Highly sensitive hydrogen sulfide (H₂S) gas sensors from viral-templated nanocrystalline gold nanowires

As depicted in figures 1(a)–(d), to form H₂S gas sensors, bio-templated gold nanowires were assembled on electrodes fabricated on a Si/SiO₂ substrate using a modification of a previously reported procedure [30, 31]. A gold-binding [26] M13 bacteriophage was used as the template. This particular clone displayed an 8-mer peptide (VSGSSPDS) with an affinity for gold on the N-terminus of each of 2700 copies of the pVIII major coat protein [26, 32]. The 300 nm thermal oxide layer electrically insulated the viral-templated nanowires from the underlying Si substrate. The Ti/Au (20 nm/180 nm) electrodes, which were 50 μm wide and separated by a 3 μm gap, were fabricated using standard photolithography, electron beam deposition, and lift-off techniques. Prior to nanowire assembly, the substrates with patterned electrodes were solvent cleaned with ultrasonication in acetone, isopropanol, and deionized water and activated with O₂ plasma using a reactive ion etching (RIE, Surface Technology Systems) system at 100 W, 100 mT for 30 s. This plasma treatment was critical for uniform and non-specific adhesion of the viral template to the patterned substrates. The pre-patterned substrates were then incubated with gold-binding phage in tris-buffered saline (TBS, 50 mM Tris–HCl, 150 mM NaCl, pH 7.5) for 10 min. During this step, the gold-binding phages were adsorbed onto the substrate. The substrate was then washed and rinsed in TBS with 0.7% Tween 20 and deionized water. To control the density of phage on the substrate surface, and ultimately the number of parallel electrical connections between the electrodes, samples were made with three different phage concentrations: 1 × 10⁸ pfu μl⁻¹, 3 × 10⁸ pfu μl⁻¹, and 5 × 10⁸ pfu μl⁻¹. Gold nanoparticles were selectively bound to the gold-binding phage by submerging the substrate in a 5 nm diameter gold colloid solution of 5 × 10¹³ particles ml⁻¹ (BBI Solutions.) for 1 h. The substrate was rinsed 3 times with deionized water and gently dried with air. The nanoparticles bound to the phages were used as seeds for electroless deposition of gold using Nanoprobes GoldEnhance^TM LM solutions. The electroless deposition time was varied from 3 to 12 min to control the nanoparticle size and electrical resistance of the gold nanowires formed. The viral-templated gold nanowire devices are hereafter referred to as 'as-assembled' devices. As-assembled devices were treated with O₂ plasma for 30 s at 100 mT and 100 W using the RIE system, dipped in ethanol (Sigma-Aldrich) for 10 min, and gently blown dry with air. This two-step process removed the viral template and surface organics, as well as Au₂O₃ which may have been generated by exposure to O₂ plasma [33, 34]. These devices will hereafter be referred to as 'ethanol-treated' devices and were used to evaluate the contribution of the viral template to device resistance and sensing performance.

Zoom In Zoom Out Reset image size

Scanning electron microscopy (SrEM, Phillips XL30 FEG) was used to determine the morphology and spatial distribution of the gold nanowires on the substrate. The number of seed particles per phage was quantified. A short electroless deposition of 1 min was used to slightly enlarge the gold nanoparticle seeds without merging them, making them easier to observe with SEM. The particles on 15 individual templates were counted for range and average. In addition, the areal fill factors of viral-templated gold nanowires on the Si/SiO₂ substrate were determined for devices assembled with phage concentrations of 1 × 10⁸ pfu μl⁻¹, 3 × 10⁸ pfu μl⁻¹, and 5 × 10⁸ pfu μl⁻¹. For each concentration, a minimum of 10 locations were imaged and analyzed. Furthermore, the width of the gold nanocrystal components of the viral-templated nanowires was measured at various electroless deposition times. Approximately 100 gold nanoparticles were analyzed for each electroless deposition time. Transmission electron microscopy (TEM, Phillips Tecnai 12) was used to analyze the morphology and connection between the nanoparticles within the gold nanowires. For TEM sample preparation, gold nanowires were fabricated on unpatterned SiO₂ substrates and dispersed in deionized water by ultrasonication. The dispersed nanowires were then loaded onto carbon-coated copper grids and dried in vacuum.

The room temperature resistance of each as-assembled viral-templated gold nanowire device was determined using two-terminal, current–voltage (I−V ) measurements in which the current was recorded as the applied voltage was swept from −0.3 to 0.3 V (Keithley 2636A sourcemeter) in 30 mV increments. The same procedure was used to measure device resistance after ethanol treatment.

Devices selected for gas sensor measurements were wire-bonded (West-bond Inc. 7499D) at room temperature to a copper printed circuit board (PCB) with 1% Si/Al wire before sensing analysis. Wire-bonded sensors were placed in a flow cell chamber with a gas inlet and outlet. A constant bias of 0.15 V was applied to each device and, after establishing a stable baseline resistance, sensing analysis was performed at ambient temperature and pressure under a constant flow rate of 200 sccm. The resistance change of each viral-templated gold nanowire sensor was measured with exposure to H₂S gas. To vary analyte concentration, H₂S gas was diluted using dry air as the carrier gas and each sensor was alternately exposed to H₂S gas at the specified concentration and dry air for time intervals of 15 min and 30 min, respectively. A mass flow controller with LabView interface was used to control the H₂S concentration and exposure time. A similar procedure was followed for selectivity analysis using NH₃ and NO₂ as the gas analytes.

The morphology of the viral-templated gold nanowires which comprise the sensors was examined with SEM to study the effect of electroless deposition time and phage concentration on the as-assembled gold nanowires. A representative image of a sensor assembled with a phage concentration of 1 × 10⁸ pfu μl⁻¹ and an electroless deposition time of 3 min is shown in figure 2(a). Gold nanowires slightly less than 1 μm in length, composed of well-defined nanoparticles were seen randomly distributed on the substrate in addition to a few individual gold nanoparticles. The number of isolated nanoparticles was small in comparison to those incorporated in the nanowires. As shown in the high magnification SEM image in figure 2(b), the width and connectivity varied along individual nanowires, as well as from nanowire to nanowire. Further SEM analysis revealed that the number of gold nanoparticle seeds per template ranged from 31 to 61 with an average of 42. We, therefore, attributed the largest deviations in nanowire width and connectivity to differences in the density and arrangement of gold nanoparticle seeds along the gold-binding phage prior to electroless gold deposition. The size and morphological changes, which accompanied increased electroless deposition time, can be seen in figures 2(b)–(d). As the electroless deposition time increased from 3 to 12 min, the overall width and connectivity of the gold nanowires also increased. The average width of the resulting nanocrystals on nanowires for 3, 7, and 12 min of gold deposition were 29 ± 7 nm, 60 ± 13 nm, and 82 ± 16 nm, respectively. No difference in the morphology, structure, or distribution of nanowires on the substrate surface was observed with SEM after template removal by ethanol treatment as shown in the supplementary data (available at stacks.iop.org/Nano/25/135205/mmedia), figure S1. The TEM image in figure 3 revealed the detailed structure of the nanowires. Mostly shorter fragments (<200 nm) of nanowires were observed with TEM. This is likely because the fragile nanowires were fractured during sample preparation. The nanoparticles within the nanowires were polydisperse. Both single chain and multi-chain nanoparticle arrangements were observed. At higher magnifications near-point-contact connections between nanoparticles that, in many cases, structurally and electrically held the viral-templated gold nanowires together were observed.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figures 4(a)–(c) show SEM images of the viral-templated gold nanowire sensors fabricated with increasing phage concentration and an electroless deposition time of 3 min. Phage concentrations of 1 × 10⁸ pfu μl⁻¹, 3 × 10⁸ pfu μl⁻¹, and 5 × 10⁸ pfu μl⁻¹ yielded devices with an average gold nanowire surface coverage of 16%, 39%, and 49%, respectively. At a phage concentration of 1 × 10⁸ pfu μl⁻¹ both isolated and small clusters of nanowires were observed within the 3 μm gap between the electrodes. At higher phage concentrations, 3 × 10⁸ pfu μl⁻¹ and 5 × 10⁸ pfu μl⁻¹, gold nanowires formed a continuous, mesh-like structure between the electrodes. Given the relative size of the gold nanowires and electrode gap, multiple nanowires were required to physically bridge the gap between the two electrodes. As a result, a lower density of complete, physical connections was observed for the 1 × 10⁸ pfu μl⁻¹ phage concentration, as compared to the 3 × 10⁸ pfu μl⁻¹ and 5 × 10⁸ pfu μl⁻¹ phage concentrations.

Zoom In Zoom Out Reset image size

The effects of electroless deposition time and phage concentration on sensor resistance were studied. Devices displayed Ohmic behavior within the −0.3 and 0.3 V two-terminal measurement range at room temperature for all conditions investigated. Viral-templated sensor resistance is shown as a function of gold deposition time in figure 5(a) along with the device yield for each condition. Device yield is defined as the ratio of devices with measurable resistance (less than 10 GΩ) to the total number of devices fabricated. The sensor resistance decreased as the electroless deposition time increased, which is consistent with a previous report [30]. As evidenced by SEM analysis, at longer times more gold was deposited on the nanoparticle seeds resulting in more continuous, thicker nanowires. We attribute the increase in percentage yield to the increased physical continuity of the nanowires. Furthermore, we ascribe the reduced device resistance to the increase in cross-sectional area of the nanowires, in addition to the enhanced continuity caused by deposition. At all deposition times, the device-to-device resistance varies by 1–4 orders of magnitude. The large range of resistances observed at a given deposition time is attributed to the previously mentioned variations in nanowire thickness and connectivity associated with differences in density and arrangement of gold nanoparticle seeds along the gold-binding phage.

Zoom In Zoom Out Reset image size

As expected, the phage concentration also influenced sensor resistance. The median resistances for devices assembled with phage concentrations of 1 × 10⁸ pfu μl⁻¹, 3 × 10⁸ pfu μl⁻¹, and 5 × 10⁸ pfu μl⁻¹ and an electroless deposition of 3 min is shown in figure 5(b), in addition to the device yield at each concentration. As the concentration of phage increased, the median device resistance decreased. As previously discussed, the morphology of the sensors varies greatly with phage concentration. A low concentration of phage (1 × 10⁸ pfu μl⁻¹) resulted in devices with nanowires that produced relatively few physical connections bridging the electrode gap and higher concentrations of phage (3 × 10⁸ pfu μl⁻¹ and 5 × 10⁸ pfu μl⁻¹) produced a fairly dense and continuous network of gold nanowires across the entire gap. As the gold nanowire density increased the number of available conductive pathways also increased, causing the overall sensor resistance to be reduced. Furthermore, the reduced yield at the low concentration was attributed the low probability of creating a physical connection between the two electrodes. To briefly evaluate the stability of the as-assembled devices the resistance was re-measured after storage under ambient conditions for 4–5 months. The resistance of all re-measured devices increased, most by no more than a few percent but a handful by as much as an order of magnitude. More studies are required to explore the cause of the resistance change and the source of device-to-device variations.

To better understand the impact of the phage template and surface organics on sensor resistance, two-terminal measurements were also performed on ethanol-treated devices. Resistance distributions of as-assembled and ethanol-treated sensors fabricated using a 3 × 10⁸ pfu μl⁻¹ phage concentration and 3 min electroless deposition are shown in figure 6. As previously discussed, a large distribution of resistances was observed in the as-assembled devices with the peak of the distribution between 10⁸ and 10⁹ Ω. Following ethanol treatment the distribution remained broad; however the peak resistance decreased to 10−100 Ω. Nonconductive organic ligands act as energy barriers to charge transport via electron hopping in metal nanoparticle films and chains, often resulting in highly resistive materials [35–37]. The large decrease in resistance observed in ethanol-treated sensors was attributed to the removal of organic components such as peptides and viral template from gold nanoparticle surfaces resulting in reduced nanoparticle-to-nanoparticle energy barriers and enhanced charge transport [37–39].

Zoom In Zoom Out Reset image size

A characteristic real-time sensing response, defined as the change in resistance relative to the baseline resistance (ΔR/R₀), and a calibration curve with respect to gas concentration for as-assembled gold nanowire sensors are represented in figures 7(a) and (c). These devices were assembled with a phage concentration of 3 × 10⁸ pfu μl⁻¹ and electroless deposition time of 3 min. After exposure to dry air flow for 5 h with constant bias applied, a stable resistance baseline was established for each device. However, within the first minute the resistance dropped sharply and then continued to slowly decline; by 10 min of dry air flow, all devices were stable with less than 5% change in resistance over time. Sensor resistance increased with exposure to H₂S concentrations ranging from 0.025 to 0.5 ppm. This chemiresistive behavior was consistent with other reports of H₂S sensors composed of gold NP chains [4, 5, 16], and films [17] in which charge transport between neighboring nanoparticles or nanocrystals is impeded by adsorption of H₂S onto the gold surface causing resistance to increase. The sensitivity, defined as the slope of the linear region on the calibration curve, was 654%/ppm, within the linear range from 0 to 0.025 ppm. This sensitivity is more than an order of magnitude greater than other gold-based room temperature H₂S sensors [4, 13, 17]. Saturation of sensor response was observed at concentrations above 0.025 ppm. The lowest detection limit, defined as the concentration at which the response is 3 times the signal-to-noise ratio, was 2 ppb. This value is lower than that achieved by H₂S sensors composed of electrophoretically-assembled glycine-stabilized gold nanoparticles [4] and gold nanoparticles on 1-pyrenesulfonic acid-coated carbon nanotube templates [16], and comparable to the 3 ppb limit reported for sensors assembled from carbon nanotubes decorated with electrochemically deposited gold [13]. The response and recovery times, which are defined as the time to reach 95% of saturation resistance and the time for resistance to recover to 10% above the baseline resistance, were greater than 15 min and 30 min, respectively. Seventy percent recovery was observed within 30 min for all as-assembled devices, indicating desorption of the gas analytes from the surface. Faster recovery was observed with exposure to lower analyte concentrations such that devices exposed to 0.025 ppm H₂S experienced up to 70% recovery within 9 min. The sensors exhibited some cross-sensitivity to NH₃ and NO₂, toxic gases which are also frequently present in water treatment and mining industries. Figure 8 shows the response of the devices to 0.5 ppm of H₂S NH₃ and NO₂.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To highlight the importance of the gold-binding peptides and viral template on device behavior, the sensing performance of ethanol-treated devices, in which organics were removed, was also analyzed. A characteristic real-time sensing response and a calibration curve for ethanol-treated gold nanowire sensors are shown in figures 7(b) and (d). Ethanol-treated devices increased in resistance with exposure to H₂S concentrations ranging from 0.025 to 40 ppm. These sensors maintained a linear response up to 0.5 ppm with a sensitivity of 6%/ppm, and a 6 ppb lower limit of detection. Like the as-assembled sensors, response times were greater than 15 min; however, a considerably slower initial response rate was observed for the ethanol-treated than for the as-assembled devices. No recovery was observed in these sensors. This irreversible behavior suggests the continued presence of analyte on the gold nanowire sensor surface. Similar behaviors have been reported at room temperature for H₂S sensors based on ligand-free gold thin films [17] and citrate-coated gold nanoparticles [5, 16], due to strong Au–S affinity. Recovery was only attained in these devices at temperatures >140 ^°C.

The sensing behaviors of as-assembled and ethanol-treated sensors were notably different. As-assembled devices, in which the viral template and gold-binding peptides were intact, exhibited a slightly decreased lower detection limit (3×), decreased dynamic range (20×), and a substantial sensitivity increase (100×) compared to ethanol-treated devices. Moreover, the analyte adsorption and desorption rates of the as-assembled sensors were markedly faster than the ethanol-treated sensors. This behavior suggests that, although the viral template with the gold-binding peptide was intended primarily for structural assembly of nanowires, it also played an active role in device–analyte interaction.

Biological molecules incorporate a range of chemical moieties which enable extraordinary diversity and specificity for in vivo processes. These same attributes have proven useful in discrete sensor and electronic nose applications. Biological molecules including proteins [40, 41], peptides [42–44], antibodies [45], and DNA [46] have been successfully integrated into gas or vapor phase sensors to impart analyte specificity. Of particular relevance to these studies is the use of peptides. For example, piezoelectric-based sensors have utilized molecular modeling in conjunction with oligopeptide mimics of human olfactory [41] and dioxin [42] receptors to detect vapor phase acetic acid [41], ammonia [41], and dioxin [42]. The same oligopeptides have also been used to sensitize silicon nanowire-based chemiresistive sensors [44]. Furthermore, peptides identified with a combinatorial phage display library [47], in a process very similar to that used to select the gold-binding peptide found in these studies, have been used for detection of trinitrotoluene (TNT) with both fluorescence quenching [48] and conductance-based field effect transistor (FET) [49] device platforms. In the former, the entire virus with high copy peptide fusions was incorporated into the device [48]. Indirect evidence exists that, indeed, the gold-binding peptide may have an affinity for sulfur found in H₂S gas. Specifically, the pVIII major coat protein of the unmodified or wild-type M13 virus has been recently reported to display an affinity for sulfur [50]. The carboxyl groups associated with acidic amino acids such as aspartate (D) and glutamate (E) found within the wild-type pVIII coat donate electrons to sulfur, creating S–O and C–S bonds [50]. Like the wild-type M13, the gold-binding phage template displays acidic amino acid (aspartate, D) on its pVIII coat and may, therefore, share this attraction to the H₂S analyte, potentially increasing device sensitivity and response rate. In addition, amine groups have demonstrated an affinity for H₂S [51–53]. Each of 2700 copies of the pVIII protein found along the length of the viral template is terminated with an amine group, which may again cause increased sensitivity and enhanced response rate. Moreover, very rapid initial response rates were observed in H₂S sensors fabricated from glycine-stabilized gold nanoparticles on which amine groups were also displayed [4].

Alternatively, the gold-binding peptide fusion and/or phage template may interact with the H₂S analyte or its decomposed, adsorbed products without exhibiting a specific affinity. The gold-binding phage template displays an 8-mer peptide known to preferentially bind to and reduce gold in solution [26]. Specifically, hydroxyl-containing amino acids such as serine (S) can act as anchoring sites for peptide adsorption to Au surfaces [26, 54, 55]. The presence of the high affinity gold-binding peptide, rich in serine (S), on the surface of the as-assembled devices, may sterically hinder or decrease available adsorption sites and weaken analyte binding to the gold surface resulting in reduced dynamic range and enabling device recovery. As described here, a few potential mechanisms, which may act alone or in combination, could account for the active role of the viral template within the H₂S sensor. Yet, further studies are necessary to fully understand the function of the viral template and gold-binding peptides in sensor performance, as well as in adsorption and desorption kinetics.