Hydrogen Sensors and Switches from Electrodeposited Palladium Mesowire Arrays

See allHide authors and affiliations

Science  21 Sep 2001:
Vol. 293, Issue 5538, pp. 2227-2231
DOI: 10.1126/science.1063189


Hydrogen sensors and hydrogen-activated switches were fabricated from arrays of mesoscopic palladium wires. These palladium “mesowire” arrays were prepared by electrodeposition onto graphite surfaces and were transferred onto a cyanoacrylate film. Exposure to hydrogen gas caused a rapid (less than 75 milliseconds) reversible decrease in the resistance of the array that correlated with the hydrogen concentration over a range from 2 to 10%. The sensor response appears to involve the closing of nanoscopic gaps or “break junctions” in wires caused by the dilation of palladium grains undergoing hydrogen absorption. Wire arrays in which all wires possessed nanoscopic gaps reverted to open circuits in the absence of hydrogen gas.

Chemical sensors based on nanowires usually operate through a change in resistance induced by the surface adsorption of analyte molecules. Tao and co-workers (1) have demonstrated that the conductivity of gold nanowires changes upon exposure to molecules capable of chemisorbing to gold surfaces such as thiols and amines. An analogous effect has been observed for single-walled carbon nanotubes that exhibit a resistance that changes upon exposure to gaseous oxygen (2), water (3), and amines (4).

We describe a wire-based sensor for the detection of hydrogen gas (H2) that is based upon resistivity changes caused not by surface adsorption, but on changes in the structure of the wire itself. These sensors consist of up to 100 Pd mesoscopic (5) wires (henceforth, “mesowires”) arrayed in parallel (Fig. 1, A and B). Like conventional H2sensors based on macroscopic Pd resistors, the Pd mesowire array (PMA) in this sensor exhibits a resistance change upon exposure to H2. In contrast to all existing resistance-based H2 sensors, however, the resistance of PMAs decreases instead of increasing in the presence of H2(6). We propose a new mechanism to account for this “inverse” response.

Figure 1

(A) Schematic diagram of a PMA-based hydrogen sensor or switch. (B) SEM image [400 μm (h) by 600 μm (w)] of the active area of a PMA-based hydrogen sensor. (C) PMAs were prepared by electrochemical step edge decoration at graphite surfaces and transferred to a cyanoacrylate film.

Palladium mesowires were electrodeposited from aqueous solutions of Pd2+ onto step edges present on a graphite surface, using a two-step procedure (Fig. 1C) (7). Starting with a freshly cleaved graphite surface, a 5-ms nucleation pulse of –0.2 V [versus saturated calomel electrode (SCE)] was first applied. As shown in Fig. 2, this potential is well negative of the reversible potential for Pd deposition in these solutions (+0.6 to +0.7 V versus SCE). After this nucleation pulse, the growth of Pd mesowires was carried out using potentials in the ranges shown in gray in Fig. 2. These deposition potentials produced deposition current densities ranging from 30 to 60 μA cm−2, and deposition times for 200-nm-diameter mesowires were 10 min (8). Attempts to accelerate wire growth by using more negative potentials resulted in discontinuous structures.

Figure 2

Cyclic voltammograms for a graphite electrode in two aqueous Pd plating solutions, as indicated. The potential range used for the growth of Pd mesowires is indicated in gray.

The morphology of the Pd wires obtained by electrodeposition was dependent on the identity of the electrolyte present in the plating solution. Mesowires deposited from HCl solutions (Fig. 3, A and B) were rough and granular. The dimensions of the grains in these polycrystalline wires, as estimated from scanning electron microscope (SEM) images, ranged from 50 to 300 nm. The smallest continuous wires obtained from this solution were 150 nm in diameter. Deposition from HClO4 solutions (Fig. 3, C and D) yielded wires with a smoother morphology. The grains in these wires were 10 to 50 nm in diameter. This smoother morphology permitted nanowires as narrow as 55 nm to be deposited. Both types of wires produced sharp electron diffraction patterns characteristic of face-centered cubic (fcc) Pd metal (8). Rough and smooth wires prepared using these two plating solutions behaved identically in the H2 sensors and switches described below.

Figure 3

(A and B) SEMs of Pd mesowires prepared by electrodeposition from aqueous 2.0 mM PdCl2 and 0.1 M HCl. E dep = 0.3 V,t dep = 900 s. (C andD) SEMs of mesowires electrodeposited from aqueous 2.0 mM Pd(NO3)2 and 0.1 M HClO4.E dep = 0.3 V,t dep = 150 s.

Freshly deposited Pd mesowires were transferred from the graphite electrode surface onto a glass slide coated with cyanoacrylate (Fig. 1C). When the cyanoacrylate film had hardened (8 hours), arrays of Pd mesowires were contacted with silver epoxy. A SEM image of the active area of a sensor (Fig. 1B) shows a PMA as horizontal lines contacted on each side by silver paint. In this device, the span between silver contacts was 300 to 500 μm. In other devices, this gap ranged from 100 μm to 1.0 mm. Only mesowires long enough to span these distances were involved in sensor function. The number of such mesowires, estimated from SEM images, varied from 20 to 100. The overall success rate for preparing H2 sensors using this procedure was better than 50%.

PMAs were operated as H2 sensors by applying a constant voltage of 5 mV between the silver contacts and measuring the current, which was between 1 and 20 μA. Two different modes of sensor operation were observed. “Mode I” sensors remained conductive in the absence of H2. The resistance of a mode I sensor decreased in the presence of H2 (Fig. 4A), with the decrease related to the H2 concentration. In nitrogen carrier gas at atmospheric pressure and room temperature, the limit of detection for a mode I sensor was 0.5% H2. The sensor exhibited a sigmoidal response curve (Fig. 4C) with a minimum resistance at 5 to 10% H2.

Figure 4

(A) Current response of a mode I sensor to hydrogen/nitrogen mixtures (concentration of H2in percentage as shown). Data were acquired in random order of H2 concentration. (B) Current response of a mode II sensor to hydrogen/nitrogen mixtures (concentration of H2 in percentage as shown). Data were acquired in random order of H2 concentration. (C) Current amplitude versus H2 concentration for the two sensors of C (mode I) and D (mode II). (D) Sensor resistance versus time response for a mode I sensor. This is the only experiment of those shown in this figure for which the gas transport system was optimized for fast time response. However, even in this case, it is likely that the switching time for valves was close to the apparent rise time of the sensor resistance (70 to 75 ms).

Mode II sensors were hydrogen-activated switches. In the absence of H2, the resistance of a mode II sensor became large (>10 megaohms; switch open). In this “wait state,” the sensor dissipated no power and produced no noise. Typical data for a mode II sensor are shown in Fig. 4B. Above a threshold of approximately 2% H2, the switch closed and a device resistivity became measurable. Above this threshold H2 concentration, the same sigmoidally shaped response curve seen for mode I sensors (Fig. 4C) was obtained.

Sensors were insensitive to a variety of gases other than H2, including Ar, He, N2, water vapor, and O2. The amplitude of the sensor response was unaffected by the presence of CO and CH4 at concentrations up to 3%, but the response time to H2 in the presence of CO was increased. The response of sensors to D2 was identical to that observed for H2.

A rise-time (baseline to 90% signal saturation) of less than 80 ms has been observed for the response of mesowire-based sensors to 5% H2 (Fig. 4D). This value is approximately the response time of the gas flow system used for these measurements and represents an upper limit to the true response time of these sensors. If the true response time is dictated by the rate at which H2 can diffusionally saturate Pd grains in the mesowire, a faster response is expected. For 200-nm-diameter grains, for example, H2 must diffuse r = 100 nm. The time, τ, required for this diffusional transport can be estimated from the diffusion coefficient for hydrogen in Pd, D, using τ = r 2/2D. Assuming a mean value for D of 10−7 cm2s−1 (9), τ = 0.5 ms.

Macroscopic Pd resistors are commonly used as H2 sensors, but in these devices, exposure to H2 causes an increase in the resistance by a factor of up to 1.8 at 25°C (6). This resistance increase is caused by the increased resistivity of Pd hydride relative to pure Pd. A different mechanism must operate in the Pd mesowire-based devices described here.

The mechanism we propose is shown in Fig. 5. All of the PMAs we have investigated have been conductive before an initial exposure to H2(i.e., all devices were initially mode I). The first exposure to H2 irreversibly modified the sensor: Either an increase in the baseline resistance (in air) of a sensor was observed for mode I devices or the resistance became large (the mode I device was converted into a mode II device). A resistance versus time transient for this conversion is shown in Fig. 5. After the first exposure to H2, exposure to air opens nanoscopic gaps in some (mode I) or all (mode II) mesowires in the sensor. These gaps open when the hydrogen-swollen Pd grains in each mesowire return to their equilibrium dimensions in the absence of hydrogen. Subsequently, it is the closing of these gaps or “break junctions” in the presence of H2 that account for the decreased resistance of the sensor. Many or all of the mesowires in the array exhibit this switching behavior in mode I and mode II devices, respectively. This break junction mechanism provides an immediate explanation for the device thresholds seen at 1 to 2% H2: At room temperature, the transition from α-phase to β-phase occurs atP H2 = 8 torr in bulk Pd (6) which, at atmospheric pressure, corresponds to ≈1% H2.

Figure 5

The first exposure of a new sensor to hydrogen. In this case, an irreversible transition from mode I to mode II operation was observed (bottom). Shown at top is the mechanism proposed for mode II sensor operation. Mode I sensors operate by an identical mechanism, except that some mesowires remain conductive in the absence of hydrogen.

Acting in opposition to this “break junction” effect is an increase in resistance of each “closed” mesowire caused by the increased resistivity of Pd hydride. Mesowire-based sensors do not manifest this resistance effect in any observable way (e.g., sloping device current plateaus or baselines). This difference suggests that the change in mesowire resistance occurs on a time scale equal to, or faster than, the time required for the closing and opening of break junctions in these wires. As indicated earlier, this expectation is reasonable based on the calculated rate of diffusion.

Direct evidence for the break junction mechanism comes from atomic force microscopy (AFM) observations of individual Pd mesowires. Four AFM images of a Pd mesowire on a graphite surface (Fig. 6) track the structure of this mesowire in air (Fig. 6, A and C) and in pure H2 (Fig. 6, B and D). The gap present in the Pd mesowire shown in Fig. 6A alternatively closes and reopens as the ambient gas is changed from air to hydrogen and back to air (10). After several exposures to H2, SEM images of mesowires (Fig. 1B) reveal the presence of 2 to 10 of these gaps for every 10 μm of wire length. A more detailed examination of the AFM data reveals that multiple grains slide in a concerted fashion to effect the opening and closing of these nanoscopic gaps. For the mesowire shown in Fig. 6, for example, the three grains to the left of the break junction move to the right by ≈50 nm to close the gap. The two grains to the right of the gap remain stationary (“pinned”). A second ensemble of at least four grains also moves to the right by 30 to 50 nm.

Figure 6

Atomic force microscope images of a Pd mesowire on a graphite surface. Images (A) and (C) were acquired in air; images (B) and (D) were acquired in a stream of hydrogen gas. A hydrogen-actuated break junction is highlighted. (E) Plot of the mesowire topography along its axis showing the direction of grain sliding (red arrows) associated with opening and closing of the break junction.

The motion of mesowire segments seen in Fig. 6 is the direct result of the swelling of individual Pd grains in the presence of H2. X-ray crystallographic data for the Pd hydride system (5) indicates that the lattice constant of Pd is 3.889 Å, whereas that for β-phase PdH0.7, which is the thermodynamically stable form of Pd in equilibrium with 1.0 atm H2 at 25°C, is 4.025Å—an increase of 3.5%. For “pinned” grains which have an orientation on the surface that remains unchanged during switching, AFM images revealed a height increase of 2 to 3% in the presence of H2 (Fig. 6E). Grains involved in sliding showed apparent height increases that ranged from 4 to 15%. These results may indicate a propensity of some Pd grains to be squeezed slightly up and out of the plane of the surface in the presence of H2.

The reversible and reproducible break junction dynamics we describe for electrodeposited Pd mesowires provides a new mechanism by which mesowires can operate as chemical sensors. The performance of the mesowire array–based sensors documented here challenges existing H2 sensing technologies. In particular, mesowire array–based H2 sensors possess four attributes: (i) fast response; (ii) room-temperature operation; (iii) diminutive power requirements of less than 100 nW; and (iv) resistance to poisoning by reactive gases, including O2, CO, and CH4.

  • * To whom correspondence should be addressed. E-mail: rmpenner{at}


Stay Connected to Science

Navigate This Article