Converting Silver Electrodes into Porous Gold Counterparts: A Strategy to Enhance Gas Sensor Sensitivity and Chemical Stability via Electrode Engineering

Abstract

This chapter describes a strategy for sensitivity and chemical stability enhancement of chemiresistive gas sensors via electrode engineering. In this strategy, flexible chemiresistive gas sensors were fabricated by uniformly depositing functionalized semiconducting carbon nanotubes (CNTs) on a polyimide substrate via a novel layer-by-layer wet chemical method, followed by inkjet printing fine-featured silver interdigitated electrodes (IDEs) on the substrate. The electrode engineering was realized by converting the inkjet-printed IDEs into their highly porous and chemically stable gold counterparts via a mild and facile two-step process, with the substrate-IDE adhesion retained. As a proof-of-concept demonstration, a diethyl ethylphosphonate (DEEP, a simulant of the nerve agent sarin) sensor equipped with inkjet-printed dense silver IDEs was converted into its counterpart equipped with highly porous gold IDEs. The resulting gold-electrode gas sensor exhibited sensitivity to DEEP of at least fivefold higher than a similar sensor electrode with the dense silver IDEs. The sensitivity enhancement was probably due to the catalytic activity of the resulting gold IDEs, as well as the creation of the nano−/micro-scale pores in the gold IDEs that increased the Schottky contacts between the gold IDEs and the semiconducting CNTs.

1. Introduction

Metal nanoparticle-based inkjet inks have been formulated from a number of metals such as silver [1], gold [2, 3], copper [4, 5], and nickel [6]. Among these metals silver is the best heat and electricity conductor [7] and silver nanoparticle (SNP)-based inks not only are the most commonly used with a well-settled technology but also show the most commercial significance, as indicated by their highest sales volume among all metal-based inks [8]. However, due to some pollutants such as carbonyl sulfide (OCS) and hydrogen sulfide (H₂S) presented in air, silver (especially silver nanoparticles) tarnishes under ambient and even dry conditions [9, 10], resulting in significantly reduced conductivity [11]. Particularly, the concerns on poor conductivity of inkjet-printed silver traces in an electrochemical device are severe when used in aqueous environments, due to their oxidation and degradation under an applied electrical potential [3]. To prevent/minimize such oxidation−/degradation-based tarnishing, some passivation treatments, such as coating the silver structures with nickel [12] or polymers [13], have been successfully attempted.

In contrast to silver, gold is one of the most stable and inert chemical elements, the most ductile and malleable of all metals, and in the meantime an excellent electricity conductor. Accordingly, gold is preferable to silver for a number of applications. For instance, gold is preferred to silver to make printed circuit board electrodes for medical use (such as electrodes for high-resolution gastrointestinal electrical mapping), due to the fact that the oxidizing agents used to sterilize the electrodes at a low temperature, such as hydrogen peroxide and ozone, would oxidize silver but not gold [14, 15]. As another reported example, gold prevailed over silver as the electrode material for zinc oxide-based chemiresistive CO and NO₂ sensors for high sensitivity and short recovery time, due to that fact that gold was resistant to the poisoning or oxidizing of the target gases but silver was not [16].

Increasing evidence has shown that the device performance can be changed significantly by the contact geometry between the metal electrodes and the semiconducting materials. The sensor performance can sometimes be radically affected by the contact resistance formed in the electrode-semiconductor interface [17, 18], which has been conventionally ignored [19]. It has been shown that a Schottky barrier was formed in the contacts of semiconducting carbon nanotubes (CNTs) with gold [20] and that an increase in the Schottky contact area between semiconducting CNTs and gold/chromium electrodes resulted in a radical increase in the sensitivity of some biosensors [21].

Gold nanoparticle-based inks, however, have not been used as commonly as SNP inks. Compared with silver nanoparticles, gold nanoparticles are extremely expensive. Meanwhile, the relative immature techniques for preparing and inkjet printing gold nanoparticle-based inks are also a concern. Additionally, due to the fact that some of the most commonly used flexible substrates (such as polyethylene, terephthalate, and polyethylene naphthalate films and a number of papers such as copy, filter, and photo papers) have a relatively low maximum working temperature, the high sintering temperature (>190°C) needed for gold nanoparticle-based inks has restrained their applications in printing of flexible electronic devices. In contrast, a temperature of as low as 120°C can be used to sinter SNP-based inks, which is well suitable for most commonly-used flexible substrates.

In order to inkjet-print reliable, low-cost and high-performing flexible electronic devices, it is desirable to make use of the exceptional chemical stability of gold nanoparticles, as well as the well-developed formulation technology and low sintering temperature of SNP inks. In addition, compared with their dense counterparts, highly porous gold electrodes create much more Schottky contacts with a semiconducting sensing material (such as semiconducting CNTs), which drastically benefits sensing applications. To meet all these desires, one way that can be done is to inkjet-print an SNP ink, followed by sintering the resulting silver traces at a low temperature (e.g., 120°C) for desired conductivity, and finally chemically converting the resulting dense silver traces into their porous gold counterparts. Some one-step processes to chemically convert silver structures into their gold counterparts have been reported [22, 23, 24]. These processes, however, were not only performed under relatively harsh conditions (at 100°C and in an aqueous solution) but also not able to enhance porosity in the resulting gold components. For applications in printed electronic devices, a serious concern arises on whether the adhesion between the silver patterns and the substrate can survive the harsh conversion conditions. As a matter of fact, for inkjet-printed devices/structures, the trace substrate adhesion has always been a constant concern, especially when they use a smooth-surfaced substrate (polyimide, PET and PEN films, silicon wafer, and glass plates.), and/or have to work in relatively harsh environments (such as high humidity, water- or organic solvent-based solutions, elevated temperatures, vibration, and bending). Surprisingly, as shown in a recent report, even buffers were able to kill the adhesion between a plastic substrate (PET or PEN) and inkjet-printed silver interdigitated electrodes (IDEs) [3], let alone chemically turning silver patterns printed on a plastic substrate to their gold counterparts in an aqueous solution at a high temperature.

Among the most commonly used flexible substrates for inkjet-printed electronic devices, Kapton^® films exhibit excellent flexibility and mechanical robustness, as well as exceptional thermal and chemical stability. Some types of Kapton^® films, such as Kapton^® HN and HA films, have a slip additive incorporated in the polyimide matrix [25] to enhance their mechanical properties and reduce their resistance to sliding over themselves or parts of converting equipment. For example, a Kapton^® 500 HN polyimide film (Figure 1a) has a slip additive, which has been identified as CaCO₃ particles [1]. The additive particles in a 500 HN film made its surface look granular under an optical microscope (Figure 1b), which seemingly contributed to increased surface roughness. However, as shown in Figure 1c, the surface of a Kapton^® 500 HN film was actually very smooth, with an arithmetic (Ra) and quadratic (Rq) mean surface roughnesses of 0.67 and 0.89 nm, respectively [1].

Figure 1.

Optical (a), optical microscopic (b) [26]), and atomic force microscopic (c) [1]) images of a blank Kapton^® 500 HN polyimide film (panels b) and c) were licensed under creative commons attribution 4.0 international license and with permission from the Royal Society of Chemistry, respectively).

Apparently, one solution to ensure the IDE-substrate adhesion to survive the silver-to-gold conversion is to reduce the harshness of the conversion conditions and/or enhance the adhesion by performing surface modification to the substrate prior to inkjet printing.

This chapter describes a strategy to enhance the adhesion between the Kapton^® 500 HN polyimide substrate and inkjet-printed silver IDEs by chemically surface-modifying the substrate prior to inkjet printing, and promote the sensitivity and the chemical stability of flexible chemiresistive gas sensors through altering the material and the porosity of their electrodes. To begin with, semiconducting single-wall carbon nanotubes (SWCNTs) complexed with a chemoselective compound (selector), which functioned as the sensing material of the sensors, were chemically deposited, in a layer-by-layer fashion, on a piece of Kapton^® 500 HN polyimide film with a home-developed chemical process. Secondly, an array of chemiresistive gas sensors was fabricated by inkjet printing fine-featured silver IDEs on the resulting CNT-terminated polyimide substrate. Finally, the silver IDEs were chemically converted into their highly porous and chemically stable gold counterparts under ambient conditions via a mild two-step process, without losing the IDE-substrate adhesion. After these processes, the chemical stability of the sensor IDEs and the contact area of the IDEs with the sensing element were drastically enhanced. Additionally, due to the fact the sensing material (i.e., the selector-functionalized CNTs) was chemically and uniformly deposited on the substrate, the individual difference among all the sensors fabricated on the same piece of substrate was minimized, thus allowing for a subsequent fair performance comparison between a sensor with the original dense silver electrodes and one with the converted highly porous gold electrodes.

To exhibit the utility of the strategy mentioned above, two sensors of different types, one electrode with the original inkjet-printed dense silver electrodes, and the other with the converted porous gold electrodes, were placed side by side in a sensor box and exposed to the vapor of diethyl ethylphosphonate (DEEP) generated from a well-calibrated gas generator. The sensing profiles of the two sensors were then compared. DEEP is a simulant of sarin (a G-type nerve agent with a military designation of GB). Sarin is one of the most toxic of the known chemical warfare agents and has been used multiple times by terrorist organizations and rogue states to attack civilians and military troops, resulting in severe mass casualties [27, 28, 29]. Exposure to sarin can cause death in minutes. Detecting sarin or its simulants to provide early warnings is becoming more and more of a concern for both civilian and military personnel. This work demonstrated that a sensor with the converted highly porous gold electrodes was at least fivefold more sensitive to DEEP than its counterpart with the original inkjet-printed dense silver electrodes. The possible mechanism responsible for the sensitivity enhancement is discussed.

2. Fabrication of chemiresistive sensors

2.1 Functionalization of semiconducting SWCNTs

Functionalization of SWCNTs with a selector was realized by immobilizing a hexafluoroisopropanol group-containing compound to semiconducting SWCNTs (a hexafluoroisopropanol group has been shown to absorb the target analyte DEEP via hydrogen bonding [1]). Specifically, a small sheet of semiconducting SWCNTs was immersed in dimethylformamide (DMF) solvent, followed by a gentle sonication with a probe sonicator to break the sheet into small pieces. A short centrifugation (4, 500 x g, 1 min) was conducted, and the resulting SWCNT pellets were collected and then incubated for 2 hours in an incubator shaker (25°C, 90 rpm) with a selector solution (5 mg/ml solution of 2-(2-hydroxy-1, 1, 1, 3, 3, 3-hexafluoropropyl)-1-naphthol in DMF). This incubation process allowed for the binding of the selector molecules to the SWCNTs via π-π interaction. A centrifugation (4500 x g, 5 min) was performed and the supernatant (i.e., unbound selector in DMF) was removed. The pellets were collected, washed three times with DMF solvent (with a 4500 x g/5 min centrifugation between each wash), and finally re-suspended in DMF. The suspension was further sonicated multiple times with the probe sonicator until a homogeneous solution was obtained.

2.2 Functionalization of substrate

A Kapton^® 500 HN polyimide sheet was cleaned by sonication first with an aqueous suspension of powdered precision cleaner for 10 minutes and then with acetone for 10 minutes in an ultrasonic cleaner. A layer-by-layer wet chemical process, which consisted of multiple steps as shown in Figure 2, was then used to deposit the SWCNT-selector particles (i.e., SWCNTs functionalized with the selector 2-(2-hydroxy-1, 1, 1, 3, 3, 3-hexafluoropropyl)-1-naphthol) on the cleaned polyimide substrate under ambient conditions. In brief, the following steps were performed with three rinses with DMF after each step: (1). The cleaned polyimide was exposed to a solution of 10 wt% tris (2-aminoethyl) amine in DMF for 1 hour (step (a) in Figure 2) to introduce –NH₂ groups to the substrate surface. (2). The resulting amine-functionalized substrate was exposed to a solution of 5 mg/ml 1-pyrenebutyric acid N-hydroxysuccinimide ester (PBSE) in DMF for 1 hour (step (b) in Figure 2) to allow for the covalent bonding of PBSE to the polyimide via the nucleophilic attack reaction depicted in Figure 3, resulting in pyrene-terminated polyimide substrate. (3). The pyrene-terminated substrate was exposed to the DMF-based SWCNT-selector solution for 1 hour (step (c) in Figure 2) to allow for the binding of SWCNT-selector particles to the polyimide substrate via the π-π interaction between the SWCNTs and the pyrene groups on the substrate (this step deposited the first layer of SWCNT-selector particles on the polyimide substrate). (4). The resulting structure was exposed to the PBSE solution for 1 hour (step (d) in Figure 2) to immobilize PBSE to the substrate via the π-π interaction between the SWCNTs and the pyrene groups in PBSE. (5). The resulting PBSE-functionalized structure was then exposed to a DMF-based 10 wt% solution of tris (2-aminoethyl) amine for 1 hour (step (e) in Figure 2) to introduce –NH₂ groups to the substrate via the nucleophilic attack reaction between the –NH₂ group (s) in tris (2-aminoethyl) amine and the succinimidyl ester end of PBSE (as depicted in Figure 4). (6). The resulting amine-terminated structure was then incubated with the PBSE solution for 1 hour (step (f) in Figure 2) to allow for the introduction of pyrene groups to the substrate surface. (7). The resulting structure was then exposed to the SWCNT-selector solution for 1 hour (step (g) in Figure 2) to allow for the binding of the second layer of the SWCNT-selector particles to the substrate. 8). Steps (d), (e), (f), and (g) in Figure 2 were repeated eight times to deposit more layers of SWCNT-selector particles on the substrate.

Figure 2.

Schematic of layer-by-layer deposition of the SWCNT-selector particles on a flexible Kapton^® 500 HN polyimide substrate [30] (licensed under creative commons attribution 4.0 international license).

Figure 3.

Reaction of 1-pyrenebutyric acid N-hydroxysuccinimide ester (PBSE) with the amine groups on the amine-functionalized Kapton^® 500 HN polyimide substrate, which resulted in pyrene-terminated polyimide substrate.

Figure 4.

Schematic of the reaction of tris (2-aminoethyl) amine with PBSE-terminated polyimide substrate, which resulted in amine-terminated polyimide substrate.

2.3 Inkjet printing of silver-based interdigitated electrodes and chemical conversion of dense silver electrodes into porous gold counterparts

Silver-based IDEs were inkjet-printed with a drop-on-demand piezoelectric inkjet printer (DMP-2831, Fujifilm Dimatix, Inc., Santa Clara, CA, USA) for five passes on the functionalized polyimide substrate with a commercial silver nanoparticle-based ink, followed by a 120°C/3 h sintering process. Both the IDE finger width and the spacing between two adjacent fingers were 100 μm.

The conversion of inkjet-printed dense silver electrodes of a sensor into their porous gold counterparts was conducted under ambient conditions with a two-step process. In the first step, the sensor was incubated with a 3 wt% solution of gold(III) chloride in diluted HCl for 1 hour in an incubator shaker with a temperature of 25°C and a speed of 90 rpm, followed by rinsing three times with DI water. In the second step, the resulting structures were exposed overnight to an aqueous saturated solution of sodium chloride in the incubator shaker (25°C/90 rpm), followed by rinsing with DI water and drying with flowing air.

Figure 5a shows an optical image of an array of inkjet-printed sensors on a piece of functionalized Kapton^® polyimide substrate, while Figure 5b shows a scanned image of one of the sensors in the array. The electrodes after the two-step conversion (Figure 5c) were morphologically similar to those before the conversion (Figure 5b). However, the electrodes before and after the two-step conversion exhibited the typical silver and gold colors, respectively.

Figure 5.

Optical and scanned images of inkjet-printed sensors before and after the two-step wet chemical conversion of the silver IDEs into their porous gold counterparts. (a) Optical image of an array of sensors before the conversion. (b) and (c) scanned images of a single sensor before and after, respectively, of the conversion [30] (licensed under creative commons attribution 4.0 international license).

Scanning electron microscopy (SEM) images of an inkjet-printed sensor (before the silver-to-gold conversion), and the energy-dispersive X-ray spectroscopy (EDX) pattern of its silver IDEs are shown in Figure 6. Under the scanning electron microscope, the silver nanoparticles were densely packed and virtually spherical in shape with a diameter of ~150 nm after a 120°C/3 hour annealing process (Figure 6b). The SWCNT-selector particles between two adjacent IDE fingers were randomly and virtually evenly distributed on the surface of the polyimide substrate (Figure 6c). The EDX analyses on the silver IDEs show the presence of the elements silver and carbon (Figure 6d). It is worth mentioning that the carbon peak in the Figure 6d originated from the carbon-containing functionalized polyimide substrate and the carbon film that was sputter-coated on the sensor to make the sample conductive prior to the SEM and EDX analyses, not from the silver IDEs.

Figure 6.

SEM and EDX analyses of a sensor with inkjet-printed silver-based IDEs. (a) Low magnification SEM image of the sensor. (b) High magnification SEM image focusing on the silver nanoparticles in an IDE finger. (c) High magnification SEM image focusing on the individual SWCNT-selector particles between two adjacent fingers. (d) EDX pattern of the specimen shown in panel b) [30] (licensed under creative commons attribution 4.0 international license).

The incubation of an inkjet-printed sensor with the gold(III) chloride solution brought in a drastic change in the morphology of its IDEs. That is, the particles turned irregular in shape and much larger than the starting silver nanoparticles (Figure 7a). The elements gold, silver, and chlorine were detected in the resulting IDEs (Figure 7c). The subsequent incubation with a saturated sodium chloride solution brought in another drastic change in the morphology of the IDEs, the particles became significantly shrank in size and much more loosely packed, resulting in highly porous IDEs (Figure 7b). Elemental analyses of the porous IDEs showed that, compared with the elemental composition of the starting silver-based IDEs (Figure 6d), the elements silver and chlorine disappeared but gold was retained (Figure 7d), which means that the incubation with the saturated sodium chloride solution and the subsequent rinsing selectively removed the side product(s) of the silver-to-gold converting reaction. Again, the carbon peak in Figure 7d came from the functionalized polyimide substrate and the sputter-coated carbon film on the porous IDEs.

Figure 7.

SEM and EDX analyses of the sensor IDEs at their different stages after the silver-to-gold conversion. (a) SEM image of the IDEs after the silver-to-gold conversion but before the selective removal of the side product(s). (b) SEM image of the IDEs after both the silver-to-gold conversion and the side product removal. (c) and (d) EDX analyses of the specimens shown in a) and b), respectively [30] (licensed under creative commons attribution 4.0 international license).

Figure 8 shows the X-ray diffraction (XRD) patterns of a sensor at its different fabrication stages. Unfortunately, the XRD peaks of gold and silver overlapped in the 2θ range scanned in this work (i.e., 20° - 50°) [31, 32], and accordingly, XRD analyses performed on the sensor were unable to differentiate silver from gold. Fortunately, EDX analyses can easily distinguish the two elements. The following conclusions can be reached when combining the information drawn from the EDX (Figures 6d, 7c, and d) and XRD (Figure 8) analyses of sensor IDEs:

Figure 8.

XRD analyses of (a) the polyimide substrate functionalized with SWCNT-selector particles. (b) the functionalized polyimide substrate with inkjet-printed IDEs that have been subjected to the 120°C/3 hour annealing process. (c) the functionalized polyimide substrate with inkjet-printed IDEs that have been subjected to the 120°C/3 hour annealing process and the silver-to-gold conversion process. (d) the functionalized polyimide substrate with IDEs that have been subjected to the 120°C/3 hour annealing process, the silver-to-gold conversion process, and the side product removal process [30] (licensed under creative commons attribution 4.0 international license).

1. The inkjet-printed IDEs (before the silver-to-gold conversion) were made of silver (as shown in Figures 6d and 8b).

2. The gold(III) chloride solution converted silver into gold and a side product, silver chloride (as shown in Figures 7c and 8c).

3. The saturated sodium chloride solution selectively dissolved silver chloride, leaving porous gold behind (Figures 7d and 8d).

With these characterization results, it can be concluded that the starting inkjet-printed dense silver IDEs were converted into their porous gold counterparts by the two-step process under ambient conditions.

3. Characterization of chemiresistive sensors

3.1 Assessments of IDE-substrate adhesion

Visual inspection and Scotch^®-tape peel testing were utilized to assess the adhesion between the metal traces (dense silver or porous gold IDEs) and the functionalized polyimide substrate. Visual inspection was performed from different angles while slowly finger-bending a sensor to a radius of curvature of ~1 cm, first in tension and then in compression. The strength of the adhesion of the silver or gold IDEs to the functionalized polyimide substrate was assessed by Scotch^®-tape peel testing. Briefly, the back of a sensor was glued or taped to a smooth- and flat-surfaced plastic piece, and then the adhesive side of a piece of Scotch^® magic tape (3 M company) was finger-pressed firmly against the front of the sensor (Figure 9a and c). Finally, the tape was peeled off slowly from the sensor at an angle of ~90°.

Figure 9.

Scotch^®-tape peel testing on the sensors with the starting dense silver IDEs and with the porous gold IDEs. (a) and (c) optical images of the sensors with the dense silver IDEs and with the porous gold IDEs, respectively, that have been firmly stuck to a piece of scotch^® magic tape. (b) and (d) optical images of the sensors with the silver IDEs and with the porous gold IDEs, respectively, after the removal of the tape [30] (licensed under creative commons attribution 4.0 international license).

By visual inspection, no detachment was observed between the functionalized polyimide substrate and the dense silver or the porous gold IDEs, which was an indication that the adhesion between the inkjet-printed silver IDEs and the polyimide substrate survived the silver-to-gold conversion and the subsequent removal of the side product silver chloride. For Scotch^®-tape peel testing, as shown in Figure 9b, with the tape taken off, the dense Ag IDEs on the polyimide substrate were essentially unimpaired (there was only one very small silver particle with a diameter of ~0.2 mm taken by the tape). By contrast, the porous gold IDEs seemed to be delaminated by the removal of the type. This is, quite a number of top layer gold residues originally on the polyimide substrate were taken by the tape, with the gold nanoparticles that remained on the polyimide substrate still forming an essentially intact IDE pattern (Figure 9d).

3.2 Gas sensing

A DEEP permeation tube (KIN-TEK Laboratories, Inc.), a FlexStream™ Gas Standards Generator (KIN-TEK Laboratories, Inc.), and nitrogen gas with a flow rate of 500 sccm were used to generate the 2.0 ppm DEEP vapor stream. A home-developed sensing system automated with LabVIEW-based programs was used to monitor the real-time changes in the electrical resistance of the sensors in the sensor box. A schematic diagram of the gas sensing setup is shown in Figure 10. In a typical sensing trial, two sensors, one with inkjet-printed dense silver IDEs and the other with porous gold IDEs, were accommodated side by side in the sensor box. Upon exposing the sensors to either the carrier gas nitrogen or the 2.0 ppm DEEP balanced with nitrogen, their electrical resistance changes with time were automatically recorded by the system.

Figure 10.

Schematic diagram of the setup for the sensing of DEEP vapor.

As revealed by Figure 11, neither the sensor electrode with the original dense silver IDEs nor the sensor electrode with the porous gold IDEs was responsive to the carrier gas nitrogen (0–10 min time range) that was launched prior to the onset of the DEEP vapor. Upon launching the 2.0 ppm DEEP vapor, the electrical resistance of both sensors began to increase and continued increasing until the DEEP release was stopped. During the duration of DEEP release (10–73 min time range), there was a much faster increase in the electrical resistance of the sensor electrode with the porous gold IDEs than its counterpart with the dense silver IDEs. At the end of the DEEP release, the relative sensitivity (S) of the sensor with the porous gold IDEs reached ~47% (Figure 11 solid line), compared to only ~9% for the sensor with the dense silver IDEs (Figure 11 dash-dot line). That is to say, the conversion of the dense silver IDEs of a sensor into their porous gold counterparts enhanced the sensitivity of the sensor by more than fivefold. The relative sensitivity S is defined as

Figure 11.

Typical sensing profiles of the SWCNT-based sensors with the inkjet-printed dense silver IDEs (dash-dot line) and with the porous gold IDEs (solid line) upon exposure to the carrier gas nitrogen (0–10 min and 73–100 min) and the 2.0 ppm DEEP vapor (10–73 min).

$$\text{S}=\frac{R-R_0}{R_0}$$E1

where R₀ and R denoting the electrical resistances of a sensor right before and at a particular time after, respectively, the sensor was exposed to the DEEP vapor. Once the release of the DEEP vapor was stopped, the electrical resistance of both sensors began to decrease slowly, indicating the slow and partial recovery of the sensors.

4. Discussion

It has been reported that the following chemical reaction (Eq. (2)) took place during a one-step conversion of silver nanostructures into their gold counterparts (with no extra pores created) at an elevated temperature of 100°C [22, 23, 24]:

$$\text{3Ag}\left(\text{s}\right)+{\text{HAuCl}}_4\left(aq\right)=>Au\left(\text{s}\right)+\text{3AgCl}\left(aq\right)+HCl\left(aq\right)$$E2

where the silver chloride formed in the reaction was completely dissolved under the experimental conditions. Different from these previous reports, in the current work the silver-to-gold conversion was performed at room temperature. Based on the EDX (Figures 6d, 7c, and d) and XRD (Figure 8) analyses of the IDEs at different stages, it can be concluded that at room temperature, a silver-to-gold converting reaction similar to the one presented in Eq. (2) took place, but the silver chloride was produced in solid form. That is to say, in the current work the following reaction (Eq. (3)) was responsible for the room temperature silver-to-gold conversion:

$$\text{3Ag}\left(\text{s}\right)+{\text{HAuCl}}_4\left(aq\right)=>Au\left(\text{s}\right)+\text{3AgCl}\left(\text{s}\right)+HCl\left(aq\right)$$E3

The fact that the silver chloride in Eq. (3) was in solid form and incorporated in the gold matrix turned out to be beneficiary for sensitivity enhancement since the nano−/micro-scale pores in the gold IDEs were mainly created by the selective dissolution of the silver chloride that was performed following the silver-to-gold conversion. In addition to the silver chloride dissolution, the size difference between silver and a gold atom, which had a Van der Waals radius of 172 and 166 pm, respectively, also contributed to the pore creation in the gold IDEs, since a slight shrinkage occurred in the atomic size when a silver atom was converted into a gold atom. It was difficult, however, to experimentally determine the porosity of the porous gold IDEs, since the gold IDEs were not free-standing but on a polyimide film. Nevertheless, a rough estimation of the porosity of the gold IDEs can be calculated theoretically based on Eq. (3) and some inherent properties of the elements gold and silver. Assuming the starting inkjet-printed silver IDEs were fully dense, the total pore volume of the porous gold IDEs is the sum of the pore volumes created by the dissolution of the silver chloride and by the atomic shrinkage when three moles of silver was chemically converted into one mole of gold (based on the stoichiometry of Eq. (3)). With the densities of gold and silver being 19.32 and 10.49 g/cm³, respectively and their atomic masses 196.97 and 107.87, respectively, the porosity of the porous gold IDEs was calculated to be about 67%.

The Scotch^®-tape peel testing delaminated the porous gold IDEs, but the adhesion between the substrate and the porous gold IDE seemed unimpaired by the peeling-off of the tape. This means the creation of nano−/micro-scale pores in the gold IDEs weakened the connection between the gold nanoparticles.

The mechanism for the sensitivity enhancement due to the change in the electrode material, and geometry is still not well understood. Based on the experimental observations described in this work and some relevant literature, it is speculated that the “electrode effects” and the “porosity effects” were responsible for the significant sensitivity enhancement. The “electrode effects” refer to the fact that the catalytic activity of the gold electrodes contributes to the enhanced sensitivity, even though such effects are not conventionally considered to contribute to the electrical resistance change of chemiresistive sensors [33]. An unequivocal agreement has not been reached so far regarding how the sensing behavior of a chemiresistive sensor is affected by the material that its electrodes are made of [34], but more and more evidence shows that different electrode metals react with an target analyte differently [16, 35, 36]. Sensor electrodes that are made of a noble metal (such as palladium and gold) can act as catalysts and contribute to the overall electrical conductance and the sensor sensitivity [35, 37, 38, 39, 40]. “Porosity effects” refer to the fact that increased number of nano−/micro-scale pores in the electrodes lead to increased Schottky contacts that contribute to the enhanced sensitivity. In the current work, the Schottky contacts between the porous gold electrodes and the semiconducting SWCNTs probably contributed more to the overall electrical resistance change of a sensor than the bulk SWCNTs. As a matter of fact, there has been increasing evidence indicating that the interface between metal electrodes and semiconducting CNTs might contribute significantly to the performance of electronic devices such as gas sensors [18] and transistors [41].

There are only a small number of reports in the literature on DEEP detection. Most of these reports utilized mass spectrometry, which featured high sensitivity and selectivity, to detect DEEP. For example, active capillary plasma ionization coupled to an ion trap mass spectrometer [42] and selected ion flow tube mass spectrometry (SIFT-MS) [43] have been used to detect DEEP, with a detection limit of 0.15 ppb and 45 pptv, respectively. Additionally, DEEP vapor of 135 ppm has been detected with optical reflectivity spectroscopy [44]. Reports on detection of DEEP vapor with a chemiresistive sensor have been very scarce. As an example, a flexible chemiresistive sensor based on non-functionalized reduced graphene oxide has been used to sense 2.0 ppm DEEP vapor with a low relative sensitivity of 5.2% [1]. In terms of sensitivity to DEEP vapor, chemiresistive sensors are apparently not as good as mass spectrometry. Nevertheless, mass spectrometry is associated with tedious sample preparation, solvent management, and bulky equipment. In contrast, chemiresistive sensors described in this work are ultra-lightweight, flexible, miniature-sized, and wearable, do not require sample preparation or solvents, and can be readily integrated into wireless sensing platforms to realize wireless detection.

Based on the generic nature of the facile and mild method described in this work to increase the electrode chemical stability and Schottky contacts of a sensor, the method can be readily applied to other semiconductor-based chemiresistive sensors for the purpose of sensitivity and chemical stability enhancement.

5. Conclusions

Functionalized SWCNTs were virtually evenly deposited on a polyimide film through a novel layer-by-layer wet chemical method. Flexible chemiresistive sensors were then fabricated by inkjet printing fine-featured silver IDEs on the resulting functionalized substrate. Sensors with highly porous (with a calculated porosity of ~67%) and chemically stable gold IDEs were fabricated via a facile and ambient condition two-step wet chemical process. The adhesion between the inkjet-printed silver IDEs and the substrate survived the two-step dense silver to porous gold conversion process, but the creation of the nano−/micro-scale pores in the resulting gold IDEs weakened the connection between gold nanoparticles.

The dense silver to porous gold conversion turned to be very efficient in enhancing the sensitivity of the sensors to DEEP (a simulant of the nerve agent sarin). A sensor equipped with the converted porous gold IDEs exhibited a sensitivity to DEEP of more than five times higher than a similar sensor equipped with the original dense silver IDEs. The drastic sensitivity increase was probably due to the catalytic activity of the gold IDEs and the increased Schottky contacts between the porous gold electrodes and the semiconducting SWCNTs.

The electrode engineering idea described in this work is generic and readily applicable to other semiconductor-based chemiresistive sensors to enhance their sensitivity and chemical stability.

Conflict of interest

The author declares no conflict of interest.