ARTICLES PUBLISHED ONLINE: 29 NOVEMBER 2009 | DOI: 10.1038/NNANO.2009.349
Dispersed nanoelectrode devices Antonio Tricoli and Sotiris E. Pratsinis* The enhanced performance and reduced scale that nanoparticles can bring to a device are frequently compromised by the poor electrical conductivity of nanoparticle structures or assemblies. Here, we demonstrate a unique nanoscale electrode assembly in which conduction is carried out by one set of nanoparticles, and other device functions by another set. Using a scalable process, nanoparticles with tailored conductivity are stochastically deposited above or below a functional nanoparticle film, and serve as extensions of the bulk electrodes, greatly reducing the total film resistance. We apply this approach to solid-state gas sensors and achieve controlled device resistance with an exceptionally high sensitivity to ethanol of 20 ppb. This approach can be extended to other classes of devices such as actuators, batteries, and fuel and solar cells.
R
educing the size of material components to the nanoscale has led to new functionalities in electronic devices such as sensors1, solar cells2,3, batteries4 and fuel cells5. The integration of nanostructures into devices is not trivial, however, and requires creative new methods for nanomaterial deposition6, stabilization7 and processing8. Furthermore, not all nanoscale properties improve with decreasing size. In particular, electrical conductivity is reduced drastically through increases both in surface energy barriers and the density of boundaries9. This inverse trend between size and conductivity limits the further downscaling of electronic components in all applications where electrical conductivity determines device performance and industrial feasibility. In fact, the performance of dye-sensitized solar cells can be greatly increased with nanostructured TiO2 films, but electrical losses limit cell efficiency2. Similarly, the discharge capacity of lithium–manganese batteries can be increased by using LiMn2O4 nanoparticles in a carbon network10. The high mass of carbon required, however, by the low conductivity of such nanoparticles drastically reduces the specific capacity of the composite. The sensitivity of chemo-resistive gas sensors is drastically increased1–11 by reducing the grain size of the sensing elements below 10 nm; however, this is at the expense of increased resistivity, which again limits their commercial application7. This poor conductivity of nanostructured materials challenges the fabrication of microdevices that integrate functional nanostructures with standard electronic components. In fact, state-of-the-art electrodes in microdevices consist of dense films that are deposited by sputtering, screen printing, chemical vapour decomposition/ physical vapour deposition (CVD/PVD) and other methods compatible with complementary metal oxide semiconductor (CMOS) technology12. Typically, a functional nanostructured film is deposited between these electrodes13. Film properties and electrode geometry determine the film resistance. Because the properties of a film depend on its composition, the minimal film resistance is determined primarily by the electrode distance and area ratio, which are quite restrictive for wide-bandgap semiconductors227. Recently, unique electronic components with high current-carrying capacities have been obtained by using dense, thin films of aligned carbon nanotubes between electrodes8. Here, we explore a new concept to control the resistance of a nanostructured film that can overcome some of the limitations of conventional electrodes, in which nanostructured films are deposited that contain functional (for example, semiconducting) and
conductive domains. Optimal functionality can therefore be achieved by decreasing the semiconducting grain size, while the presence of interspersed conductive domains (the so-called nanoelectrodes) among them reduces the high film resistance and associated electrical losses. Metallic or low-bandgap metal-oxide nanoparticles serve as the nanoelectrodes. As a proof of concept, high-performance solid-state gas sensors comprising widebandgap nanoparticle films are assembled with or without nanoelectrodes using dry (flame) aerosol technology (wet chemistry routes can also be used). The effect of nanoelectrode thickness and layout on sensor resistance and performance is investigated and compared with those made with standard electrodes.
Nanoelectrode assembly strategies Figure 1 shows a schematic of standard electrodes without (Fig. 1a) and with nanoelectrode (Fig. 1b,c) layouts for integration of functional nanostructured films in electronic devices. Typically (standard electrodes), functional films of (semiconducting) nanoparticles were deposited between substrate electrodes (Fig. 1a). These are usually either interdigitated noble metal macroelectrodes sputtered on inert (for example, Al2O3) substrates11 or microelectrodes micromachined onto silicon wafers13. Synthesis and deposition of the functional (semiconducting) nanoparticle films was achieved by a wet (for example, sol–gel) or dry (for example, aerosol) process. The resulting film was calcined or annealed in situ. In the present device, the functional nanoparticles were made using a scalable flame aerosol process14 and consisted of 10-nm SnO2 crystals (Fig. 1a) having conductivity controlled by surface states and therefore sensitive to analytes (such as CO or ethanol vapour)11. The resulting sensor resistance (Fig. 1d) of a corresponding ideal circuit depends on film porosity and thickness, SnO2 resistivity and macroelectrode geometry9. Although in the present setup gas sensors have been investigated, the same method could be used for deposition of other semiconductors such as WO3 , ZnO and TiO2 which are attractive for use in sensors, solar cells and other devices. Nanoelectrodes were also deposited using the above aerosol process either before (Fig. 1b, ‘bottom’ layout) or after (Fig. 1c, ‘top’ layout) deposition of the functional SnO2 film. The total film resistance could be controlled with either layout by varying the relative size of the functional (semiconductive) and conductive domains, as shown in the equivalent ideal circuits (Fig. 1e,f ). Film
Particle Technology Laboratory, Institute of Process Engineering, Department of Mechanical and Process Engineering, ETH Zu¨rich, CH-8092 Zu¨rich, Switzerland. *e-mail:
[email protected] 54
NATURE NANOTECHNOLOGY | VOL 5 | JANUARY 2010 | www.nature.com/naturenanotechnology
NATURE NANOTECHNOLOGY a
ARTICLES
DOI: 10.1038/NNANO.2009.349
Standard
c
With nanoelectrodes
b Bottom
+
d
+
−
Functional particles
Top
−
Conductive particles
e Rcond
Rfunct
+
−
Rfunct
Rcond
Rfunct
f
Rfunct
Rcond
+
+
−
Rfunct
Rcond
+
−
−
Figure 1 | Schematic of possible layouts for functional nanostructured films integrated in microcircuitry (such as solid-state gas sensors). a, Standard chemoresistive gas sensor consisting of a functional film of semiconducting nanoparticles (for example, SnO2 , WO3 or TiO2) deposited between a pair of electrodes. In this configuration, the functional material properties influence both the performance and resistivity of the nanostructured film. b,c, In contrast, for dispersed, asymmetric nanoelectrodes, the film resistivity can be tailored by choosing conductive (silver, platinum, gold, CuO) particles with appropriate characteristics. d–f, This is shown by equivalent circuits for the different layouts: the standard electrode circuit (d) and those containing a bottom (e) or top (f) layout. In e and f, the total resistance (R) is a function of the material properties and the relative thickness between the conductive and functional domains. In the bottom layout, a continuous nanoelectrode layer (e, broken line) would shunt the functional film and destroy the film performance.
c
a
105
60
Sensor resistance, Rair (MΩ)
104 +
−
40
Substrate
103 102
20 1
10
2 μm 100 0
b
200
400
600
Sensor response, S = Rair /REtOH − 1
320 °C, dry air, 10 ppm EtOH
BT = 100 nm
0 800
SnO2 bulk thickness, BT (nm)
BT = 600 nm
d 240
Time (s)
180
120
60 Recovery time Response time 0 0
200
400
600
800
SnO2 bulk thickness, BT (nm)
Figure 2 | Performance of solid-state gas sensors with standard electrodes. a,b, SEM images of SnO2 nanostructured films deposited by FSP with specific bulk thicknesses (BT) of 100 nm (a) and 600 nm (b). c, Increasing the amount (bulk thickness) of SnO2 decreases the sensor resistance (triangles), which asymptotically reaches a minimum value determined by the substrate geometry, particle properties and film porosity but decreases also the sensor response (circles) to analytes (for example, EtOH). Triangles correspond to the sensor response (right vertical axis) and circles correspond to resistance (left vertical axis). d, The response (triangles) and recovery (circles) times to 10 ppm EtOH are reduced remarkably by decreasing the SnO2 bulk thickness.
performance is influenced enormously by the choice of layout, as will be shown. Narrow-bandgap metal-oxide (for example, CuO) or metallic (for example, silver, gold, platinum) particles can be used as the nanoelectrodes (Fig. 1b,c). Often, incorporation of CuO in sensor films such as SnO2 is used to increase the sensitivity
and selectivity to some analytes (for example, H2S) through the formation of p–n junctions15. This, however, increases the film resistance instead of decreasing it through electron/hole recombination in the junction15. Here, the formation of such p–n junctions is minimized by depositing, separately, the ‘conductive’ (p-type) and
NATURE NANOTECHNOLOGY | VOL 5 | JANUARY 2010 | www.nature.com/naturenanotechnology
55
ARTICLES
NATURE NANOTECHNOLOGY
a
b
DOI: 10.1038/NNANO.2009.349
EDXS of top CuO CuLa
+
−
Substrate
SnO2 CuO O Ka
Au 10 μm
Al2O3 substrate
c
EDXS of SnO2 film
AlKa
CuKa
CuLI
N Ka C Ka
SnLa
AlKa SiKa
SnLI
SnLb SnLb SnLg
1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 keV
d
109
nm (BT) SnO2 CuO 100 100 + 100 100 + 600 50 + 600
Sensor resistance, Rair (Ω)
108
O Ka
SnLa CuLa N KaCuLI SiKa C Ka
SnLI
SnLb SnLb SnLg
CuKb
107 106 105 104 103
CuKa CuKb
1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 keV
102 200
300
400 Temperature (°C)
500
Figure 3 | Control of the resistance of the device by deposition of top-layout nanoelectrodes. a, SEM image of a CuO nanoelectrode (100 nm bulk thickness, BT) on a highly porous SnO2 nanostructured film (100 nm bulk thickness) deposited on a sensor substrate with gold macro-electrodes. b,c, The SnO2 film lies between the CuO particle film and the gold electrodes, as indicated by EDXS spot analysis of the nanoelectrode (b) and functional (c) films, confirming the dominant presence of copper and tin, respectively, in each film. d, After deposition of 100–600-nm-thick CuO nanoelectrode films on 50–100-nm-thick SnO2 films, the sensor resistance drops by between one and four orders of magnitude at all temperatures. Nevertheless, in this top-layout configuration, the sensor response decreases with increasing nanoelectrode bulk thickness due to the catalytic activity of the CuO nanoparticles.
‘functional’ (n-type) domains. The film resistance is therefore reduced by using (p-type) CuO as the nanoelectrode and (n-type) SnO2 as the functional material. This requires much less conductive oxide nanoelectrodes than in conventional mixed oxides where the conductive and functional particles are deposited simultaneously16. Control of the nanoelectrode layout and deposition rate allows the desired electrical properties to be achieved. For the top layout (Fig. 1c), the nanoelectrode domains form a nearly continuous layer as electron flow is forced through part of the functional film. In contrast, for the bottom layout (Fig. 1b), a continuous nanoelectrode layer would shunt the functional film, ruining its performance. Nevertheless, for both layouts too few nanoelectrodes would have only a limited influence on the total film resistance. An optimal assembly therefore requires knowledge of the film functionality (current generation, gas sensing and so on) and deposition dynamics.
Standard electrode performance Figure 2 shows SnO2 sensing films arranged on standard macroelectrodes for the detection of CO (ref. 7) and EtOH (ref. 11). A continuous semiconducting film (bulk thickness, 50–600 nm) was deposited on the substrate with gold macro-electrodes by an orthogonally impinging aerosol of SnO2 nanoparticles17, resulting in a lace-like network of nanostructures (Fig. 2a) that evolves towards a cauliflower-like structure following in situ flame annealing7 (Fig. 2b). These structures form closed electrical paths between the macro-electrodes (Fig. 2a,b). 56
With increasing film thickness (and SnO2 specific volume), the sensor resistance (Fig. 2c, triangles) decreases asymptotically from 5 104 (bare substrate) to just 6.3 MV at 320 8C in dry air. The asymptotic value depends on the electrode geometry and film temperature. In contrast to conductive dense films, here the film resistance is not simply inversely proportional to film thickness and cannot be estimated by simple geometry9. In fact, the conductivity of a nanostructured, highly porous film is calculated by an array of in-parallel and in-series resistances9. Although it is possible to decrease the sensor resistance (Fig. 2c, triangles) by increasing the film thickness, such an approach is limited. Furthermore, this is not optimal, as maximal performance (Fig. 2c, circles) is obtained by the thinnest films, which also have the highest resistance (triangles) and cannot be easily integrated in portable devices7. Here, the sensor response to 10 ppm EtOH (Fig. 2c, circles) was increased five times by decreasing the thickness from 600 to 50 nm. These results are attributed to the higher penetration of analyte (for example, EtOH) in thin rather than in thick films18. Furthermore, decreasing the bulk thickness of the SnO2 sensors (Fig. 2d) from 800 to 50 nm decreased the response (triangles) and recovery (circles) times to 10 ppm EtOH from 81 and 217 s to 40 and 28 s, respectively. This is attributed to faster analyte penetration and adsorbed species (for example, OH) diffusion in thin (for example, 50 nm bulk thickness) than in thick films. The recovery time (28 s) of the present 50 nm bulk thickness SnO2 sensor is much faster than that (130 s) of nano-TiO2 sensors19, and the
NATURE NANOTECHNOLOGY | VOL 5 | JANUARY 2010 | www.nature.com/naturenanotechnology
NATURE NANOTECHNOLOGY
ARTICLES
DOI: 10.1038/NNANO.2009.349
a
a
105
Sensor resistance, Rair (MΩ)
2 µm +
104
50 nm SnO2 deposition
103
+
102
101
–
Substrate
320 °C dry air 0
b
–
Substrate
1
2
3
4
CuO bulk thickness, BT (nm)
CuO 2 µm
Sensor response, S = Rair /REtOH − 1
b
50 nm SnO2
300
50 nm SnO2 + 1.1 nm BT CuO 50 nm SnO2 + 2.3 nm BT CuO 50 nm SnO2 + 4.5 nm BT CuO
200
100
c 0 0
10
20
30
40
50
EtOH concentration (ppm) +
− Dielectric substrate
CuO nanoparticles
Figure 4 | Deposition of bottom-layout nanoelectrodes. a,b, SEM images of CuO (a) nanoelectrodes of a few nanometres deposited (1.1 nm equivalent bulk thickness) on Al2O3 sensor substrates (b, higher magnification). An open-circuit layer of CuO (b) nanoparticles is visible on the Al2O3 substrate grains. c, Schematic overview showing the standard electrodes (yellow) as well as the asymmetrically dispersed nanoelectrodes (green).
present response time (40 s) is comparable to that of the latter (30 s)19 and silver-doped SnO1.8 sensors (60 s)20.
Nanoelectrode performance for the top layout In the top layout, nanoelectrode particles are made using a flame aerosol process and directly deposited7 on the functional film of the sensors (Fig. 2). Figure 3a shows the resulting nanostructure morphology near a (gold) macro-electrode covered by a 100-nm thin porous SnO2 film followed by a 100-nm thin CuO nanoelectrode layer, as confirmed by energy-dispersive X-ray spectroscopy (EDXS) analysis (Fig. 3b,c). The nanostructured SnO2 film with the top nanoelectrode layer extends homogeneously over the substrate, following closely the shape of the macroelectrode. Deposition of these CuO particles from the aerosol does not damage the pre-deposited SnO2 film17. Figure 3d shows the effect of the top nanoelectrode thickness on film resistance. After deposition of just a 100-nm layer (bulk thickness) of CuO, the standard electrode sensor resistance (triangles) is decreased by at least one order of magnitude (diamonds) at all temperatures. Increasing the CuO top layer thickness from 100 to 600 nm (circles) decreases the sensor resistance by two to three orders of magnitude over that temperature range. Using a thinner
Figure 5 | Performance of bottom-layout nanoelectrodes. a,b, Sensor resistance (a) as a function of CuO bulk thickness (BT) and response (b) as a function of EtOH concentration. Up to 1.1 nm CuO bulk thickness, the sensor resistance (a, triangles) corresponds to the substrate resistance (a, triangles, 0 nm bulk thickness) indicating that an open-circuit CuO nanoelectrode layer is obtained. After deposition of a 50-nm (bulk thickness) functional SnO2 film (a, open symbols), the sensor resistance decreases by two to three orders of magnitude. Deposition of a 1.1-nm CuO bottom layer reduces the SnO2 sensor resistance (a, squares) by more than an order of magnitude. The sensor response to EtOH of that 1.1 nm bottomlayout nanoelectrode sensor (b, squares) is superior to that of pure SnO2 (b, circles), whereas those at 2.3 and 4.5 nm bulk thickness are remarkably lower (b, diamonds), indicating shunting of the sensor and the existence of an optimal bulk thickness or spatial distribution of CuO nanoelectrodes.
SnO2 layer (for example, 50 nm), the addition of CuO reduces the sensor resistance even more (squares). These low film resistances (,1 107 V) are convenient for portable micro-gas sensors, because they reduce the power required to heat the film to convenient temperatures (,350 8C) and are compatible with standard electronic microcircuitry21. Nevertheless, silver, platinum, gold and CuO nanoparticles tend to catalyse gas-sensing reactions22. Therefore, it may be challenging to achieve both low resistance and maximal sensing performance with a top nanoelectrode layout that could compete chemically with the semiconducting oxide. In fact, the sensor response to 50 ppm EtOH, here, was reduced from 150 at a SnO2 bulk thickness of 100 nm to just 1.8 by increasing the bulk thickness of the top nanoelectrode to 100 nm as with excess22 platinum on SnO2 sensors of CO. Deposition of 100 nm bulk thickness, top layout nanoelectrodes decreased the response time of the 50 nm bulk thickness SnO2 sensor from 40 to 10 s, most probably due to the catalytic consumption of the analyte that reaches the functional film. In contrast, the recovery time was not affected, probably because too little
NATURE NANOTECHNOLOGY | VOL 5 | JANUARY 2010 | www.nature.com/naturenanotechnology
57
ARTICLES
NATURE NANOTECHNOLOGY
DOI: 10.1038/NNANO.2009.349
Table 1 | Comparison of sensing response with various ethanol vapour concentrations (ppm) by different nanostructured film deposition methods. Deposition process
Material
Flame spray pyrolysis þ nanoelectrodes at bottom layout
SnO2
Combustion-assisted chemical vapour deposition
SnO2 SnO2 SnO2
Flame spray pyrolysis
Si:SnO2 Hot wall aerosol reactor with low-pressure impactor
Ag:SnO1.8
Molecular beam epitaxy Organometallic chemical vapour deposition Plasma-enhanced chemical vapour deposition
SnO2 SnO2 Pd:SnO2 SnOx SnO2 SnOx
Pulsed laser deposition Radiofrequency sputtering
analyte residue was formed on the functional film to affect its performance. The nearly 100-fold smaller bulk thickness of bottom-layout nanoelectrode when compared with top-layout nanoelectrodes is attributed to the higher layer density on flat (Al2O3 substrate) films than on highly porous SnO2 films.
Nanoelectrode performance for the bottom layout CuO nanoelectrodes (Fig. 4a,b) were deposited (forming a ‘layer’ of 1.1 nm equivalent bulk thickness) on the Al2O3 substrate containing gold macroelectrodes. This discontinuous layer of nanoelectrodes (Fig. 4b,c) corresponds to an ideal open electrical circuit (Fig. 1b, green particles). As discussed previously, this is mandatory when using a bottom layout in sensor applications, or the functional film will be shunted (Fig. 1e, broken line). A functional SnO2 film is deposited above and in between the nanoelectrodes and its morphology is identical to that of SnO2 films deposited on substrates with standard electrodes (Fig. 2a,b). A small thickness of SnO2 (50 nm) was chosen to obtain the highest sensitivity (Fig. 2c, circles). Figure 5 shows the electrical characterization of such ‘open circuit’ bottom nanoelectrode gas sensors as a function of the nanoelectrode bulk thickness. After deposition of a CuO ‘layer’ with equivalent bulk thickness of 2.3 nm, the measured resistance (Fig. 5a, filled triangles) decreases only slightly from 6 to 4 104 MV, indicating that the CuO particles barely improve the conductivity of the substrate. Nevertheless, this resistance (Fig. 5a, triangles) is only 30% smaller than that of bare substrates. So, an electrical path (‘closed circuit’) is still not formed between the macroelectrodes. Furthermore, increasing the thickness to 4.5 nm barely decreases the resistance (Fig. 5a, triangles). A continuous electrical path is detected only after deposition of 9 nm CuO (not shown), when a resistance of 2 104 MV is measured. Even if a closed circuit is formed, such high resistances (.1 104 MV) are feasible, because the bulk thickness of the CuO layer is only 9 nm and the CuO contact surface is much smaller than in dense films due to the granular shape of these aerosol-deposited films (Fig. 4b). The electrical circuit was closed by deposition of 1–4.5 nm equivalent bulk thickness CuO layers on top of the bare substrate, followed by a 50-nm bulk thickness of SnO2 film. The sensor resistance then decreased by nearly two orders of magnitude (Fig. 5a, circle) below that of the bare substrate and by three orders of magnitude when 1.1-nm equivalent bulk thickness of CuO nanoelectrodes (Fig. 5a, square) was deposited, confirming the 58
Response (R air/R EtOH 2 1) to ethanol (ppm) 0.5 (0.02) 2 (0.1) 50 (10) 307 (50) 673 (100) 279 (100) 5.3 (10) 20 (10) 45 (50) 50 (10) 318 (50) 2 (0.1) 30 (100) 1.8 (100) 8 (100) 14 (50) 17 (50) 6 (50) 1.5 (100)
Ref. This work
26 17 11 11 20 27 28 29 30 31 32
formation of a conductive path. Increasing the CuO bulk thickness from 1.1 to 4.5 nm did not significantly decrease the sensor resistance (30–50 MV). The interface between the thick SnO2 film and the thin (or spotty) CuO layer mainly affects the sensor resistance, reducing it by about an order of magnitude (Fig. 5a, square) as a result of patchy CuO deposits (1.1 nm bulk thickness). The bulk thickness of the bottom layout nanoelectrode affects, most remarkably, the sensor performance. In fact, the sensor response to 50 ppm EtOH with nanoelectrodes of 1.1 nm equivalent bulk thickness (Fig. 5b, squares) was nearly twice that of pure SnO2 (Fig. 5b, circles). Increasing that bulk thickness to 4.5 nm decreased the sensor response to 50 ppm ethanol from 306 to 31 (Fig. 5b, squares and diamonds) suggesting that the SnO2 film is partially shunted by the formation of dense nanoelectrode regions, even before reaching the percolation limit (for example, at 9 nm bulk thickness here) corresponding to a continuous current pathway through the entire film. In fact, close to the substrate surface, the nanoelectrode density is at its maximum with the stochastic (random) particle deposition used here and can increase rapidly above23 the percolation limit (16 vol%)24 for randomly dispersed spheres. This reduction points to the existence of an optimal nanoelectrode distribution for achieving the best performance in the gas sensors. Increasing the amount of CuO at the bottom makes the distance between the nanoelectrodes too short, so that the performance of the semiconductor is essentially swamped or shunted. It should be noted that any contribution of more p–n junctions (by increasing the bulk thickness of the CuO) to EtOH detection would have increased further the film resistance, thereby increasing the sensor response15, a feature that was not observed here. Nevertheless, the sensor response obtained here with 1.1 nm equivalent bulk thickness of CuO is extremely high (Table 1) and comparable only to that obtained by direct flame deposition11 of silicon-doped SnO2. In fact, too great thickness (for example, 800 nm bulk thickness) of flame spray pyrolysis (FSP)-made, pure SnO2 films led to a smaller sensor response to ethanol11–17 than here, as ethanol combusted18 on the electrical inert top layers of these films. Figure 6a shows the resistance of a (50 nm bulk thickness) SnO2 sensor with 1.1 nm bulk thickness, bottom-layout CuO nanoelectrodes in dry air in the presence of low (ppb) EtOH concentrations. The high sensitivity of these sensors (Fig. 5b, squares) allows detection of 20 ppb EtOH (Fig. 6a) with a good signal-to-noise ratio. The response (Fig. 6b) to 100 ppb EtOH is 2.1 and comparable to highly
NATURE NANOTECHNOLOGY | VOL 5 | JANUARY 2010 | www.nature.com/naturenanotechnology
NATURE NANOTECHNOLOGY
DOI: 10.1038/NNANO.2009.349
a
Sensor resistance (MΩ)
40
30 20 ppb EtOH 20
Dry air, 320 °C Bottom layout 50 nm BT SnO2 + 1.1 nm CuO
100 ppb EtOH
10
0
60
120
180
Time (min)
b Sensor response, Rair/REtOH − 1
100 ppb EtOH 2
1
20 ppb EtOH
0
0
60
120
Methods
c 120
Time (s)
nanoelectrode domains, because the response and recovery times of CuO are longer than those of SnO2 (not shown). Nevertheless, the response (60 s) and recovery (38 s) times of the 1.1 nm bulk thickness sensor are still comparable to the literature20 and smaller than that obtained with thick SnO2 sensors here (Fig. 2d, bulk thickness, 600 nm). In summary, the electrical properties of highly resistive nanoparticle films can be controlled by deposition of conductive domains above or below such films. Here, this was demonstrated for solidstate gas sensors made of highly sensitive SnO2 nanoparticles. Stochastic and asymmetric deposition of CuO nanoelectrodes at the bottom between conventional standard electrodes and functional SnO2 nanoparticles reduced the sensor resistance by an order of magnitude even at 1.1 nm CuO equivalent bulk thickness. This led to an assembly of ultrathin solid-state gas sensors with extremely high sensitivity to ethanol, close to the best in the literature (Table 1). Increasing, however, the bottom equivalent bulk thickness of CuO shunted the functional SnO2 film, indicating the existence of an optimal nanoelectrode bulk thickness or spatial distribution. The sensor resistance was decreased by up to three orders of magnitude after deposition of CuO nanoparticles in the top layout at up to 600 nm bulk thickness. Although, for sensors, the top-layout nanoelectrodes decreased sensing performance through catalytic combustion of the analyte, this may be quite appealing in other applications such as dye-sensitized solar cells. Finally, applications of these nanoelectrodes are not limited to gas sensors but include all devices containing semiconducting films in which electrical conduction plays an important role, for example in batteries, micro-/nanocircuitries, and solar and fuel cells.
180
Time (min)
60
Response time Recovery time 0 0
ARTICLES
2
4
CuO bulk thickness, BT (nm)
Figure 6 | Detection of ultralow EtOH concentrations by solid-state gas sensors with bottom-layout nanoelectrodes. a, Resistance of a SnO2 sensor (50 nm bulk thickness, BT) with 1.1 nm bottom nanoelectrodes in dry air and in the presence of 100 and 20 ppb EtOH. b, The small film thickness and SnO2 particle size allows detection of 20 ppb with a good signal-to-noise ratio. c, Response (circles) and recovery (triangles) times to 10 ppm EtOH increase with increasing nanoelectrode bulk thickness, but are always below those of thick (600 nm) SnO2 films (Fig. 2d).
sensitive sensors made of silver-doped and size-selected SnO2 nanoparticles (Table 1)20. Furthermore, minimal doping of the SnO2 nanoparticles with silicon11, platinum22 or silver20 could further increase the sensor response. Figure 6c shows the response (circles) and recovery (triangles) times of a 50-nm-thick SnO2 sensor as a function of the bulk thickness of the bottom-layout nanoelectrode. Increasing the bulk thickness to 4.5 nm increased the response and recovery times to 135 and 71 s, respectively. This is attributed to formation of larger
Particle and film synthesis. A FSP reactor was used in combination with a temperature-controlled substrate holder (120 8C) for synthesis and direct deposition of semiconducting SnO2 and conductive CuO nanoparticle layers onto Al2O3 substrates featuring a set of interdigitated gold electrodes and a sensing area of 1 cm2 (ref. 11). After deposition, the resulting films were mechanically stabilized by in situ annealing with a xylene-fed spray flame7. The SnO2 films were obtained by spraying and combusting tin (II)-ethylhexanoate (Aldrich, purity .98%) diluted in xylene with a total metal atom concentration of 0.5 M. Thick (.10 nm bulk thickness) CuO conductive layers were obtained by spraying 0.5 M copper (II) 2-ethylhexanoate (Sigma Aldrich) in xylene (Fluka, purity .98%), and thin CuO layers (1–5 nm bulk thickness) were made by spraying 0.01 M Soligen Copper 8 (OMG Borchers GmbH) in ethanol (Fluka, purity .98%). The spray was formed by feeding 5 ml min21 of precursor solution through a nozzle and dispersing it with 5 l min21 of oxygen (Pan Gas, 99.5%) with a constant pressure drop of 1.5 bar at the nozzle outlet. This spray was ignited by an annular pre-mixed methane/oxygen flame (CH4 ¼ 1.5 l min21 and O2 ¼ 3.2 l min21) surrounding the nozzle. Additionally, 5 l min21 of oxygen were supplied by a surrounding porous ring to ensure overstoichiometric and complete combustion. The substrate was placed at a height of 20 cm above the nozzle and kept at 120 8C by cooling with water during film deposition7. For stabilization, the substrate was set at a height of 14 cm above the nozzle and annealed in situ7 with an impinging xylene spray-flame fed at 12 ml min21 for 30 s. Particle and film characterization. During film deposition, non-deposited FSPmade particles were collected on a glass-fibre filter 30 cm downstream of the substrate. X-ray diffraction patterns were obtained using a Bruker, AXS D8 Advance diffractometer operated at 40 kV, 40 mA at 2u (Cu Ka) ¼ 15–758, step ¼ 0.038 and scan speed ¼ 0.68 min21. The crystal size (dXRD) was determined using the Rietveld fundamental parameter method with their structural parameters25. The powder specific surface area (SSA) was measured by nitrogen adsorption and Brunauer, Emmett and Teller (BET) analysis7 using a Micromeritics Tristar 3000. The BET equivalent diameter was calculated using the density of SnO2 (6.85 g cm23) and CuO (6.46 g cm23). Transmission electron microscopy (TEM) was conducted in a Hitachi H600 at 100 kV. The morphology, patterning characteristics and thickness of deposited layers were investigated by scanning electron microscopy (SEM) and EDXS with a Quanta 200F (operated at 20 kV), LEO 1530 Gemini (Zeiss/LEO, Oberkochen) and a Tecnai F30 microscope (FEI, Eindhoven; field-emission cathode, operated at 300 kV). The sensor measurements were performed as described elsewhere19. The film conductivity was computed from the electrode distance-to-length ratio, the film bulk thickness and resistance17. The analyte mixture consisted of EtOH (105 ppm+3% synthetic air, Pan Gas 5.0 or 10 ppm+3% synthetic air, Pan Gas 5.0) in synthetic air (20.8%+2% O2 rest nitrogen,
NATURE NANOTECHNOLOGY | VOL 5 | JANUARY 2010 | www.nature.com/naturenanotechnology
59
ARTICLES
NATURE NANOTECHNOLOGY
Pan Gas 5.0) and the temperature was measured with an n-type thermocouple19. The response time was defined as the time to reach 90% of the sensor response to 10 ppm EtOH. The recovery time was defined as the time to drop 90% of the sensor response to 10 ppm EtOH19. For bottom-layout nanoelectrodes, after increasing the CuO layer thickness to 50 nm the sensor resistance was reduced to 1 103 V, indicating a continuous conducting film. The resistance of this 50 nm bulk thickness CuO film corresponds to a conductivity of 0.45 S cm21, which is six orders of magnitude higher than that (0.52106 S cm21) of a SnO2 film having the same thickness. Furthermore, the CuO (50 nm bulk thickness) sensor response (for example, 5.6 at 100 ppm EtOH) was more than one order of magnitude smaller than that (for example, 165 at 100 ppm EtOH) of SnO2 at any EtOH concentration. These results show that the CuO nanoparticles exhibit better conductivity, but the functional SnO2 is better at gas sensing. These CuO films were chemically stable, recovering the baseline completely after EtOH/air cycling and indicating no reduction to metallic copper. Annealing the CuO films above 450 8C resulted in nanoparticle sintering and thus film shrinkage and/or cracking. This increased the film resistance and destabilized the sensor signal. Nevertheless, at the lower temperatures (for example, 320 8C) used here, the sensor signal in dry air or in the presence of EtOH was stable.
Received 3 August 2009; accepted 20 October 2009; published online 29 November 2009
References 1. Wu, N. L., Wang, S. Y. & Rusakova, I. A. Inhibition of crystallite growth in the sol-gel synthesis of nanocrystalline metal oxides. Science 285, 1375–1377 (1999). 2. O’Regan, B. & Gratzel, M. A low-cost, high-efficiency solar cell based on dyesensitized colloidal TiO2 films. Nature 353, 737–740 (1991). 3. Wang, P. et al. A stable quasi-solid-state dye-sensitized solar cell with an amphiphilic ruthenium sensitizer and polymer gel electrolyte. Nature Mater. 2, 402–407 (2003). 4. Poizot, P., Laruelle, S., Grugeon, S., Dupont, L. & Tarascon, J. M. Nano-sized transition-metal oxides as negative-electrode materials for lithium-ion batteries. Nature 407, 496–499 (2000). 5. Muecke, U. P. et al. Micro solid oxide fuel cells on glass ceramic substrates. Adv. Funct. Mater. 18, 3158–3168 (2008). 6. Kim, H. et al. Parallel patterning of nanoparticles via electrodynamic focusing of charged aerosols. Nature Nanotech. 1, 117–121 (2006). 7. Tricoli, A. et al. Micropatterning layers by flame aerosol deposition-annealing. Adv. Mater. 20, 3005–3010 (2008). 8. Kang, S. J. et al. High-performance electronics using dense, perfectly aligned arrays of single-walled carbon nanotubes. Nature Nanotech. 2, 230–236 (2007). 9. Barsan, N. & Weimar, U. Conduction model of metal oxide gas sensors. J. Electroceram. 7, 143–167 (2001). 10. Patey, T. J. et al. Flame co-synthesis of LiMn2O4 and carbon nanocomposites for high power batteries. J. Power Sour. 189, 149–154 (2009). 11. Tricoli, A., Graf, M. & Pratsinis, S. E. Optimal doping for enhanced SnO2 sensitivity and thermal stability. Adv. Funct. Mater. 18, 1969–1976 (2008). 12. Hagleitner, C. et al. Smart single-chip gas sensor microsystem. Nature 414, 293–296 (2001). 13. Ku¨hne, S. et al. Wafer-level flame-spray-pyrolysis deposition of gas-sensitive layers on microsensors. J. Micromech. Microeng. 18, 035040 (2008). 14. Mueller, R., Ma¨dler, L. & Pratsinis, S. E. Nanoparticle synthesis at high production rates by flame spray pyrolysis. Chem. Eng. Sci. 58, 1969–1976 (2003). 15. Tamaki, J., Maekawa, T., Miura, N. & Yamazoe, N. CuO–SnO2 element for highly sensitive and selective detection of H2S. Sens. Actuat. B 9, 197–203 (1992). 16. Chou, T. P., Zhang, Q., Russo, B. & Cao, G. Enhanced light-conversion efficiency of titanium-dioxide dye-sensitized solar cells with the addition of indium-tinoxide and fluorine-tin-oxide nanoparticles in electrode films. J. Nanophoton. 2, 023511 (2008).
60
DOI: 10.1038/NNANO.2009.349
17. Ma¨dler, L. et al. Direct formation of highly porous gas-sensing films by in situ thermophoretic deposition of flame-made Pt/SnO2 nanoparticles. Sens. Actuat. B 114, 283–295 (2006). 18. Becker, T., Ahlers, S., Bosch-v.Braunmuhl, C., Muller, G. & Kiesewetter, O. Gas sensing properties of thin- and thick-film tin-oxide materials. Sens. Actuat. B 77, 55–61 (2001). 19. Teleki, A., Pratsinis, S. E., Kalyanasundaram, K. & Gouma, P. I. Sensing of organic vapors by flame-made TiO2 nanoparticles. Sens. Actuat. B 119, 683–690 (2006). 20. Joshi, R. K. & Kruis, F. E. Influence of Ag particle size on ethanol sensing of SnO1.8:Ag nanoparticle films: a method to develop parts per billion level gas sensors. Appl. Phys. Lett. 89, 153116 (2006). 21. Graf, M., Gurlo, A., Barsan, N., Weimar, U. & Hierlemann, A. Microfabricated gas sensor systems with sensitive nanocrystalline metal-oxide films. J. Nanopart. Res. 8, 823–839 (2006). 22. Ma¨dler, L. et al. Sensing low concentrations of CO using flame-spray-made Pt/SnO2 nanoparticles. J. Nanopart. Res. 8, 783–796 (2006). 23. Rodriguez-Perez, D., Castillo, J. L. & Antoranz, J. C. Relationship between particle deposit characteristics and the mechanism of particle arrival. Phys. Rev. E 72, 021403 (2005). 24. Pecharroma´n, C., Esteban-Betego´n, F., Bartolome´, J. F., Lo´pez-Esteban, S. & Moya, J. S. New percolative BaTiO3–Ni composites with a high and frequency-independent dielectric constant (1r ¼ 80000). Adv. Mater. 13, 1541–1544 (2001). 25. Bolzan, A. A., Fong, C., Kennedy, B. J. & Howard, C. J. Structural studies of rutile-type metal dioxides. Acta Crystallogr. B 53, 373–380 (1997). 26. Liu, Y., Koep, E. & Liu, M. L. Highly sensitive and fast-responding SnO2 sensor fabricated by combustion chemical vapor deposition. Chem. Mater. 17, 3997–4000 (2005). 27. Kroneld, M. et al. Gas sensing properties of SnO2 thin films grown by MBE. Sens. Actuat. B 118, 110–114 (2006). 28. Bruno, L., Pijolat, C. & Lalauze, R. Tin dioxide thin-film gas sensor prepared by chemical vapour deposition: influence of grain size and thickness on the electrical properties. Sens. Actuat. B 18, 195–199 (1994). 29. Lee, Y. C., Huang, H., Tan, O. K. & Tse, M. S. Semiconductor gas sensor based on Pd-doped SnO2 nanorod thin films. Sens. Actuat. B 132, 239–242 (2008). 30. Hellegouarc’h, F., Arefi-Khonsari, F., Planade, R. & Amouroux, J. PECVD prepared SnO2 thin films for ethanol sensors. Sens. Actuat. B 73, 27–34 (2001). 31. Zhao, Y., Feng, Z. & Liang, Y. SnO2 gas sensor films deposited by pulsed laser ablation. Sens. Actuat. B 56, 224–227 (1999). 32. Bittencourt, C. et al. Influence of the deposition method on the morphology and elemental composition of SnO2 films for gas sensing: atomic force and X-ray photoemission spectroscopy analysis. Sens. Actuat. B 92, 67–72 (2003).
Acknowledgements The authors acknowledge stimulating discussions with T.D. Elmoe, and thank M. Righettoni and H. Keskinen for assistance with the experiments, F. Krumeich for electron microscope analysis and Competence Centre for Materials Science and Technology (CCMX) and NANocrystalline CERamic thin film coatings (NANCER) for financial support.
Author contributions A.T. and S.E.P. conceived and designed the experiments. A.T. performed the experiments. A.T. and S.E.P. analysed the data and co-wrote the paper.
Additional information The authors declare no competing financial interests. Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/. Correspondence and requests for materials should be addressed to S.E.P.
NATURE NANOTECHNOLOGY | VOL 5 | JANUARY 2010 | www.nature.com/naturenanotechnology
