micropatterning layers by flame aerosol deposition-annealing

6
DOI: 10.1002/adma.200701844 Micropatterning Layers by Flame Aerosol Deposition-Annealing** By Antonio Tricoli, Markus Graf, Felix Mayer, Ste´phane Ku ¨ hne, Andreas Hierlemann, and Sotiris E. Pratsinis* The development of scalable methods that effectively bind nanoparticles to surfaces and provide precise patterns is a key step towards the commercial exploitation of the distinctive properties of nanoparticles. [1] In this context, the direct deposi- tion from aerosols [2] is a powerful tool that extends the downscaling prospective of solid-state devices, [3] relevant in environmental, [4] energy storage, [5] sensor, [6] catalytic, [7] and biological [8] applications. Here, the most widely used aerosol technology for the manufacture of materials (e.g., fumed silica, pigmentary titania, and carbon blacks), namely the flame aerosol reactor, [9] has been used for direct synthesis, precise deposition, and stabilization of well-defined nanoparticle patterns on microelectronic substrates. The procedure yields gas-sensitive metal oxide layers that have been micropatterned on integrated-circuitry substrates under close control of syn- thesis and operation temperatures and conditions. Further- more, the presented results have much broader implications, as they demonstrate a scalable synthesis method of layers with controlled texture, adhesion, and cohesion. Such a method will be important for a number of applications including, for example, solid-oxide fuel cells, self-cleaning photocatalytics, antireflec- tives, semiconductors, optical materials, and biosensor films. Recent advances in microtechnology [10] have led to a new generation of micromachined, smart, single-chip gas sensors that rely on complementary metal oxide semiconductor (CMOS) technology and include integrated circuitry. The realization of microelectromechanical system (MEMS) structures on CMOS substrates has significantly increased the potential of this technology. Solid-state sensor devices that rely on nanocrystal- line metal-oxide gas-sensitive films have been devised. These films usually are deposited onto the micromachined hotplates by means of a drop-coating method, a tedious serial process, using slurries of preprocessed powders. [3] In comparison to their macrosize predecessors, these microdevices offer improved features, such as low power consumption and the possibility to perform fast thermal cycling owing to the small thermal mass of the coated transducer and the optimized isolation of the heated area. Integrated electronic components can be used to amplify and stabilize the respective sensor signals. Nevertheless, the industrial applicability of such devices strictly depends on a reliable batch technology for the sensing layer deposition. [11] In fact, strict requirements are imposed by the presence of integrated circuitry components on the substrate, the microsize of transducers and components (micrometer precision of the deposition method), and the need to perform all processing steps, including micromachining and layer deposition, on the wafer-level to minimize the costs per unit. [11] Aerosol-based methods have been used for synthesis of nanostructured sensing layers on macroscopic sensor sub- strates. The applied methods included sputtering, [12] spray pyrolysis, [13] cluster beam deposition, [14] classified aerosol deposition, [15] combustion chemical vapor deposition, [16] or flame spray pyrolysis (FSP). [17] The last two methods in particular are also very attractive for application to microscale substrates, as they potentially enable a wafer-level deposition on an industrial scale. This holds particularly true since nano- particles are generated by means of scalable flame reactors. Furthermore, a broad range of materials can be prepared using these liquid-fed flames so that the limitations of conventional vapor-fed flames with regard to the availability of gaseous metal-oxide precursors can be overcome. [9] However, the integration of flame-based aerosol methods into the industrial fabrication of semiconductor devices is not trivial, as the mechanical stability of the nanostructured layers depends on the substrate temperature during deposition. At high temperatures (850 8C), mechanically stable layers are obtained, [16] but these high temperatures are incompatible with CMOS substrates that carry circuitry components, which cannot tolerate temperatures above 400 8C. The deposition or substrate temperature can be reduced, the resulting layers, COMMUNICATION [*] Prof. S. E. Pratsinis, A. Tricoli Particle Technology Laboratory Department of Mechanical and Process Engineering ETH Zurich 8092 Zurich (Switzerland) E-mail: [email protected] Dr. M. Graf, Dr. F. Mayer Sensirion AG 8712 Staefa Zurich (Switzerland) S. Ku ¨hne, Prof. A. Hierlemann Physical Electronics Laboratory Department of Physics ETH Zurich 8093 Zurich (Switzerland) [**] Financial support was provided by Nanoprim and KTI #7745.1. The authors would like to thank Dr. F. Krumeich (EMEZ ETH Zurich) for microscopic analyses. Adv. Mater. 2008, 20, 3005–3010 ß 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 3005

Upload: independent

Post on 27-Apr-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

COM

MU

DOI: 10.1002/adma.200701844 N

ICATIO

Micropatterning Layers by Flame AerosolDeposition-Annealing**

N

By Antonio Tricoli, Markus Graf, Felix Mayer, Stephane Kuhne, Andreas Hierlemann,

and Sotiris E. Pratsinis*

The development of scalable methods that effectively bind

nanoparticles to surfaces and provide precise patterns is a key

step towards the commercial exploitation of the distinctive

properties of nanoparticles.[1] In this context, the direct deposi-

tion from aerosols[2] is a powerful tool that extends the

downscaling prospective of solid-state devices,[3] relevant in

environmental,[4] energy storage,[5] sensor,[6] catalytic,[7] and

biological[8] applications. Here, the most widely used aerosol

technology for the manufacture of materials (e.g., fumed silica,

pigmentary titania, and carbon blacks), namely the flame

aerosol reactor,[9] has been used for direct synthesis, precise

deposition, and stabilization of well-defined nanoparticle

patterns on microelectronic substrates. The procedure yields

gas-sensitive metal oxide layers that have been micropatterned

on integrated-circuitry substrates under close control of syn-

thesis and operation temperatures and conditions. Further-

more, the presented results havemuch broader implications, as

they demonstrate a scalable synthesis method of layers with

controlled texture, adhesion, and cohesion. Such a method will

be important for a number of applications including, for example,

solid-oxide fuel cells, self-cleaning photocatalytics, antireflec-

tives, semiconductors, optical materials, and biosensor films.

Recent advances in microtechnology[10] have led to a new

generation of micromachined, smart, single-chip gas sensors

that rely on complementarymetal oxide semiconductor (CMOS)

technology and include integrated circuitry. The realization of

microelectromechanical system (MEMS) structures on CMOS

[*] Prof. S. E. Pratsinis, A. TricoliParticle Technology LaboratoryDepartment of Mechanical and Process EngineeringETH Zurich8092 Zurich (Switzerland)E-mail: [email protected]

Dr. M. Graf, Dr. F. MayerSensirion AG8712 Staefa Zurich (Switzerland)

S. Kuhne, Prof. A. HierlemannPhysical Electronics LaboratoryDepartment of PhysicsETH Zurich8093 Zurich (Switzerland)

[**] Financial support was provided by Nanoprim and KTI #7745.1. Theauthors would like to thank Dr. F. Krumeich (EMEZ ETH Zurich) formicroscopic analyses.

Adv. Mater. 2008, 20, 3005–3010 � 2008 WILEY-VCH Verlag G

substrates has significantly increased the potential of this

technology. Solid-state sensor devices that rely on nanocrystal-

line metal-oxide gas-sensitive films have been devised. These

films usually are deposited onto the micromachined hotplates

by means of a drop-coating method, a tedious serial process,

using slurries of preprocessed powders.[3] In comparison to

their macrosize predecessors, these microdevices offer

improved features, such as low power consumption and the

possibility to perform fast thermal cycling owing to the small

thermal mass of the coated transducer and the optimized

isolation of the heated area. Integrated electronic components

can be used to amplify and stabilize the respective sensor

signals. Nevertheless, the industrial applicability of such

devices strictly depends on a reliable batch technology for

the sensing layer deposition.[11] In fact, strict requirements are

imposed by the presence of integrated circuitry components on

the substrate, the microsize of transducers and components

(micrometer precision of the deposition method), and the need

to perform all processing steps, including micromachining and

layer deposition, on the wafer-level to minimize the costs per

unit.[11]

Aerosol-based methods have been used for synthesis of

nanostructured sensing layers on macroscopic sensor sub-

strates. The applied methods included sputtering,[12] spray

pyrolysis,[13] cluster beam deposition,[14] classified aerosol

deposition,[15] combustion chemical vapor deposition,[16] or

flame spray pyrolysis (FSP).[17] The last two methods in

particular are also very attractive for application to microscale

substrates, as they potentially enable a wafer-level deposition

on an industrial scale. This holds particularly true since nano-

particles are generated by means of scalable flame reactors.

Furthermore, a broad range of materials can be prepared using

these liquid-fed flames so that the limitations of conventional

vapor-fed flames with regard to the availability of gaseous

metal-oxide precursors can be overcome.[9]

However, the integration of flame-based aerosol methods

into the industrial fabrication of semiconductor devices is not

trivial, as the mechanical stability of the nanostructured layers

depends on the substrate temperature during deposition. At

high temperatures (850 8C), mechanically stable layers are

obtained,[16] but these high temperatures are incompatible

with CMOS substrates that carry circuitry components, which

cannot tolerate temperatures above 400 8C. The deposition or

substrate temperature can be reduced, the resulting layers,

mbH & Co. KGaA, Weinheim 3005

COM

MUNIC

ATIO

N

Figure 1. Process schematic: a) highly porous (98%) flame-made nanoparticle layers are deposited on silicon wafer through a shadowmask, and then b) insitu mechanically stabilized by an impinging, particle-free xylene flame.

Table 1. Flame conditions and SnO2 nanoparticle characteristics.

Flame A B C D

Precursor Concentration [mol l�1] 0.5 0.5 0.5 1

Precursor Flow [ml min�1] 5 8 8 8

Dispersion O2 [l min�1] 5 5 3 3

Crystallite Size, dXRD [nm] 12� 1 14� 1 18� 1 22� 1

Grain Size, dBET [nm] 9� 1 12� 1 18� 1 21� 1

Specific Flame Enthalpy [kJ l�1] 39 63 105 114

Flame Height [cm] 13 15 19 21

3006

however, feature a poor mechanical stability that is not

sufficient to survive the CMOS post-processing steps such as

dicing (water jet in the saw).

Therefore, we developed the concept of a thermally induced

deposit stabilization based on in situ, rapid flame treatment of

low-temperature-deposited nanoparticle micropatterns. This

method can be applied while maintaining low substrate tem-

peratures, and the initial crystallite size is preserved (Fig. 1). As

will be shown, the developed deposition and stabilization

concept provides a wafer-level, rapid micropatterning method

and the resulting CMOS-compatible microsensors feature

good sensitivity and stability, as will be demonstrated in gas

and liquid jet impingement test measurements. The micro-

sensors featured sensitive layers of pure or Pt-doped SnO2

nanoparticles.

Figure 1 shows a schematic of the deposition concept. The

metal oxide nanoparticle synthesis takes place by dispersion

and ignition of a precursor spray solution of the target species

(e.g., SnO2). Depending on solution composition, its combus-

tion leads to metal oxide crystallites and, if noble-metal

precursors are present, to additional noble-metal cluster

formation, typically on top of the metal oxide surface.

Further particle growth occurs by condensation, surface

growth, coagulation, and sintering, which yields nanocrystal-

line material (Fig. 1a). The particle size (Table 1) is controlled

by the particle residence time in the high temperature zone, the

metal concentration, and the droplet dispersion during FSP as

detailed previously.[18] These newly formed nanostructured

particles deposit to the substrate, resulting in a layer of high

porosity but with low mechanical stability owing to the low

temperature of the substrate (150 8C). The layer is stabilized in

a second step by applying in situ high thermal and mechanical

stress for 30 s using a particle-free xylene flame (Fig. 1b). This

drastically changes the layer texture and increases its cohesion

and adhesion to the substrate.

www.advmat.de � 2008 WILEY-VCH Verlag GmbH &

Nanoparticle patterns (Fig. 2a) down to 100mm in diameter

(Fig. 2c) have been achieved by placing a shadow mask

featuring an array of circular openings in front of a Si-wafer

(with a thin silicon nitride layer) in cross-flow to the hot,

particle-laden FSP-jet (Fig. 1a). The deposited microlayers

(Fig. 2a–c) are highly uniform, crack-free, and quite porous

with a lace-like structure (Fig. 2d). The deposition zone shape

is reasonably well defined; only few particles pass between the

shadow mask and its support structure. Outside the mask

opening, the density of the layer rapidly decreases and reaches

zero at about 10mm away from the opening edges as indicated

by the thin ‘‘halo’’ around the deposits (Fig. 2a–c).

Different magnifications of as-deposited patterns (Table 1:

flame A, 180 s deposition time) reveal the surface morphology.

The uniform, self-assembled lace-like structure extends for

several hundred micrometers and consists of an intricate net of

thin nanostructures (Fig. 2d) that are loosely connected to each

other by long particle bridges. The visible (microscopy)

intralayer pore size is in the range of 100–400 nm, while the

particle bridges appear to be rather softly agglomerated

structures of 5–50 nm thickness (Fig. 2e).

The stabilization of these deposits by in situ annealing (Fig.

1b) drastically changed the layer morphology (Fig. 3a and b).

Co. KGaA, Weinheim Adv. Mater. 2008, 20, 3005–3010

COM

MUNIC

ATIO

N

Figure 2. Micropatterning of Pt/SnO2 particles made from flame A (Table 1) using a metalshadow mask with an array of circular openings (Ø: 100–800mm) to create the respectivepatterns: a) scanning electronmicroscopy images of circular micropatterns, and b,c) well-definedspots down to a spot diameter of 100mm (c). d)The resulting self-assembled, lace-like layers arehighly uniform and porous Their pore sizes are on the order of 300 nm and e) 5–50 nm thinlyagglomerated particle bridges extend for hundreds of nanometers.

Figure 3. a,b) Pt/SnO2 layers prepared by flame A, after in situ annealing.The initial layer uniformity (Fig. 2d) is preserved (a), but its morphology isaltered to more robust, cauliflower-like structures (b). The latter are on theorder of 1mm in diameter featuring fine pores of 10–50 nm diameter.

Figure 4. A cross-sebefore and b) afteruniform thickness a(b) the layer thickne

Adv. Mater. 2008, 20, 3005–3010 � 2008 WILEY-VCH Verlag GmbH & Co. KGaA,

Although the large-scale layer uniformity

(Fig. 2d) has been preserved (Fig. 3a), the

dense net of thin bridges transforms into

more compact, cauliflower-like structures

(Fig. 3b), and the intralayer pores are

enlarged (pore size� 860 nm). These cauli-

flower-like structures with a diameter of

about 1.5mm each exhibit pores in the range

of tens of nanometers (Fig. 3b). The crystal

size, however, is the same as that measured

directly after deposition (Table 1), even for

up to 60 s in situ annealing. The high

temperature of the xylene flame initiates

sintering of the nanobridges that form

thicker structures of higher stability. The

weakest connections between bridges rapidly

shrink and break, while the rest remain

anchored at the most-stable side.

In situ annealing also reduced the layer

thickness (Fig. 4) from (28� 3)mm for the

lace-like layer (Fig. 4a) to (1.5� 0.7)mm for

the cauliflower-like one (Fig. 4b). A layer

mass of (8.6� 0.5)mg was measured after

180 s deposition with flame A on a substrate

surface of 54.5 cm2. After in situ annealing

the layer mass was (8.4� 0.5)mg. Likewise, the volume-based

layer porosity has been reduced from 98% to 62%. The

increased material density upon annealing also increases the

contact surface between nanomaterial and electrodes. This

improves layer adhesion while nanoparticle cohesion is also

increased by necking through sintering.

This open, cauliflower-like structure facilitates analyte

transport through the sensitive layer so that analyte-induced

resistance changes, which constitute the sensor response, occur

almost instantaneously in the whole sensitive layer. This

positively influences the sensor response/recovery time and has

ctional cut through a layer deposited by flame A a)in situ annealing. The as-deposited layers show and a rather compact layer structure. After annealingss is strongly reduced.

Weinheim www.advmat.de 3007

COM

MUNIC

ATIO

N

Figure 5. Equivalent cleared area diameter, dCA, upon impinging a N2 jetonto Pt/SnO2 layers deposited by flame A (diamonds, triangle), and SSA(circles) of the corresponding nanoparticles as a function of annealingtemperature in an oven during 4 h. The dCA follows the SSA, indicating thatparticle sintering increases layer cohesion and its adhesion to the sub-strate. The in situ annealed samples (triangle) do not show any layerremoval upon applying the N2 jet.

Figure 6. Sensor response of an in situ annealed Pt/SnO2 layer (flame D)upon alternating exposure to increasing CO concentrations (1–40 ppm)and pure synthetic air at 450 8C and 25% relative humidity.

3008

an impact on its sensitivity, as more sensor surface is

simultaneously exposed to the analyte.

Themechanical stability of these layers (flameA), deposited

on a Si-wafer, was measured by vertically impinging a N2 jet.[19]

The layers had been annealed either conventionally in an oven

at 400–900 8C for 4 h or by using in situ annealing with the flame

for 30 s.

The diameter of the substrate area cleared of nanoparticles

in the area of the N2 jet impingement location, dCA, (Fig. 5,

diamonds) decreased with increasing oven temperature. For

example, annealing at 900 8C decreased the dCA from about 1

to 0.3 cm. This is consistent with the expected improvement in

particle cohesion and adhesion to the wafer surface by sintering

through annealing, which mechanically stabilizes the layer. A

decreasing specific surface area (SSA) of the deposited

nanoparticles (Fig. 5, circles) with increasing annealing

temperature further supports the hypothesis of enhanced

particle sintering and better particle adhesion to the substrate.

On the other hand, in situ annealing by the xylene spray flame

drastically enhances the layer stability, so that there is not any

particle removal upon jet impingement (Fig. 5, triangle).

To further investigate this remarkable strengthening of the

layer after in situ annealing, the N2 jet was substituted by a

water jet that is similar to the one used during CMOS wafer

dicing. Without any thermal stabilization, the as-deposited

layer was immediately destroyed upon coming in contact with

water: the particles were dispersed by capillary forces. This is

an indication that the flame-deposited particles (Fig. 2a) were

only loosely attached to the substrate. In contrast, in situ

annealed layers completely preserved their structure and

integrity.

The sensing performance of the layers was thereafter

assessed using CMOS-compatible micro-machined hotplates

www.advmat.de � 2008 WILEY-VCH Verlag GmbH &

(Fig. 1, inset) on a Si-wafer. They consist of a thermally isolated

sensing area (closed membrane) with 2 interdigitated Pt

electrodes (4 fingers each as shown in Fig. 1a), two embedded

electrical heaters, and three chip temperature sensors (one is

shown).

Figure 6 shows the resistance changes of a microsensor

featuring a Pt/SnO2 layer (deposition time¼ 180 s, flame D in

Table 1) after deposition-annealing upon exposure to various

CO concentrations at a hotplate temperature of 450 8C(heating by integrated Pt heaters). The conditions of flame

D were chosen because they enable the formation of larger

particles (Table 1) and thus reduced the sensor baseline

resistance.

A high sensor baseline resistance (>6� 108Ohm) was

observed despite the high operation temperature of the

microhotplate (450 8C). The sensor showed some baseline

drift and, at concentrations higher than 20 ppm CO, the sensor

equilibrium time increased a little (Fig. 6). Microsensors are

particularly attractive for measurements in the low concentra-

tion range while higher CO concentrations may lead to a

baseline drifts owing to the strong interaction between CO and

Pt/doped SnO2.[20] The high material resistance is a conse-

quence of the low thermal conductivity of the ‘‘as-deposited’’

layer, so that a sharp temperature drop occurs toward its outer

surface. This temperature drop dramatically decreases the

layer electrical conductivity. Further evidence for this effect

was provided by the considerably lower material resistance

upon applying external heating in an oven (Fig. 7).

The results of isothermal sensor measurements with the

whole measurement chamber in an oven are remarkably

different (Fig. 7). Upon isothermal conditions the microsensor

baseline resistance is drastically reduced from about 109

(Fig. 6) to 107–108Ohm (Fig. 7a), even at chamber tempera-

tures that are lower (320–370 8C) than that of the micro-

Co. KGaA, Weinheim Adv. Mater. 2008, 20, 3005–3010

COM

MUNIC

ATIO

N

Figure 7. Microsensor response (flame D for layer preparation) to a) COand b) ethanol upon external heating (Tch¼ temperature in the chamber).The baseline resistance and the sensitivity to CO are considerably lower incomparison to the measurements with integrated heaters (Fig. 6).

hotplate surface (450 8C, Fig. 6). The microsensor baseline was

stable for all tested concentrations for both analytes, CO

(Fig. 7a) and ethanol (EtOH; Fig. 7b).

Finally, it was possible to simultaneously generate 69 SnO2

microsensor layer patterns of about (28� 3)mm thickness (Fig.

1, inset shadow mask) on a wafer within 3min using a shadow

mask featuring an array of circular openings of uniform size.

The layers have also been stabilized on the wafer by in situ

annealing. The mask openings have been aligned with the

target deposition areas over the whole wafer. The openings

have been decreased to the microscale, as detailed previously.

The number of openings and the shadow mask size can be

further increased to allow for simultaneous layer deposition on

a larger number of microsensors and on larger wafers. The

large sensor responses to different analytes and the short

response and recovery times as a consequence of the highly

Adv. Mater. 2008, 20, 3005–3010 � 2008 WILEY-VCH Verl

porous layer morphology and the small particle size are highly

advantageous for the development of portable gas micro-

sensors.

A scalable flame aerosol technology for the rapid deposition

of micropatterned stable, functional inorganic nanomaterial

layers was presented, and it is compatible with integrated-

circuit technology. Flame-made nanoparticles of carefully

controlled crystallinity and morphology have been formed and

patterned onto a substrate through a mask at a high deposition

rate. This is followed by a rapid (30 s) in situ annealing that

drastically improves the layer cohesion and its adhesion to the

substrate. During this annealing step, the layer morphology

changes from a self-assembled, lace-like texture to a robust,

cauliflower-like one. The annealing step does not alter the

nanocrystallinity, and the substrate temperature during

annealing does not exceed the CMOS limit (400 8C).The overall procedure yields stable layers that can withstand

subsequent sensor processing steps (e.g., cleaning, dicing,

machining) and exhibit excellent sensor performance for

standard analytes such as CO and ethanol. Furthermore, the in

situ annealing of micropatterns can be applied to a wide range

of nanoparticle compositions and coating processes such as

filtration[21] and electrophoretic deposition,[2] since it rapidly

transforms highly porous layer or film morphologies to denser

ones firmly adhering to the substrate. This is a key process

design feature that facilitates the development of large-scale

nanoparticle deposition methods.

Experimental

A flame spray pyrolysis (FSP) unit [17] was used in combinationwith a water-cooled substrate holder and a shadow mask (Fig. 1a) forsynthesis and direct deposition of thick (�30mm) layers of pure andPt-doped SnO2 nanoparticles. In a second step (Fig. 1b), the layerswere mechanically stabilized by in-situ annealing with a particle-free,xylene-FSP flame. Such layers were FSP-deposited on 800mm thick Siwafers featuring a radius of 9.8 cm and a 100-nm-thick Si-nitridepassivation layer. Alternatively these layers were FSP-deposited alsoon microhotplates that have been micro-machined on wafers (Fig. 1a).The membrane type microhotplates feature a sensitive area of300� 300mm2, Pt-interdigitated electrodes (4 fingers each), anembedded heater and three temperature sensors (one on themembrane, one in close proximity and one on the bulk chip), whichare electrically isolated by a Si-nitride passivation layer.

Sensing nanoparticles were prepared as follows: First, tin(II)-ethylhexanoate (Aldrich, purity> 98%) was diluted in xylene(Fluka, purity> 98.5%) with a metal-atom concentration of 0.5–1mol l�1 (Table 1). This solution was supplied at a rate of 5–8ml min�1

through a nozzle and dispersed to a fine spray with 3–5 l min�1 oxygen(pressure drop 1.5 bar). That spray was ignited by an annular-shapepremixed methane/oxygen flame (CH4¼ 1.5 l min�1, O2¼ 3.2 l min�1).An additional 5 l min�1 sheath oxygen flow was supplied from anannulus surrounding that flame to assure over-stoichiometric combus-tion conditions (Fig. 1a). Powder samples were collected with a vacuumpump (Vacuumbrand, RE 16) on a water-cooled glass-fiber filter (GF/D Whatman, 257mm diameter) 50 cm above the burner.

The temperature of the substrate during deposition (Fig. 1a) wasmaintained between 140–160 8C in order to ensure a maximumdeposition rate and to avoid water condensation that may alter the

ag GmbH & Co. KGaA, Weinheim www.advmat.de 3009

COM

MUNIC

ATIO

N

3010

layer properties. The temperature on the back side of the substrate wasmeasured by an n-type thermocouple. The nozzle-substrate (NS)distance during deposition depended on the flame enthalpy density:20 cm forA andB, 22 cm for C and 24 for D to keep the above substratetemperatures.

The hot flame gas jet, carrying the freshly formed nanoparticles,impinged on the water-cooled substrate (Fig. 1a). The temperaturegradient between the impinging gas and the substrate surface generatesa thermophoretic particle flux directed to the substrate surface withsubsequent particle deposition and layer growth. Layer microstructur-ing was achieved by using shadow masks symmetrically aligned withrespect to the particle flow to the substrate. Structuring on wafer-levelwas performed using a metallic shadow mask (Fig. 1a, inset).

The layers were annealed in-situ (Fig. 1b) by reducing the NS to14 cm and by impinging a particle-free (no metal precursor) xylenespray flame onto the shadow mask for 30–60 s. The dispersion oxygenflow rate was 5 l min�1 and the xylene feed rate was 12ml min�1. Thesupporting flame and sheath oxygen flows were the same as during FSPdeposition. The maximum temperature measured on the back of thesubstrate was 220 8C, which is far below the CMOS limit ofapproximately 400 8C.

X-ray diffraction patterns were obtained by a Bruker, AXS D8Advance diffractometer operated at 40 kV, 40mA at 2u (CuKa)¼ 15 8–75 8, step¼ 0.03 8 and scan speed¼ 0.6 8 min�1. The crystalsize (dXRD) was determined using the Rietveld fundamental parametermethod with the structural parameters of cassiterite [22]. The powderspecific surface area (SSA) was measured by BET analysis using aMicromeritics Tristar 3000. The BET equivalent diameter wascalculated using the density of SnO2 as 6.85 g cm�3. Transmissionelectron microscopy was conducted in a Hitachi H600, operated at100 kV. The morphology, patterning characteristics and thickness ofthe deposited layers were investigated using an SEM with a LEO 1530Gemini (Zeiss/LEO, Oberkochen) and an optical microscope. Thelayer porosity was determined by SEM and gravimetric analysis ofwafer sections before (e.g., 600mg) and after deposition (e.g., 610mg).

The layer mechanical stability was tested by directing a N2 or waterjet towards its surface from 0.5 cm height. The N2 jet was produced by16 l min�1 of N2 through a nozzle (1.2mm ID) at 4 bar. The water jetwas produced by connecting a water reservoir at a pressure of 20 bars toa nozzle (0.9mm ID) resulting in a flow rate of 0.3 l min�1.

The sensor measurements were performed on a gas manifoldconsisting of six mass flow controllers (SensirionAG), twomultimeters(Agilent 34401A and HP 3458A), a function generator (Agilent33220A), a data acquisition/switch unit (Agilent 34970A), threethermally isolated gas-tight measurement chambers, and a computerfor data sampling and setup control.

Sensor measurements were performed with CO (100 ppm� 2% insynthetic air, Messer, Griesheim 5.0), EtOH (1022 ppm� 3% syntheticair, Messer, Griesheim 5.0) and synthetic air (20.8% O2� 2% restnitrogen, Messer, Griesheim 5.0). The microsensor temperature wasmeasured either on the microhotplate surface with the embeddedthermosensor, when heating with the embedded Pt-heaters, or with a

www.advmat.de � 2008 WILEY-VCH Verlag GmbH &

n-type thermocouple placed in the proximity of the microsensorposition, as it was done during the isothermal measurements ofFigure 7, when the whole chamber was heated.

Received: July 27, 2007Revised: October 9, 2007

[1] V. I. Klimov, A. A.Mikhailovsky, S. Xu, A.Malko, J. A. Hollingsworth,

C. A. Leatherdale, H. J. Eisler, M. G. Bawendi, Science 2000, 290, 314.

[2] T. J. Krinke, H. Fissan, K. Deppert, M. H. Magnusson, L. Samuelson,

Appl. Phys. Lett. 2001, 78, 3708.

[3] M. Graf, D. Barrettino, S. Taschini, C. Hagleitner, A. Hierlemann, H.

Baltes, Anal. Chem. 2004, 76, 4437.

[4] M. E. Franke, T. J. Koplin, U. Simon, Small 2006, 2, 36.

[5] Z. P. Shao, S. M. Haile, J. Ahn, P. D. Ronney, Z. L. Zhan, S. A.

Barnett, Nature 2005, 435, 795.

[6] V. Maheshwari, R. F. Saraf, Science 2006, 312, 1501.

[7] S. Thybo, S. Jensen, J. Johansen, T. Johannessen, O. Hansen, U. J.

Quaade, J. Catal. 2004, 223, 271.

[8] S. J. Park, T. A. Taton, C. A. Mirkin, Science 2002, 295, 1503.

[9] H. K. Kammler, L. Madler, S. E. Pratsinis, Chem. Eng. Technol. 2001,

24, 583.

[10] C. Hagleitner, A. Hierlemann, D. Lange, A. Kummer, N. Kerness, O.

Brand, H. Baltes, Nature 2001, 414, 293.

[11] M. Graf, A. Gurlo, N. Barsan, U. Weimar, A. Hierlemann, J. Nano-

part. Res. 2006, 8, 823.

[12] L. Y. Sheng, Z. N. Tang, J. Wu, P. C. H. Chan, J. K. O. Sin, Sens.

Actuators B 1998, 49, 81.

[13] G. Korotcenkov, Sens. Actuators B 2005, 107, 209.

[14] T. Mazza, E. Barborini, I. N. Kholmanov, P. Piseri, G. Bongiorno, S.

Vinati, P.Milani, C. Ducati, D. Cattaneo, A. Li Bassi, C. E. Bottani, A.

M. Taurino, P. Siciliano, Appl. Phys. Lett. 2005, 87, 103.

[15] M. K. Kennedy, F. E. Kruis, H. Fissan, B. R. Mehta, S. Stappert, G.

Dumpich, J. Appl. Phys. 2003, 93, 551.

[16] Y. Liu, E. Koep, M. L. Liu, Chem. Mater. 2005, 17, 3997.

[17] L. Madler, A. Roessler, S. E. Pratsinis, T. Sahm, A. Gurlo, N. Barsan,

U. Weimar, Sens. Actuators B 2006, 114, 283.

[18] M. C. Heine, S. E. Pratsinis, Ind. Eng. Chem. Res. 2005, 44, 6222.

[19] B. E. Russ, J. B. Talbot, J. Adhes. 1999, 68, 257.

[20] C. Bittencourt, E. Llobet, P. Ivanov, X. Correig, X. Vilanova, J.

Brezmes, J. Hubalek, K. Malysz, J. J. Pireaux, J. Calderer, Sens.

Actuators B 2004, 97, 67.

[21] S. K. Andersen, T. Johannessen, M. Mosleh, S. Wedel, J. Tranto, H.

Livbjerg, J. Nanopart. Res. 2002, 4, 405.

[22] A. A. Bolzan, C. Fong, B. J. Kennedy, C. J. Howard, Acta Crystallogr.

Sect. B 1997, 53, 373.

Co. KGaA, Weinheim Adv. Mater. 2008, 20, 3005–3010