rinsed with DI water and IPA. The ZnO nanowire array sample was then annealed for 1 h at 600 °C in atmospheric air to prevent surface electrical insulation via the formation of zinc carbonate44. Structural characterization was performed using field emission-scanning electron microscopy (FE-SEM, JEOL, JSM7610F). Hexagonal columnar structures with a diameter of approximately 100 nm were observed, indicating that the grown ZnO nanowires had a hexagonal structure with the prism (1 0 − 1 0) plane as the main face.

Phosphonic acid SAM modification on ZnO nanowires

Solutions of 0.05 mM ODPA, MPA, or D-HDPA-d17 were prepared using toluene (ODPA and D-HDPA) or methanol (MPA) solvent to form a uniform SAM without surface ZnO dissolution during modification30. The annealed ZnO nanowire array or fabricated device sample was immersed in the solution (10 mL) at 20–25 °C for a controlled modification time. After cleaning (first with toluene and then with tetrahydrofuran) and airflow drying, the samples were annealed at 150 °C for 30 min in atmospheric air. D-HDPA was synthesized according to a modified procedure reported in the literature45. 1-bromohexadecane-d33 (250 mg, 0.74 mmol) and triethyl phosphite (0.438 mL, 2.56 mmol) were used. The peaks detected in the mass spectrum of the product were consistent with the calculated molecular weight of the protonated D-HDPA (Supplementary Fig. 13).

Sensor device fabrication

The hydrothermally grown ZnO nanowires were dispersed in acetone via sonication and then dropped onto a 100-nm SiO2/n + -type Si substrate with a 5 nm Ti adhesion layer/200 nm Pt pad electrodes, which were patterned using photolithography and RF sputtering. Photolithography was employed to fabricate the two-terminal electrical contacts on the dropped nanowires. A photoresist (Merck, AZP-1350) was coated using a spin coater (MIKASA, MS-A100), followed by photolithography exposure and development (Merck, AZ300MIF). A 200 nm layer of Pt was then deposited via RF sputtering. The fabrication process was finalized with a lift-off process in acetone and subsequent acetone cleaning. The 4-terminal devices were fabricated using electron beam (EB) lithography with an EB resist (PMMA 7 wt% in anisole solution), developer (TCI, 4-methyl-2-pentanone), and lift-off (NMP, TCI). The geometry of the fabricated device was analyzed using FE-SEM (Supplementary Fig. 14). The diameters of employed ZnO nanowires were typically ranged from 70 to 180 nm. The length of the nanowires in the sensor device is defined by the electrode gap, which is approximately 1-2 µm. In the temperature dependence of the nanowire electrical resistance, nanowire resistance significantly decreases with increasing temperature (Supplementary Fig. 15), indicating semiconducting electrical conduction properties of ZnO nanowires.

Sensing measurements

For the sensing measurements, the target aldehydes were introduced into the chamber of a temperature-controlled probe station using a bubbling system with purified air as the carrier gas (N2:O2 = 4:1). The concentrations of the gas-phase aldehydes were controlled through dilution and calibrated using GC-MS (Supplementary Fig. 1), with a constant total gas flow of 500 SCCM. Electrical characterization of the ZnO nanowire devices was conducted using a semiconductor analyzer (Keithley 4200SCS). Sensor response was determined as R0/Rg, where R0 and Rg represent the resistances under exposures of purified air and target aldehydes, respectively. Recovery time was defined as the time required for the resistance reach to 90% R0 after the end of the target aldehyde flow.

Molecular adsorption on ZnO nanowires

The ZnO nanowire array substrate was placed in a bottle filled with saturated nonanal vapor and left undisturbed for 10 min. During adsorption, the bottle containing the target molecule and the substrate were each heated to 50 °C. Nonanal-adsorbed ZnO nanowires were used for subsequent measurements.

FT-IR measurements

The FT-IR and pMAIRS spectra of the surface molecules on the ZnO nanowires were obtained at room temperature using an FT-IR spectrometer (Thermo Fisher Scientific, Nicolet iS50) equipped with a mercury–cadmium–telluride (MCT) detector. In total, 300 scans were performed to obtain each spectrum. For each pMAIRS measurement, a series of eight single-beam sample measurements were carried out at angles of incidence ranging from 44° to 9° at 5° intervals, with a resolution of 4 cm−1. The test chamber was then purged with dry air. The FT-IR or pMAIRS spectra of bare ZnO nanowires were used as background spectra for the measurements of the SAM-modified samples. For the nonanal-adsorbed samples, the spectra after SAM modification were used as the background. The obtained angle-dependent spectra were used to calculate the OP and IP spectra. The theoretical details regarding the IR pMAIRS analysis can be found in the literature39. For these IR measurements, a double-polished float-zone Si substrate was used for the ZnO nanowire array samples.

GC-MS analysis

A GC-MS instrument (Shimadzu GC-MSQP 5050 A equipped with a Supelco SLB-IL60 capillary column and an OPTIC-4 inlet temperature control system) was used to analyze the desorbed compounds at each temperature. The samples were heated to different temperatures (150, 200, 250, 300, and 350 °C) and cooled to 40 °C. The desorbed gas at each temperature was then analyzed using GC-MS. The measurements for each temperature were conducted sequentially (from 150 to 350 °C) without removing the sample from the inlet port.

DFT calculations

The ZnO (1 0−1 0) surface was modeled as a two-dimensional slab in a three-dimensional periodic cell. A (6 × 4) plane slab with a thickness of six atomic layers was constructed. To suppress interactions between the slabs, the z-directional length of the unit cell containing the slab and vacuum layers was set to 50 Å. Two types of ODPA adsorption arrangements were considered: straight and zigzag. The straight arrangement was found to be more stable and was therefore adopted.

Because various adsorption conformations of nonanal and alkyl chain orientations of ODPA were expected, we searched for the most stable surface structure using the quench dynamics method with the Forcite module implemented in the Materials Studio 2022 software46. The COMPASS III47 and Dreiding48 force fields were used under NVE ensemble conditions, with a time step of 1 fs and a simulation time of 300 ps. For the COMPASS III force field, two different calculations were performed with initial temperatures set to 400 and 600 K. For the Dreiding force field, the temperature was set to 400 K. The coordinates of all the atoms in the slab were fixed during the quench dynamics. The quench-step number was set to 10,000, resulting in 31 optimized structures obtained for each quench dynamics run. From these, the five most stable structures were selected. A more detailed description of the quench dynamics can be found in the literature49.

The candidate structures obtained from the quench dynamics simulations were further optimized using DFT. The Vienna ab initio simulation package (VASP 5.4.4)50,51 was employed to perform DFT calculations, using the Perdew–Burke–Ernzerhof exchange-correlation functional52. The Kohn–Sham equations were solved with a plane-wave basis set employing the projector-augmented wave method53. The cutoff energy for the plane-wave basis set was set to 400 eV. The convergence threshold for the self-consistent field (SCF) iteration was set to 1.0 × 10−5 eV. The atomic coordinates were relaxed until the forces on all atoms were less than 0.05 eV/Å. During optimization, the bottom two layers of the ZnO slab remained fixed. Because the computational cells were large, only the Γ-point was used to sample the Brillouin zone. A Tkatchenko-Scheffler dispersion correction scheme54 was adopted. The lowest energy structure among the candidate structures optimized with DFT was selected as the initial structure for subsequent quench dynamics. This procedure was repeated, alternating between quench dynamics and DFT calculations, until no further updates to the most stable structure were made.

We conducted a partitioning analysis of the energy difference using a modified application55 of the distortion/interaction model56 and activation strain model57 for adsorption. The surface distortion energy was estimated based on the difference between the SCF energy of the structure with the adsorbate removed from the most stable adsorption structure and that of the most stable surface structure in the absence of an adsorbate. The interaction energy between the adsorbate and surface was determined by calculating the difference between the energy of the most stable adsorption structure and the energies of the surface and adsorbate, which were calculated separately, maintaining their geometrical structures in the adsorption system. The adsorbate represented the RCHO moiety.