A Nose for Hydrogen Gas: Fast, Sensitive H2 Sensors Using

Article pubs.acs.org/accounts

A Nose for Hydrogen Gas: Fast, Sensitive H2 Sensors Using Electrodeposited Nanomaterials Reginald M. Penner* Department of Chemistry, University of California, Irvine, California 92697, United States

CONSPECTUS: Hydrogen gas (H2) is odorless and flammable at concentrations above 4% (v/v) in air. Sensors capable of detecting it rapidly at lower concentrations are needed to “sniff” for leaked H2 wherever it is used. Electrical H2 sensors are attractive because of their simplicity and low cost: Such sensors consist of a metal (usually palladium, Pd) resistor. Exposure to H2 causes a resistance increase, as Pd metal is converted into more resistive palladium hydride (PdHx). Sensors based upon Pd alloy films, developed in the early 1990s, were both too slow and too insensitive to meet the requirements of H2 safety sensing. In this Account, we describe the development of H2 sensors that are based upon electrodeposited nanomaterials. This story begins with the rise to prominence of nanowire-based sensors in 2001 and our demonstration that year of the first nanowirebased H2 sensor. The Pd nanowires used in these experiments were prepared by electrodepositing Pd at linear step-edge defects on a graphite electrode surface. In 2005, lithographically patterned nanowire electrodeposition (LPNE) provided the capability to pattern single Pd nanowires on dielectrics using electrodeposition. LPNE also provided control over the nanowire thickness (±1 nm) and width (±10−15%). Using single Pd nanowires, it was demonstrated in 2010 that smaller nanowires responded more rapidly to H2 exposure. Heating the nanowire using Joule self-heating (2010) also dramatically accelerated sensor response and recovery, leading to the conclusion that thermally activated H2 chemisorption and desorption of H2 were rate-limiting steps in sensor response to and recovery from H2 exposure. Platinum (Pt) nanowires, studied in 2012, showed an inverted resistance response to H2 exposure, that is, the resistance of Pt nanowires decreased instead of increased upon H2 exposure. H2 dissociatively chemisorbs at a Pt surface to form Pt−H, but in contrast to Pd, it stays on the Pt surface. Pt nanowires showed a faster response to H2 exposure than Pd nanowires operating at the same elevated temperature, but they had a surprising disadvantage: The resistance change observed for Pt nanowires was exactly the same for all H2 concentrations. Electron surface scattering was implicated in the mechanism for these sensors. Work on Pt nanowires lead in 2015 to the preparation of Pd nanowires that were electrochemically modified with thin Pt layers (Pd@ Pt nanowires). Relative to Pd nanowires, Pt@Pd nanowires showed accelerated response and recovery to H2 while retaining the same high sensitivity to H2 concentration seen for sensors based upon pure Pd nanowires. A new chapter in H2 sensing (2017) involves the replacement of metal nanowires with carbon nanotube ropes decorated with electrodeposited Pd nanoparticles (NPs). Even higher sensitivity and faster sensor response and recovery are enabled by this sensor architecture. Sensor properties are strongly dependent on the size and size monodispersity of the Pd NPs, with smaller NPs yielding higher sensitivity and more rapid response/recovery. We hope the lessons learned from this science over 15 years will catalyze the development of sensors based upon electrodeposited nanomaterials for gases other than H2.

1. INTRODUCTION The growth in the use of hydrogen gas (H2)-powered fuel cell electric vehicles has motivated the development of H2 sensors designed specifically to detect leaked H2 at concentrations well below the lower explosion limit of 4% (v/v) in air. Targets for H2 sensor performance, published by the U.S. Department of Energy in 2015 (Table 1), specify a response time of 1.0 s © 2017 American Chemical Society

across the concentration range from 0.1% to 10%. Such sensors must also be inexpensive, compact, power-efficient, and able to function across a wide range of ambient temperature and humidity (Table 1). Received: April 3, 2017 Published: August 4, 2017 1902

DOI: 10.1021/acs.accounts.7b00163 Acc. Chem. Res. 2017, 50, 1902−1910

Article

Accounts of Chemical Research

years,6−15 the capabilities of these H2 sensors have been advanced from rudimentary to state-of-the-art. In this Account, I describe the main discoveries responsible for this progress.

Table 1. U.S. Department of Energy Targets for Hydrogen Safety Sensor Performancea sensor metric

value or range of values

concentration range operating temperature response time gas environment lifetime

0.1−10% −30 to 80 °C 1.0 s ambient air; 10−98% humidity 10 years

2. SENSORS FROM NANOWIRE ENSEMBLES PREPARED BY ELECTROCHEMICAL STEP EDGE DECORATION (ESED) In 2000, graduate students Mike Zach and Erich Walter demonstrated that metal nanowires could be prepared using a technique called electrochemical step edge decoration (ESED).16−19 Noble metal (e.g., Pt, Pd, Au) nanowires were obtained by nucleating nanoparticles at linear step edge defects located on the (0001) plane of highly oriented pyrolytic graphite (HOPG).17,18 As these nanoparticles grew larger, they coallesced with nearest neighbors, eventually forming polycrystalline nanowires that were many micrometers or even millimeters in total length (Figure 1a). In a typical experiment, hundreds or thousands of nanowires were electrodeposited at the step-edge defects present on a typical HOPG electrode. But since these nanowires were located on a conductive surface, they could not be incorporated into devices such as sensors that exploited their electrical properties. Mike solved this problem by developing a “lift-off” process for transferring ensembles of nanowires onto glass surfaces and the opportunity to fabricate chemical sensors from electrodeposited nanowires was created (Figure 1b).16 Fred Favier, a visiting scientist from the CNRS Montpellier, France, prepared the first Pd nanowire-based H2 sensors using this process in 2001.6 Fred’s sensors consisted of ensembles of Pd nanowires that were hemicylindrical in cross-section and 200−400 nm in diameter. The fabrication of electrical contacts posed some challenges. Fred first located domains of nanowires that had nucleated in a parallel arrangement by exploring the transferred nanowires using an optical microscope. He then hand-applied dabs of silver paste using a toothpick onto the ends of these arrays to prepare electrical contacts (Figure 1c).

a Updated 2015. U.S. Department of Energy, Energy Efficiency and Renewable Energy (EERE), Fuel Cell Technologies Office. Multi-Year Research, Development, and Demonstration Plan, 2011−2020. Section 3.7 Hydrogen Safety, Codes and Standards, p. 21.

These performance metrics are extremely challenging, and few, if any, H2 sensors have so far achieved this level of performance. Among the available sensor technologies, cost and size considerations favor metal film chemiresistors based upon palladium (Pd) and Pd alloys. Work on such sensors dates to 1972 and perhaps earlier.1 A second simple and costeffective sensor technology, metal-oxide semiconductor (MOS) diodes based upon Pd, was described by Lundstrom and coworkers in 1975. 2 The pure Pd films used in early chemiresistors and MOS sensors were subject to blistering and pealing, a consequence of the strain imposed on the film− dielectric interface by the 10% volume increase that occurs when PdHx undergoes the α-to-β phase transition at ∼1% H2.3 Alloys of Pd with either silver4 or nickel5 suppressed this phase transition enabling the preparation of robust chemiresistors and MOS diodes for H2 sensors that endured exposures to high H2 concentrations. These sensors, described in the early 1990s,5 set new standards for the detection of H2, but their performance did not meet the current DOE targets (Table 1) in terms of either speed or sensitivity. Beginning in 2001, my research group started to explore the possibility of replacing metal films in H2 sensors with nanowires, prepared by electrodeposition. In the ensuing 16

Figure 1. H2 sensors prepared using ESED Pd nanowire ensembles. (a) SEM image (colorized) of Pd nanowires electrodeposited on HOPG. (b) Process flow for the fabrication of a H2 sensor using ensembles of Pd nanowires electrodeposited on HOPG. (c) SEM image (colorized) of a H2 sensor showing the ensemble of transferred Pd nanowires with silver electrical contacts. (d) Current versus time during the exposure of a sensor to pulses of H2 at the indicated concentrations. (e) Current versus H2 concentration calibration curves for H2 sensors operating in both modes. Reproduced from ref 7. Copyright 2002 American Chemical Society. 1903


Article


Figure 2. H2 sensors prepared using single LPNE Pd nanowires.10 (a) Process flow for the preparation of Pd nanowires using LPNE. (b) Schematic digram of a H2 sensor containing a single LPNE Pd nanowire. (c,d) Low magnification SEM images of such a sensor and the Pd nanowire. (e−g) SEM images and AFM cross-sectional scans (bottom) showing three Pd nanowires of varying width prepared by LPNE. Reproduced with permission from ref 10. Copyright 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.

In order to achieve higher sensitivity to H2, we realized that it would be necessary to prepare Pd nanowires that did not fracture upon exposure to H2. In this case, one could rely purely on the electrical resistance of the PdHx to detect H2, just as had been done earlier by Hughes and Schubert using Pd/Ni film resistors.5 But in spite of our best efforts, all Pd nanowires prepared by ESED, with minimum lateral dimensions in the 100 nm range, showed the undesirable fracture behavior. A new method for preparing smaller Pd nanowires was needed, a goal that we pursued for five years.

These Pd nanowire arrays showed a rapid decrease in electrical resistance when exposed to H2 at concentrations above 1% (Figure 1d,e). This was a big surprise since existing film-based H2 sensors showed an increased resistance in H2,5 deriving from the higher resistivity of PdHx, up to 1.8 times higher for x = 0.7, as compared with Pd0.3 What could cause an inverted resistance response? A crucial experiment, elucidating the mechanism by which these Pd nanowire ensembles detected H2, was conducted by Erich Walter, a graduate student in the group.7 Erich used an atomic force microscope (AFM) to observe individual Pd nanowires as they were exposed alternately to air and H2. The first surprise was that exposure of freshly prepared Pd nanowires to H2 caused them to fracture at multiple points along their axes. Each fracture was associated with a gap of 30− 70 nm between fracture faces. When the AFM was filled with H2, Erich’s images of the same nanowires showed that these gaps closed, pressing fracture faces into contact with one another and closing the circuit, increasing the net electrical conduction through the ensemble of nanowires. These “hydrogen-actuated break junctions” were animated by the increase in volume of β-phase PdHx (10%) as compared with pure Pd0.7 This mechanism is unique and interesting, but it did not provide adequate sensor performance. The main issue is that these sensors have a limit-of-detection for H2, LODH2 ≈ 1% because this is the threshold for the α-to-β phase transition for PdHx that produces the volume change.

3. PALLADIUM NANOWIRES PREPARED BY LITHOGRAPHICALLY PATTERNED NANOWIRE ELECTRODEPOSITION (LPNE) Erik Menke working with Chengxiang Xiang and Michael Thompson solved this problem in 2006 by developing a new process for electrodepositing Pd nanowires: lithographically patterned nanowire electrodeposition (LPNE).8,20,21 In LPNE, photolithography is used to prepare “artificial step edges” by photopatterning an evaporated 40−80 nm thick nickel film on glass (Figure 2a). The exposed edges of this film are then used to nucleate and electrodeposit nanowires (Figure 2a(iv)). Because the edges of the nickel film are recessed under a photoresist layer (Figure 2a(iii)), the thickness of the nanowire (perpendicular to the surface) and its width (in the plane of the surface) can be controlled separately.8,20,21 LPNE made it easy to electrodeposit individual Pd nanowires of 1904


Article


Figure 3. Performance of H2 sensors based upon single LPNE Pd nanowires.14 (a) ΔR/R0 versus time data for H2 detection using a Pd nanowire (40 nm (h) × 100 nm (w)) operating at room temperature and at four elevated temperatures as indicated. (b) ΔR/R0 versus [H2] from 4 ppm to 100%. The temperature of “HOT” nanowires varied from 384 K (94 nm (h) × 183 nm (w)) to 428 K (27 nm (h) × 75 nm (w)). (c,d) Response time (c) and recovery time (d) for three nanowires of varying size, as indicated. (e) Response rate (1/τresp) and recovery rate (1/τrec) versus surface area/ volume ratio for 12 single nanowire sensors. (f,g) Response and recovery transients for two nanowire sensors and a Pd film sensor, all with minimum lateral dimensions in the range from 11 to 14 nm. Reproduced from ref 14. Copyright 2015 American Chemical Society.

defined width and height for the first time, and relative to ESED, much smaller lateral dimensions down to 20 nm in height were possible. The discovery of LPNE has proven to be an enabling discovery for all of our subsequent work in H2 sensor development. Fan Yang prepared the first H2 sensors based upon single Pd nanowires prepared using the LPNE process in 2009 (Figure 2a,b,c).9 Fan immediately discovered that these lithographically patterned nanowires fractured in exactly the same way as those prepared earlier using the ESED process. The break junctions formed by fracturing also showed the same H2-actuated opening and closing behavior reported in our earlier work and exactly the same LODH2 ≈ 1% for the detection of H2.9 To our dismay, the LPNE process seemed to confer no advantage in terms of H2 sensing performance. But Fan persevered, fine-tuning the electrodeposition process for Pd, and he soon made a discovery: The addition of EDTA (ethylenediamine tetraacetic acid) to the plating solution complexed the free Pd2+, shifting the onset for electrodeposition of Pd (reaction 1) to more negative potentials where hydrogen evolution (reaction 2) occurs in parallel with Pd2+ reduction. At these potentials in the presence of palladium metal, protons can be directly incorporated into PdHx, which is formed in preference to Pd0 (reaction 3). Pd2 +(aq) + 2e− → Pd(s)

(1)

2H+(aq) + 2e− → H 2(g)

(2)

x H+(aq) + x e− + Pd(s) → PdHx(s)

(3)

obtained by direct Pd0 electrodeposition (reaction 1); 15 nm instead of 5 nm.9,11 But as Fan demonstrated, the key difference was that Pd nanowires synthesized in EDTA did not fracture upon exposure to H2. Moreover, such nanowires could detect H2 across the range from 2 ppm to 10% in nitrogen, a dynamic range for H2 detection of 4 orders of magnitude. By preparing and characterizing many Pd single nanowire sensors, Fan demonstrated that the sensing performance improved as the Pd nanowire got smaller, both in terms of sensitivity to H2 (Figure 3b) and in terms of the response time (Figure 3c) and the recovery time (Figure 3d).11 He also observed that these properties were correlated with the surfacearea/volume ratio of the nanowire (Figure 3e).10 Nanowires with identical heights and different widths showed significant differences in sensitivity and speed favoring narrower nanowires for this reason. For example, the data of Figure 3f,g show an acceleration in both response and recovery for an 11 nm (h) × 93 nm (w) nanowire, relative to a 14 nm (h) × 121 nm (w) nanowire and an 11 nm Pd film.10 Since the minimum distance over which H diffusion will occur (11 nm) is the same for the 11 nm Pd film and the 11 × 93 nm Pd nanowire, these data support the conclusion that the rates of sensor response and recovery are not determined by H diffusion. Instead, we believe that the rate of sensor response is limited by two other processes, depending upon the H2 concentration regime. For [H2] < 1%, the detection of H2 does not trigger the α-to-β phase transition. In this limit, we believe that the ratelimiting step in sensor response is the dissociative adsorption of H2 onto the Pd surface.10,11 This conclusion follows from the observed correlation between response and recovery rate with the surface area/volume ratio of Pd nanowires and from measurements the apparent activation energy for sensor response, Ea,resp. Within the narrow concentration range from 1% to 2%, sensor response times slow dramatically (see Figure

As soon as nanowire growth ends and the potentiostat is disconnected, the PdHx spontaneously decomposes into dihydrogen and Pd0, forming nanowires that have a subtly different morphology and a larger mean grain diameter than is 1905


Article


Figure 4. Performance of H2 sensors based upon single LPNE Pt nanowires.13 (a) Comparison of H2 detection at Pt and Pd nanowires of the same size (20 nm (h) × 130 nm (w)) both operating at T = 550 K. The hydrogen concentration program is shown in green at bottom. (b) ΔR/R0 versus [H2] calibration plot for a Pt and a Pd nanowire showing a 1000-fold lower LODH2 for Pt versus Pd. (c) Response time versus [H2] for Pt and Pd nanowires showing more rapid response of Pt for [H2] < 3%. (d) Recovery time versus [H2] for Pt and Pd nanowires showing slow, 1000−2000 s [H2]-invariant recovery of Pt. (e) Proposed mechanism for the operation of Pt nanowire H2 sensors, involving the modulation of nanowire resistance by electron surface-scattering. Reproduced from ref 13. Copyright 2012 American Chemical Society.

3c) as the α-to-β phase transition becomes rate-limiting. For [H2] > 2%, both H2 adsorption and the kinetics of the phase transition operate to limit the response speed.10 It should be emphasized that H2 safety sensors must function in air, a much more challenging ambient than N2 or Ar. Both sensitivity and speed for Pd nanowire sensors are degraded in air relative to these inert gases. Pd nanowires operating at 300 °C in N2, for example, are capable of a LODH2 of 2 ppm, while in dry air,9,11 LODH2 is increased 50-fold to 100 ppm.9,11 Similarly, response and recovery times of 40 s and 300 s for exposure to 0.10% H2 in N2 are stretched to 400 s and 1000 s in air. The oxygen in air increases the LODH2 by opening a channel for the removal of chemisorbed H from the nanowire surface as water,22 preventing its absorption to form PdHx.23,24 PdO +

3 H 2(ads) → PdH + H 2O(ads) 2

H 2O(ads) → H 2O(g)

4. PLATINUM NANOWIRES PREPARED BY LPNE Unlike Pd, platinum (Pt) does not absorb H2, but it is able to chemisorb it, forming a surface hydride. Fan Yang and I were interested in understanding whether this process altered the resistivity of the Pt nanowire, and if so, we were anxious to compare H2 sensors prepared using Pt nanowires with Pd nanowire sensors. The results certainly surprised us. The response of Pd and Pt nanowire sensors to the same sequence of H2 pulses in air (Figure 4a) shows that Pt nanowires exhibit an inverted response compared to Pd nanowires; the resistance decreases in the presence of H2 instead of increasing.13 Moreover, the amplitude of the resistance decrease for Pt nanowires was the same (ΔR/R0 ≈ −3%) across a range of [H2] from 4% to 50%. Pd nanowires showed ΔR/R0 ranging from 0 to 7%, depending on [H2]. Pt nanowires were also more sensitive than Pd nanowires, able to detect H2 in air at 10 ppm (Figure 4b), and faster, with a response time at [H2] = 1% 100× faster than a Pd nanowire of the same size (Figure 4c). But for Pt nanowires, the recovery time was 1000−2000 s independent of the [H2], much slower than for Pd nanowires. A mechanism that explains these observations is the following (Figure 4e).13 In humid air, Pd and Pt surfaces are both covered with chemisorbed oxygen in the form of MO and M−OH functionalities. Physisorbed H2O and O2 are also present on these surfaces. Exposure of Pd and Pt nanowires to H2 results in dissociation of the H2 and the formation of Pt−H and Pd−H monohydrides on the surfaces of these metals. On Pd, adsorbed

(4) (5)

Chemisorbed oxygen, Pd−O, also blocks Pd adsorption sites, impeding hydrogen adsorption and extending the time required for equilibration of hydrogen in the gas phase with the PdHx, retarding both response and recovery. The detrimental influence of air on sensor performance provides a startling confirmation that sensor metrics are strongly influenced by surface chemistry for resistance-based hydrogen gas sensors. 1906


Article


Figure 5. Performance of Pd@Pt nanowire H2 sensors.14 (a) Schematic diagram of Pd@Pt nanowire sensors illustrating the four Pt coverages explored in this study. (b) Resistance versus time data for three Pd@Pt nanowire sensors (40 nm( h) × 100 nm (w) × 50 μm (l)) with θPt = 0.0, 1.0, and 10 ML. (c) Normalized ΔR versus time plots showing the acceleration of sensor response for Pd@Pt nanowire sensors with θPt = 0.0, 1.0, and 10 ML. (d) Summary of response (left) and recovery (right) data for Pd@Pt nanowire sensors with θPt = 0.0, 0.1, and 1.0 ML. Recovery is more strongly accelerated relative to response, but both are faster for θPt = 1.0 ML. Reproduced from ref 14. Copyright 2015 American Chemical Society.

hydrogen diffuses into the bulk of the nanowire transforming it into more resistive PdHx. H2 dissociation continues until the composition of the resulting PdHx is equilibrated with the ambient partial pressure of H2.13 On Pt, however, once the surface is saturated with Pt−H, no further adsorption of H2 can occur. Although the bulk of the Pt nanowire is unchanged, a resistance decrease is nevertheless observed because the Pt−H surface is less efficient at scattering conduction electrons inside the nanowire than the MO and M−OH functionalities present on the Pt surface in air. This means that in the presence of H2, the resistance of a Pt nanowire is lower than that in air. Of course, for macroscopic wires this surface scattering effect is negligible but for metal nanowires, surface electron scattering accounts for a significant fraction, up to 5%, of the total resistance of these structures. In fact, the 3% value of ΔR/R0 that we measure experimentally is consistent with a change in the specularity parameter, p, governing this process from p = 0.44 (in air) to 0.80 (at a Pt−H saturated surface).13 A key point is that in the presence of H2, the coverage of hydride on the Pt surface increases until saturation coverage is attained. This is the reason that the resistance change observed for Pt nanowires shows no H2 concentration dependence. These physics were first described by the German physicist Klaus Fuchs in 1938, but they have not been exploited in sensors for the detection of H2 or other gases to our knowledge, until the results of these experiments were published in 2012.25

H2 sensors based upon Pt nanowires are unlikely to be very useful because they cannot discriminate different H2 concentrations. However, used in conjunction with less sensitive Pd nanowires, Pt nanowires could enable a more rapid “alarm” at lower H2 concentrations, providing the means for detecting H2 leaks much more rapidly.13

5. Pd@Pt CORE@SHELL NANOWIRES The surprising H2 sensing performance of Pt nanowires13 inspired us to prepare Pd nanowires with an ultrathin (0.1 ML < θPt < 10 ML) conformal Pt coating (Figure 5a). The goal was to prepare nanowires that retained the ability of Pd nanowires to discriminate different H2 concentrations, while accelerating sensor response relative to pure Pd nanowires.10,11 Xiaowei Li, a graduate student, lead this project, and his first task was to develop a process for conformally electrodepositing Pt onto Pd nanowires.14 In general, when a metal, M1, is deposited onto the surface of a second metal, M2, an additional increment of potential is required beyond the reversible potential, E°M′1, in order to nucleate M1 and initiate this deposition process. The extra potential is termed the nucleation overpotential, ηM1. In the case of interest, involving the electrodeposition of Pt onto a Pd nanowire, the interval between EPt °′ and (EPt °′ − ηPt) is ∼400 mV.13 At potentials within this interval, no Faradaic current is observed. Xiaowei found that the electrodeposition of Pt at potentials more negative than EPt °′ − ηPt resulted in the nucleation of three-dimensional Pt nanoparticles and the rapid 1907


Article


Figure 6. CNT@Pd NP H2 sensors.15 (a) Low magnification SEM image of a single CNT rope H2 sensor. (b,c) TEM images of a CNT rope after the electrodeposition of Pd NPs at low (b) and higher magnification (c) showing Pd NPs (inset, blue circles) and a single Pd NP (c). (d) Histograms of Pd NP diameters for electrodepositions of various quantities of Pd in terms of the Coulombic loading, ranging from 15 to 102 μC on a 50 μm length CNT rope, like that depicted in panel a. (e,f) Resistance versus time data for four CNT@PdNP sensors with QPd ranging from 15 to 102 μC. Shown are the lowest H2 concentration range (10−100 ppm, e) and the medium range (100−1000 ppm, f). (g) Plot of ΔR/R0 versus [H2] for four CNT@PdNP H2 sensors operating at room temperature (RT) and a Pd nanowire (40 nm × 100 nm × 50 μm) operating at RT, 316, and 376 K as indicated. Reproduced from ref 15. Copyright 2017 American Chemical Society.

growth of these particles at diffusion-control.14 A rough overlayer of Pt is thereby obtained that becomes many nanometers in thickness in a few seconds. This diffusionlimited growth mode for Pt and the thick Pt films it produces are of no use to us. Xiaowei further discovered that for potentiostatic deposition °′, a very small, at more positive potentials, still negative of EPt time-independent Faradaic current was observed.14 The amplitude of this current was reproducible for nanowires of the same size and length, suggesting that a surface area-limited deposition process could be responsible for it. This hypothesis was supported by Xiaowei’s observation that Pd films subjected to this process were coated by a spatially uniform and conformal Pt layer that was just a few Pt monolayers in thickness, based upon SEM, EDX, and XPS measurements.14 Collectively, these data suggested that at low deposition overpotentials, the growth of Pt on Pd surfaces occurred by a layer-by-layer or Frank van der Merwe mechanism.26 We used this process for preparing Pd@Pt core@shell nanowires where the Pt coverage (θPt) was in the range from θPt ≈ 0.1−10 ML. How did these ultrathin Pt overlayers influence H2 sensing? Raw data for three Pd@Pt nanowire H2 sensors, with θPt = 0.0, 1.0, and 10 ML, at five temperatures (Figure 5b) show that elevated temperature exerts a strong influence on sensing, producing more rapid response and recovery.14 But, as we reported earlier,10 elevation of the temperature also decreases the amplitude of ΔR/R014 as the solubility of H in PdHx in the heated nanowire is depressed.3 At 376 K, one also observes a low drift of the baseline to lower R caused by grain growth within the nanowire. An operating temperate of 316 K is close

to optimal, producing accelerated speed with a minimal loss of sensitivity and a stable baseline resistance.14 At T = 316 K, a Pd@Pt nanowire with θPt = 1.0 ML shows an additional boost in speed relative to pure Pd nanowires (Figure 5c). But response and recovery do not benefit equally from the Pt overlayer; recovery is preferentially accelerated (Figure 5c).14 The strong acceleration of recovery induced by the surface Pt monolayer is puzzling. On the one hand, pure Pt nanowires show unusually slow recovery with τrec = 1000−2000 s across all concentrations (Figure 4d). On the other hand, a single monolayer of Pt on a Pd nanowire accelerates recovery by a factor of 10 or more (Figure 5c,d).14 The difference between a pure Pt nanowire and a Pd@Pt nanowire is that the former derives signal from surface scattering and responds only to the composition of adsorbates on the Pt surface. The Pd@Pt nanowire, on the other hand, acts like a pure Pd nanowire in the sense that its resistance is dominated by the bulk PdHx formed upon H2 exposure. The conclusion is that Pt accelerates the formation of H2O on the sensor surface, and the removal of H as water, accelerating the conversion of bulk PdHx to Pd0 at Pd@Pt nanowires and accelerating recovery. At a pure Pt nanowire, however, it is the desorption of the surface H and the restoration of the products of O2 adsorption (e.g., Pt = O and Pt-OH) at the nanowire surface that must occur before the baseline resistance of the nanowire is restored. The presence of a surface Pt monolayer modifies the H2 sensing behavior of a pure Pd nanowire in two other ways:14 First, the amplitude of ΔR/R 0 measured at each H 2 concentration is reduced at low temperatures (T = 294 and 303 K) and enhanced at higher temperatures (T = 316, 344, and 376 K) for θPt = 1.0 ML. Second, at θPt = 10 ML, the 1908


Article


Most importantly, CNT@Pd NP H2 sensors are faster than Pd nanowires and faster than prior H2 sensors based upon CNTs. Rise and recovery times were as fast as 60 s (for [H2] = 1000 ppm), whereas Pd nanowires are a factor of 3−5 slower than this at room temperature. Xiaowei examined a relatively narrow range of mean Pd NP diameters, from 4.5 to 5.8 nm, but the speed of CNT@Pd NP H2 was correlated with Pd NP diameter across all four mean particle diameters. As in our earlier work involving Pd nanowires,10,11 we conclude that surface area/volume ratio of the sensing element(s), the Pd NPs in this case, is a determining factor in defining the rate of sensor response and recovery. We were very surprised to discover that the performance of CNT@Pd NP H2 sensors eclipses that of Pd single nanowire sensors and Pd@Pt nanowire sensors at room temperature.

reduced permeability of the Pt shell to H interferes with H2 detection leading to a LODH2 of ∼2% versus ∼0.2% for θPt = 1 ML. Pd@Pt nanowires with θPt = 10 ML are much too insensitive to H2 to function as sensors. This work teaches us an important lesson: The response/ recovery speed of nanoscale chemical sensors is prone to retardation by rate-limiting surface chemical kinetics, because as the critical dimension of the sensor is reduced the diffusional flux of molecules to sensor surfaces are increased. Diffusion limits are replaced by kinetic limits for the rate of surface chemical reactions that are required for sensor function. For this reason, we expect that the performance of nanoscale sensors should be strongly influenced by catalysts that accelerate surface chemical reactions, such as adsorption/ desorption. The influence of a Pt monolayer provides a graphic demonstration of this principle: The H2 response and recovery kinetics for a highly optimized H2 sensor consisting of a Jouleheated Pd nanowire can be significantly accelerated by the addition of minute quantities (e.g., 1 ML) of a Pt metal catalyst to the nanowire surface.

7. SUMMARY The field of electrical H2 gas sensing has advanced enormously over the past 25 years. With recent work in the past 15 years, a high level of performance has been achieved as a consequence of a progression of innovationssome of these described in this Account. It is tempting to predict that H2 sensing technology is approaching maturity and that future improvements will be incremental, but as we recently discovered in the case of CNT@Pd NP H2 sensors, disruptive discoveries likely remain to be made. A 1.0 s response time at any H2 concentration (Table 1) remains a daunting challenge that, to our knowledge, is not achieved by any electrical H2 sensor described so far.

6. PALLADIUM NANOPARTICLE-DECORATED CARBON NANOTUBE ROPES In 1995, Hongjie Dai and co-workers27 demonstrated that films of carbon nanotubes (CNTs) decorated with Pd nanoparticles could function as H2 sensors. Relative to the metrics of Table 1, these sensors were slow, with response times of several minutes or more across all concentrations, but they were the very first gas sensors to exploit CNTs. In many subsequent papers on this subject, the performance of these systems have certainly improved, but response and recovery speeds have remained too slow for reasons that have not been evident.15 Based upon our prior work, which implicated dissociative H2 chemisorption onto the Pd surface as the rate-limiting step in sensor response,10,11,14 we formulated a hypothesis that the diameter of the Pd nanoparticles was critically important in determining the response and recovery properties of this class of H2 sensors. Earlier this year, Xiaowei Li in my group devised experiments designed to test this hypothesis. Xiaowei first patterned thick CNT “ropes”, essentially films with a width of a few micrometers, onto glass surfaces using dielectrophoresis in conjunction with a microfabricated LPNE nickel electrode (Figure 6a).15 Then, he removed the template, attached electrical contacts to the CNT rope, and electrodeposited Pd nanoparticles using a pulsed electrodeposition procedure. The key point is that Xiaowei was able to produce Pd NPs dispersed within the CNT rope (Figure 6b) that were narrowly dispersed in diameter, with a diminutive mean diameter of

A Nose for Hydrogen Gas: Fast, Sensitive H2 Sensors Using

Recommend Documents